[misc]: re-run pre-commit after the 2026-10-07 merges

Today's merges (33d81730..c520878f) left main failing the CI pre-commit
job, while start-of-day main 33d81730 passed every hook. This runs
`pre-commit run --all-files --hook-stage manual` again (Python 3.12, as
in CI) and fixes what it reports. Behaviour is unchanged.

yapf (the hook's in-place reformatting):
- apps/dreamverse/dreamverse/config.py
- apps/dreamverse/dreamverse/mock_server.py
- fastvideo/distributed/device_communicators/ulysses_a2a.py
- fastvideo/entrypoints/video_generator.py
- fastvideo/envs.py
- fastvideo/pipelines/stages/denoising.py
- fastvideo/registry.py
- fastvideo/train/models/hunyuan15/hunyuan15.py
- fastvideo/train/models/ltx2/ltx2.py
- fastvideo/train/utils/dataloader.py
- fastvideo/training/cosmos2_5_training_pipeline.py
- fastvideo/training/ltx2_training_pipeline.py
- fastvideo/training/training_utils.py
- fastvideo/worker/ray_distributed_executor.py
In fastvideo/registry.py the Cosmos 2.5 DFD detector lambda becomes a
small nested _cosmos25_dfd_detector(path) with the same logic, which
avoids yapf's deep hanging layout.

ruff (SIM108; not auto-fixed because of the comment):
- fastvideo/train/utils/performance.py: the if/else that sets
  max_frames is now a conditional expression, with the comment kept
  above it.

codespell (flags "THW" as a misspelling of THE/THAW):
- fastvideo/pipelines/stages/denoising.py: the two Cosmos25 RoPE
  comments now say "(T*H*W, D)".

pymarkdown, actionlint, mypy and check-filenames already passed.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

.agents/skills/add-model-01-prep/scripts/clone_reference_repo.py

@@ -10,9 +10,7 @@ from pathlib import Path
 
 
 def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(
-        description="Clone a reference repo for FastVideo parity tests."
-    )
+    parser = argparse.ArgumentParser(description="Clone a reference repo for FastVideo parity tests.")
     parser.add_argument("repo_url", help="Official reference repository URL")
     parser.add_argument("target_dir", help="Directory to clone into")
     parser.add_argument("--branch", help="Branch or tag to clone")
@@ -62,9 +60,7 @@ def gitignore_entry_for(target: Path) -> str:
     try:
         relative = resolved.relative_to(root)
     except ValueError as exc:
-        raise ValueError(
-            "--update-gitignore requires target_dir to be under the current directory"
-        ) from exc
+        raise ValueError("--update-gitignore requires target_dir to be under the current directory") from exc
 
     text = relative.as_posix().rstrip("/")
     return "/" + text + "/"

.agents/skills/add-model-01-prep/scripts/download_hf_weights.py

@@ -8,14 +8,12 @@ import os
 import sys
 from pathlib import Path
 
-
 HF_TOKEN_ENV_KEYS = ("HF_TOKEN", "HUGGINGFACE_HUB_TOKEN", "HF_API_KEY")
 
 
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
-        description="Download a HF model snapshot or selected files into a local directory."
-    )
+        description="Download a HF model snapshot or selected files into a local directory.")
     parser.add_argument("repo_id", help="HF repo id, for example Org/Model")
     parser.add_argument("local_dir", help="Destination directory")
     parser.add_argument("--repo-type", default="model", help="HF repo type (default: model)")

.agents/skills/add-model-01-prep/scripts/inspect_hf_layout.py

@@ -10,7 +10,6 @@ import sys
 from pathlib import Path
 from typing import Any
 
-
 HF_TOKEN_ENV_KEYS = ("HF_TOKEN", "HUGGINGFACE_HUB_TOKEN", "HF_API_KEY")
 RAW_WEIGHT_SUFFIXES = (".safetensors", ".pt", ".pth", ".ckpt", ".bin")
 KNOWN_COMPONENTS = {
@@ -34,8 +33,7 @@ KNOWN_COMPONENTS = {
 
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
-        description="Classify a HF repo or local directory as Diffusers, raw, custom, or unknown."
-    )
+        description="Classify a HF repo or local directory as Diffusers, raw, custom, or unknown.")
     parser.add_argument("source", help="HF repo id or local weights directory")
     parser.add_argument("--repo-type", default="model", help="HF repo type (default: model)")
     parser.add_argument("--revision", help="HF revision to inspect")
@@ -94,14 +92,12 @@ def load_remote_files(
 ) -> list[str]:
     from huggingface_hub import list_repo_files
 
-    return sorted(
-        list_repo_files(
-            repo_id,
-            repo_type=repo_type,
-            revision=revision,
-            token=token,
-        )
-    )
+    return sorted(list_repo_files(
+        repo_id,
+        repo_type=repo_type,
+        revision=revision,
+        token=token,
+    ))
 
 
 def load_remote_model_index(
@@ -215,24 +211,24 @@ def build_result(args: argparse.Namespace) -> dict[str, Any]:
         "components_seen": components,
         "file_count": len(files),
         "file_scan_truncated": truncated,
-        "files_sample": files[: args.sample_limit],
+        "files_sample": files[:args.sample_limit],
     }
 
 
 def print_human(result: dict[str, Any]) -> None:
     for key in (
-        "source",
-        "source_kind",
-        "repo_type",
-        "revision",
-        "token_env",
-        "source_layout",
-        "needs_conversion",
-        "model_index_class",
-        "model_index_diffusers_version",
-        "model_index_error",
-        "file_count",
-        "file_scan_truncated",
+            "source",
+            "source_kind",
+            "repo_type",
+            "revision",
+            "token_env",
+            "source_layout",
+            "needs_conversion",
+            "model_index_class",
+            "model_index_diffusers_version",
+            "model_index_error",
+            "file_count",
+            "file_scan_truncated",
     ):
         value = result.get(key)
         if value is not None:

.agents/skills/add-model-02-parity/templates/component_parity_test.py

@@ -18,7 +18,6 @@ import pytest
 import torch
 from torch.testing import assert_close
 
-
 os.environ.setdefault("MASTER_ADDR", "localhost")
 os.environ.setdefault("MASTER_PORT", "29519")
 os.environ.setdefault("DISABLE_SP", "1")
@@ -35,15 +34,10 @@ FASTVIDEO_CONFIG_CLASS = "<FastVideoConfig>"  # TODO.
 FASTVIDEO_MODEL_MODULE = "fastvideo.models.<bucket>.<module>"  # TODO.
 FASTVIDEO_MODEL_CLASS = "<FastVideoModel>"  # TODO.
 
-OFFICIAL_REF_DIR = Path(
-    os.getenv("<FAMILY_UPPER>_OFFICIAL_REF_DIR", REPO_ROOT / "<ReferenceDir>")
-)
-LOCAL_WEIGHTS_DIR = Path(
-    os.getenv("<FAMILY_UPPER>_LOCAL_WEIGHTS_DIR", REPO_ROOT / "official_weights" / FAMILY)
-)
-CONVERTED_WEIGHTS_DIR = Path(
-    os.getenv("<FAMILY_UPPER>_CONVERTED_WEIGHTS_DIR", REPO_ROOT / "converted_weights" / FAMILY)
-)
+OFFICIAL_REF_DIR = Path(os.getenv("<FAMILY_UPPER>_OFFICIAL_REF_DIR", REPO_ROOT / "<ReferenceDir>"))
+LOCAL_WEIGHTS_DIR = Path(os.getenv("<FAMILY_UPPER>_LOCAL_WEIGHTS_DIR", REPO_ROOT / "official_weights" / FAMILY))
+CONVERTED_WEIGHTS_DIR = Path(os.getenv("<FAMILY_UPPER>_CONVERTED_WEIGHTS_DIR",
+                                       REPO_ROOT / "converted_weights" / FAMILY))
 
 
 def _resolve_hf_token() -> str | None:
@@ -99,18 +93,14 @@ def _load_official_model(device: torch.device, dtype: torch.dtype) -> torch.nn.M
     model = OfficialClass()  # TODO: pass official config kwargs.
     state_dict = {}  # TODO: load official state dict from LOCAL_WEIGHTS_DIR.
     missing, unexpected = model.load_state_dict(state_dict, strict=True)
-    assert not missing and not unexpected, (
-        f"official load mismatch missing={missing[:5]} unexpected={unexpected[:5]}"
-    )
+    assert not missing and not unexpected, (f"official load mismatch missing={missing[:5]} unexpected={unexpected[:5]}")
     return model.to(device=device, dtype=dtype).eval()
 
 
 def _load_fastvideo_model(device: torch.device, dtype: torch.dtype) -> torch.nn.Module:
     """Load the FastVideo component with the same tensor content."""
     if not CONVERTED_WEIGHTS_DIR.exists() and not LOCAL_WEIGHTS_DIR.exists():
-        pytest.skip(
-            f"No FastVideo loadable weights: {CONVERTED_WEIGHTS_DIR} or {LOCAL_WEIGHTS_DIR}"
-        )
+        pytest.skip(f"No FastVideo loadable weights: {CONVERTED_WEIGHTS_DIR} or {LOCAL_WEIGHTS_DIR}")
 
     # TODO: replace with the bucket-specific FastVideo config/class/loader.
     # DiT examples:
@@ -127,8 +117,7 @@ def _load_fastvideo_model(device: torch.device, dtype: torch.dtype) -> torch.nn.
     state_dict = {}  # TODO: load converted or directly mapped state dict.
     missing, unexpected = model.load_state_dict(state_dict, strict=True)
     assert not missing and not unexpected, (
-        f"FastVideo load mismatch missing={missing[:5]} unexpected={unexpected[:5]}"
-    )
+        f"FastVideo load mismatch missing={missing[:5]} unexpected={unexpected[:5]}")
     return model.to(device=device, dtype=dtype).eval()
 
 
@@ -187,11 +176,9 @@ def test_component_parity():
 
     assert official_out.shape == fastvideo_out.shape
     diff = (official_out - fastvideo_out).abs()
-    print(
-        f"official abs_mean={official_out.abs().mean().item():.6f} "
-        f"fastvideo abs_mean={fastvideo_out.abs().mean().item():.6f} "
-        f"diff_max={diff.max().item():.6f} diff_mean={diff.mean().item():.6f}"
-    )
+    print(f"official abs_mean={official_out.abs().mean().item():.6f} "
+          f"fastvideo abs_mean={fastvideo_out.abs().mean().item():.6f} "
+          f"diff_max={diff.max().item():.6f} diff_mean={diff.mean().item():.6f}")
 
     # TODO: pick tolerance by scope:
     # - single block / same kernel: 1e-4

.agents/skills/add-model-07-conversion/templates/family_to_diffusers.py

@@ -27,7 +27,6 @@ try:
 except ImportError:  # pragma: no cover - optional local conversion dependency
     snapshot_download = None
 
-
 # TODO: fill with authoritative component prefixes for monolithic checkpoints.
 # Example: {"model.model.": "transformer", "pretransform.model.": "vae"}
 COMPONENT_PREFIXES: dict[str, str] = {}
@@ -47,10 +46,7 @@ SKIP_PATTERNS: tuple[str, ...] = ()
 
 
 def _hf_token() -> str | None:
-    return (
-        os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACE_HUB_TOKEN")
-        or os.environ.get("HF_API_KEY")
-    )
+    return (os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACE_HUB_TOKEN") or os.environ.get("HF_API_KEY"))
 
 
 def resolve_src(src: str, revision: str | None) -> Path:
@@ -95,11 +91,10 @@ def apply_mapping(key: str) -> str | None:
     return key
 
 
-def split_monolithic(
-    state: dict[str, torch.Tensor],
-) -> dict[str, OrderedDict[str, torch.Tensor]]:
+def split_monolithic(state: dict[str, torch.Tensor], ) -> dict[str, OrderedDict[str, torch.Tensor]]:
     components: dict[str, OrderedDict[str, torch.Tensor]] = {
-        name: OrderedDict() for name in set(COMPONENT_PREFIXES.values())
+        name: OrderedDict()
+        for name in set(COMPONENT_PREFIXES.values())
     }
     intentionally_skipped: list[str] = []
     unowned: list[str] = []
@@ -117,10 +112,8 @@ def split_monolithic(
             unowned.append(key)
     if unowned:
         sample = ", ".join(unowned[:10])
-        raise ValueError(
-            f"Unowned monolithic keys: {len(unowned)}. "
-            f"Add COMPONENT_PREFIXES or SKIP_PATTERNS entries. Sample: {sample}"
-        )
+        raise ValueError(f"Unowned monolithic keys: {len(unowned)}. "
+                         f"Add COMPONENT_PREFIXES or SKIP_PATTERNS entries. Sample: {sample}")
     if intentionally_skipped:
         print(f"Intentionally skipped {len(intentionally_skipped)} keys")
     return {name: weights for name, weights in components.items() if weights}
@@ -143,8 +136,12 @@ def build_component_configs(_src_dir: Path) -> dict[str, dict[str, Any]]:
     # TODO: emit config content accepted by FastVideo loaders. Most components use
     # config.json; schedulers use scheduler_config.json.
     return {
-        "transformer": {"_class_name": "<FastVideoTransformerClass>"},
-        "vae": {"_class_name": "<FastVideoVAEClass>"},
+        "transformer": {
+            "_class_name": "<FastVideoTransformerClass>"
+        },
+        "vae": {
+            "_class_name": "<FastVideoVAEClass>"
+        },
     }
 
 
@@ -177,19 +174,13 @@ def build_model_index(
     }
     if revision:
         index["_fastvideo_converted_revision"] = revision
-    return {
-        key: value
-        for key, value in index.items()
-        if key.startswith("_") or key in available_components
-    }
+    return {key: value for key, value in index.items() if key.startswith("_") or key in available_components}
 
 
 def validate_component_configs(configs: dict[str, dict[str, Any]]) -> None:
     # TODO: instantiate each FastVideo config and call update_model_arch(...) or
     # update_model_config(...) with this JSON so unknown emitted keys fail here.
-    placeholder_configs = [
-        name for name, config in configs.items() if "<" in json.dumps(config)
-    ]
+    placeholder_configs = [name for name, config in configs.items() if "<" in json.dumps(config)]
     if placeholder_configs:
         raise ValueError(f"Replace config placeholders for: {placeholder_configs}")
 
@@ -201,9 +192,7 @@ def verify_conversion(
     del dst_dir, components
     # TODO: load each emitted stateful component through its production loader and
     # assert strict load, or document exact allowed missing/unexpected keys.
-    raise NotImplementedError(
-        "Implement production config validation and strict-load checks"
-    )
+    raise NotImplementedError("Implement production config validation and strict-load checks")
 
 
 def write_component(
@@ -216,9 +205,7 @@ def write_component(
     if component_dir.exists() and any(component_dir.iterdir()):
         shutil.rmtree(component_dir)
     component_dir.mkdir(parents=True, exist_ok=True)
-    save_file(
-        dict(state), str(component_dir / "diffusion_pytorch_model.safetensors")
-    )
+    save_file(dict(state), str(component_dir / "diffusion_pytorch_model.safetensors"))
     if config is not None:
         config_path = component_dir / config_filename(name)
         with config_path.open("w", encoding="utf-8") as f:
@@ -261,9 +248,7 @@ def convert(
 
     if layout in {"monolithic", "raw_official"}:
         # TODO: replace model.safetensors with the official monolithic file name.
-        components = split_monolithic(
-            load_checkpoint(default_monolithic_checkpoint(src_path))
-        )
+        components = split_monolithic(load_checkpoint(default_monolithic_checkpoint(src_path)))
     elif layout in {"separate_components", "mixed"}:
         if not src_path.is_dir():
             raise ValueError(f"{layout} layout requires a source directory: {src_path}")
@@ -271,9 +256,7 @@ def convert(
     else:
         raise ValueError(f"Unsupported template layout: {layout}")
 
-    copied = (
-        copy_passthrough(src_path, dst_dir) if src_path.is_dir() else []
-    )
+    copied = (copy_passthrough(src_path, dst_dir) if src_path.is_dir() else [])
     configs = build_component_configs(src_path if src_path.is_dir() else src_path.parent)
     validate_component_configs(configs)
     for name, state in components.items():
@@ -289,9 +272,7 @@ def convert(
 
 def main() -> None:
     parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument(
-        "--src", required=True, help="HF repo id, local dir, or checkpoint path"
-    )
+    parser.add_argument("--src", required=True, help="HF repo id, local dir, or checkpoint path")
     parser.add_argument("--revision", help="HF branch, tag, or commit for repo sources")
     parser.add_argument(
         "--dst",

.agents/skills/add-model-08-trace/templates/block_trace_debug.py

@@ -24,8 +24,8 @@ from typing import Any
 
 import torch
 
-FAMILY: str = "<family>"       # e.g. "magi_human", "ltx2", "wan"
-COMPONENT: str = "<component>" # e.g. "dit", "vae", "encoder"
+FAMILY: str = "<family>"  # e.g. "magi_human", "ltx2", "wan"
+COMPONENT: str = "<component>"  # e.g. "dit", "vae", "encoder"
 DRILL_LAYER_ENV: str = "<FAMILY>_DEBUG_DRILL_LAYER"
 HYPOTHESIS_ENV: str = "<FAMILY>_DEBUG_PATCH_<HYPOTHESIS>"
 REL_THRESHOLD: float = 0.005  # 0.5% abs_mean drift flags a block as divergent
@@ -94,6 +94,7 @@ def _attach_block_hooks(
     handles: list[Any] = []
 
     def _hook(name: str):
+
         def fn(_module, _inputs, outputs):
             t = outputs[0] if isinstance(outputs, tuple) else outputs
             if not torch.is_tensor(t):
@@ -101,6 +102,7 @@ def _attach_block_hooks(
             log.append({"side": label, **_stat(name, t)})
             if tensors is not None:
                 tensors[name] = t.detach().float().cpu()
+
         return fn
 
     def _pre_hook(name: str):
@@ -114,6 +116,7 @@ def _attach_block_hooks(
             log.append({"side": label, **_stat(key, t)})
             if tensors is not None:
                 tensors[key] = t.detach().float().cpu()
+
         return fn
 
     # TODO: adapt attribute paths to your model. Remove adapter block if absent.
@@ -131,43 +134,21 @@ def _attach_block_hooks(
             # magi-human uses: attention, mlp.pre_norm, mlp.up_gate_proj,
             # mlp.down_proj (pre+post), mlp, attn_post_norm, mlp_post_norm.
             if hasattr(layer, "attention"):
-                handles.append(
-                    layer.attention.register_forward_hook(_hook(f"{tag}.attention"))
-                )
+                handles.append(layer.attention.register_forward_hook(_hook(f"{tag}.attention")))
             if hasattr(layer, "mlp"):
                 mlp = layer.mlp
                 if hasattr(mlp, "pre_norm"):
-                    handles.append(
-                        mlp.pre_norm.register_forward_hook(_hook(f"{tag}.mlp.pre_norm"))
-                    )
+                    handles.append(mlp.pre_norm.register_forward_hook(_hook(f"{tag}.mlp.pre_norm")))
                 if hasattr(mlp, "up_gate_proj"):
-                    handles.append(
-                        mlp.up_gate_proj.register_forward_hook(
-                            _hook(f"{tag}.mlp.up_gate_proj")
-                        )
-                    )
+                    handles.append(mlp.up_gate_proj.register_forward_hook(_hook(f"{tag}.mlp.up_gate_proj")))
                 if hasattr(mlp, "down_proj"):
-                    handles.append(
-                        mlp.down_proj.register_forward_pre_hook(
-                            _pre_hook(f"{tag}.mlp.down_proj")
-                        )
-                    )
-                    handles.append(
-                        mlp.down_proj.register_forward_hook(_hook(f"{tag}.mlp.down_proj"))
-                    )
+                    handles.append(mlp.down_proj.register_forward_pre_hook(_pre_hook(f"{tag}.mlp.down_proj")))
+                    handles.append(mlp.down_proj.register_forward_hook(_hook(f"{tag}.mlp.down_proj")))
                 handles.append(mlp.register_forward_hook(_hook(f"{tag}.mlp")))
             if hasattr(layer, "attn_post_norm"):
-                handles.append(
-                    layer.attn_post_norm.register_forward_hook(
-                        _hook(f"{tag}.attn_post_norm")
-                    )
-                )
+                handles.append(layer.attn_post_norm.register_forward_hook(_hook(f"{tag}.attn_post_norm")))
             if hasattr(layer, "mlp_post_norm"):
-                handles.append(
-                    layer.mlp_post_norm.register_forward_hook(
-                        _hook(f"{tag}.mlp_post_norm")
-                    )
-                )
+                handles.append(layer.mlp_post_norm.register_forward_hook(_hook(f"{tag}.mlp_post_norm")))
     return handles
 
 
@@ -193,11 +174,9 @@ def _write_log(entries: list[dict], path: Path) -> None:
     path.parent.mkdir(parents=True, exist_ok=True)
     with open(path, "w") as f:
         for e in entries:
-            f.write(
-                f"{e['name']} {e['shape']} "
-                f"{e['abs_mean']:.8f} {e['sum']:.4f} "
-                f"{e['min']:.6f} {e['max']:.6f}\n"
-            )
+            f.write(f"{e['name']} {e['shape']} "
+                    f"{e['abs_mean']:.8f} {e['sum']:.4f} "
+                    f"{e['min']:.6f} {e['max']:.6f}\n")
 
 
 def _sort_key(name: str, drill_layer: int) -> tuple:
@@ -205,9 +184,14 @@ def _sort_key(name: str, drill_layer: int) -> tuple:
         return (0, "")
     if name.startswith(f"L{drill_layer:02d}."):
         sub_order = {
-            "attention": 0, "attn_post_norm": 1, "mlp.pre_norm": 2,
-            "mlp.up_gate_proj": 3, "mlp.down_proj<in>": 4,
-            "mlp.down_proj": 5, "mlp": 6, "mlp_post_norm": 7,
+            "attention": 0,
+            "attn_post_norm": 1,
+            "mlp.pre_norm": 2,
+            "mlp.up_gate_proj": 3,
+            "mlp.down_proj<in>": 4,
+            "mlp.down_proj": 5,
+            "mlp": 6,
+            "mlp_post_norm": 7,
         }.get(name.split(".", 1)[1], 9)
         return (1, f"block[{drill_layer:02d}]", sub_order)
     if name.startswith("block["):
@@ -216,10 +200,8 @@ def _sort_key(name: str, drill_layer: int) -> tuple:
 
 
 def _print_table(by_name: dict[str, dict], drill_layer: int) -> int | None:
-    hdr = (
-        f"{'name':<18} {'up_shape':<22} {'up_absmean':>12} {'fv_absmean':>12} "
-        f"{'absmean_diff':>14} {'rel%':>8} {'up_sum':>14} {'fv_sum':>14} {'sum_diff':>12}"
-    )
+    hdr = (f"{'name':<18} {'up_shape':<22} {'up_absmean':>12} {'fv_absmean':>12} "
+           f"{'absmean_diff':>14} {'rel%':>8} {'up_sum':>14} {'fv_sum':>14} {'sum_diff':>12}")
     print(f"\n{hdr}\n{'-' * len(hdr)}")
     first_div: int | None = None
     for name in sorted(by_name.keys(), key=lambda n: _sort_key(n, drill_layer)):
@@ -235,11 +217,9 @@ def _print_table(by_name: dict[str, dict], drill_layer: int) -> int | None:
             flag = " <<< DIVERGE"
             if first_div is None:
                 first_div = int(name[len("block["):-1])
-        print(
-            f"{name:<18} {str(up['shape']):<22} {up['abs_mean']:>12.6f} "
-            f"{fv['abs_mean']:>12.6f} {am_diff:>14.6f} {am_rel * 100:>7.3f}% "
-            f"{up['sum']:>14.4f} {fv['sum']:>14.4f} {sum_diff:>12.4f}{flag}"
-        )
+        print(f"{name:<18} {str(up['shape']):<22} {up['abs_mean']:>12.6f} "
+              f"{fv['abs_mean']:>12.6f} {am_diff:>14.6f} {am_rel * 100:>7.3f}% "
+              f"{up['sum']:>14.4f} {fv['sum']:>14.4f} {sum_diff:>12.4f}{flag}")
     return first_div
 
 
@@ -255,10 +235,8 @@ def _print_elementwise(up_t: dict[str, torch.Tensor], fv_t: dict[str, torch.Tens
             continue
         diff = (a - b).abs()
         rel = (diff.mean().item() / max(a.abs().mean().item(), 1e-9)) * 100
-        print(
-            f"{name:<30} {str(tuple(a.shape)):<22} "
-            f"{diff.max().item():>12.6f} {diff.mean().item():>12.6f} {rel:>9.4f}%"
-        )
+        print(f"{name:<30} {str(tuple(a.shape)):<22} "
+              f"{diff.max().item():>12.6f} {diff.mean().item():>12.6f} {rel:>9.4f}%")
 
 
 def main() -> None:

.agents/skills/add-model-09-pipeline/templates/pipeline_parity_test.py

@@ -43,12 +43,10 @@ def _add_official_to_path() -> Path:
 
 def _log_tensor_stats(label: str, tensor: torch.Tensor) -> None:
     value = tensor.detach().float()
-    print(
-        f"[{_MODEL_FAMILY} PIPELINE] {label}: shape={tuple(tensor.shape)} "
-        f"dtype={tensor.dtype} device={tensor.device} "
-        f"min={value.min().item():.6f} max={value.max().item():.6f} "
-        f"mean={value.mean().item():.6f} std={value.std().item():.6f}"
-    )
+    print(f"[{_MODEL_FAMILY} PIPELINE] {label}: shape={tuple(tensor.shape)} "
+          f"dtype={tensor.dtype} device={tensor.device} "
+          f"min={value.min().item():.6f} max={value.max().item():.6f} "
+          f"mean={value.mean().item():.6f} std={value.std().item():.6f}")
 
 
 def _extract_tensor(output: Any, key: str) -> torch.Tensor:
@@ -73,10 +71,8 @@ def _run_official_pipeline(
     device: torch.device,
 ) -> Any:
     del official_path, params, device
-    pytest.skip(
-        "TODO: import the official pipeline/factory, load official weights, "
-        "run with params, and return the comparison target."
-    )
+    pytest.skip("TODO: import the official pipeline/factory, load official weights, "
+                "run with params, and return the comparison target.")
 
 
 def _run_fastvideo_pipeline(model_path: Path, params: dict[str, Any]) -> Any:
@@ -146,8 +142,6 @@ def test_todo_model_family_pipeline_official_parity() -> None:
     assert official_tensor.shape == fastvideo_tensor.shape
 
     diff = (official_tensor - fastvideo_tensor).abs()
-    print(
-        f"diff max={diff.max().item():.6f} "
-        f"mean={diff.mean().item():.6f} median={diff.median().item():.6f}"
-    )
+    print(f"diff max={diff.max().item():.6f} "
+          f"mean={diff.mean().item():.6f} median={diff.median().item():.6f}")
     assert_close(fastvideo_tensor, official_tensor, atol=1e-2, rtol=1e-2)

.agents/skills/ci-runner/SKILL.md

@@ -0,0 +1,99 @@
+---
+name: ci-runner
+description: Work on FastVideo's Slurm-only, change-aware GPU CI lanes, static Buildkite graph, trusted ci-runner policy, lane scripts, and GB200 validation.
+---
+
+# Slinky Slurm CI lanes
+
+FastVideo's `ci-runner` Buildkite queue is the control plane for all active
+GPU CI. A private host-owned dispatcher leases GPUs from the Slinky Slurm tray
+and runs the immutable PR SHA inside an isolated Enroot container. Buildkite
+pipeline upload and Slurm submission occur on the login plane; every test
+payload executes on Slurm compute.
+
+The files under `fastvideo/tests/modal/` and `.buildkite/scripts/pr_test.sh`
+are dormant rollback code. Never add an active Buildkite or slash-command
+route to them. `pr_test.sh` must continue to reject Buildkite invocations.
+
+The private operator bundle is deliberately outside this repository because
+it contains site paths and credentials. See
+`docs/contributing/ci_architecture.md`; this skill covers the repository half
+and the coordination contract with that bundle.
+
+## Invariants
+
+- `.buildkite/pipeline.yml` contains exactly one static step for every active
+  GPU lane. Each step pins a unique key and label, a 90-minute timeout, the
+  trusted `/opt/fastvideo-ci-runner/run-ci` command (`run-unit` is the one
+  compatibility wrapper), step-level internal `TEST_TYPE`, and
+  `queue: "ci-runner"`.
+- Active CI contains no `pr_test.sh` command, Modal invocation, default queue,
+  Buildkite plugin, `soft_fail`, or job-controlled artifact glob.
+- The six Fastcheck lanes use `:microscope:` labels. Full-Suite-only lanes use
+  `:test_tube:` or `:bar_chart:` so direct reruns update the right aggregate.
+- SSIM and vanilla training request all four GPUs. Keep both in the
+  `fastvideo/slinky/whole-tray` Buildkite concurrency group with a limit of one
+  so the second job does not consume an agent or command timeout while waiting
+  for the same tray.
+- `/test full` schedules all twenty lanes. `/merge`, `ready`, and new pushes to
+  ready PRs use the trusted base-branch planner in
+  `.github/scripts/plan_merge_ci.py`: automatic Fastcheck remains the universal
+  six-lane baseline, and the merge build adds only path-relevant integration
+  lanes. Unknown source/build paths fail closed to all fourteen additive lanes.
+  The trusted uploader still normalizes and validates the complete static graph
+  before Buildkite evaluates its plan conditions.
+- Focused merge builds may pass allowlisted golden-gate and SSIM test basenames.
+  The private host validates the lane plan and basenames before staging them,
+  and the in-container scripts validate them again. Direct `/test ssim`,
+  explicit `/test full`, and the weekly main-branch schedule run the complete
+  SSIM matrix.
+- The trusted uploader serves exactly three entry pipelines:
+  `pr-fastcheck` for automatic PR builds, `ci` for slash-command/ready-label
+  API builds, and `fastvideo-performance-lane` for the weekly schedule. Keep
+  incoming GitHub webhook processing disabled on `ci` so it cannot duplicate
+  `pr-fastcheck` on every PR update.
+- Test payloads live in `.buildkite/scripts/unit_test.sh` or executable
+  `.buildkite/scripts/lanes/<lane>.sh`. Backend policy (GPU count, extras,
+  secrets, kernel build, artifacts) stays in the agent-owned lane table.
+- Tests must preserve an inherited `MASTER_PORT`. Packed containers share the
+  tray network namespace, so the private runner assigns a distinct port range
+  per GPU lease and the SSIM scheduler assigns task offsets within its range.
+- The ARM64 runner image includes the pinned FA4 CuTe overlay validated on
+  GB200. Keep SSIM at `FASTVIDEO_FA4=1` because its references were seeded with
+  FA4; keep lanes with FA2 baselines at `FASTVIDEO_FA4=0`. A runner image change
+  must revalidate both the FA4 import and an actual GB200 forward kernel.
+- `fastvideo/tests/ssim/ci_runner.py` is the active four-GPU SSIM scheduler.
+  New SSIM files are discovered through `REQUIRED_GPUS` and
+  `*_MODEL_TO_PARAMS`; do not wire them through the dormant Modal scheduler.
+- The host policy fail-closes unknown tuples. A repository-side lane change is
+  inert until the operator updates the private lane table and uploader policy
+  in the same rollout.
+
+## Adding or changing a lane
+
+1. Read the closest `AGENTS.md` and the domain-specific testing guide.
+2. Add or update the executable lane payload under `.buildkite/scripts/`.
+   Keep it deterministic and free of host-specific paths or credential fetches.
+3. Add the static pipeline step and canonical `/test <name>` mapping. Keep the
+   `<name>-ci` alias only when compatibility requires it.
+4. Add its source/test path ownership to `.github/scripts/plan_merge_ci.py`.
+   Prefer the narrowest correctness-preserving lane set; leave unknown paths
+   fail-closed. Extend `fastvideo/tests/contract/test_ci_test_collection.py`,
+   `test_merge_ci_plan.py`, and focused CPU-only scheduler/policy tests.
+5. Coordinate the private lane row: GPU count (1-4), wall time, script, scope
+   pairs, step key, command, HF cache/token, tracking mode, extras, attention
+   backend policy, kernel policy, and artifact relay. Active training lanes
+   keep W&B offline and do not stage a W&B credential.
+6. Update the trusted pipeline-uploader schema. A mismatch must reject the
+   pipeline rather than silently skip a lane.
+7. Run `pre-commit run --files <changed paths>`, the planner's representative
+   diff matrix, contract tests, private driver tests, and a real GB200 canary.
+   Multi-GPU, hardware-reference, training, performance, and SSIM changes need
+   their own target-hardware evidence.
+
+## Rollback
+
+Rollback the Slurm routing/configuration change or pause the `ci-runner` queue.
+Do not silently reactivate Modal. A manual Modal experiment requires the
+explicit local opt-in documented in `ci_architecture.md`; returning it to
+production CI needs a separate reviewed decision.

.agents/skills/env-var-conventions/SKILL.md

@@ -0,0 +1,79 @@
+---
+name: env-var-conventions
+description: Add, read, rename, or remove an environment variable in FastVideo, or change the environment-variable policy. Use before touching fastvideo/envs.py, os.environ, os.getenv, or monkeypatch.setenv in fastvideo/, and when fastvideo/tests/contract/test_env_policy.py fails.
+---
+
+# Environment Variable Conventions
+
+## Purpose
+
+FastVideo registers its environment variables as typed fields in
+`fastvideo/envs.py`. The policy that governs them is
+`docs/contributing/env_vars.md`, and the contract test
+`fastvideo/tests/contract/test_env_policy.py` enforces the policy in the unit
+CI lane. This skill routes an environment-variable change through that policy.
+The policy doc is the single source of the rules; read it instead of relying
+on a summary here.
+
+## Prerequisites
+
+- Read `docs/contributing/env_vars.md` in full.
+- Decide whether the setting belongs in an environment variable or an argument
+  (rule 5 in the policy doc). Settings that users change per deployment are
+  arguments; add them through `fastvideo/fastvideo_args.py` instead.
+
+## Inputs
+
+| Parameter  | Required | Description                                                    |
+| ---------- | -------- | -------------------------------------------------------------- |
+| `change`   | Yes      | Add, read, rename, or remove a variable, or change the policy. |
+| `variable` | Yes      | The variable name, with the `FASTVIDEO_` prefix.               |
+
+## Steps
+
+1. **Declare or edit the variable in `fastvideo/envs.py`.**
+   - Pick the field type and category that the policy doc lists.
+   - Write a description that states what the variable does and its units.
+   - To rename, keep the old name in `deprecated_names`. To remove, add the
+     name to `DEPRECATED_VARIABLES`. Update the uses in `examples/`,
+     `scripts/`, `docs/`, `apps/`, and the tests.
+2. **Read the variable with `envs.NAME.get()` inside a function.**
+   - In tests, change the value with `envs.NAME.override(value)`, and a variable
+     outside the registry with `envs.override_external(name, value)`; the
+     `env_overrides` fixture keeps either until the end of the test.
+   - Name a variable that only tests read `FASTVIDEO_TEST_*`.
+   - Do not call `os.environ`, `os.getenv`, or `monkeypatch.setenv` for a
+     FastVideo variable.
+   - To set a variable that another tool reads, call `envs.set_external`,
+     `envs.setdefault_external`, or `envs.unset_external`.
+3. **Regenerate the table in the policy doc.**
+   - Run `python fastvideo/tests/contract/test_env_policy.py`.
+4. **Run the contract test.**
+   - Run `pytest fastvideo/tests/contract/test_env_policy.py`.
+   - When the test reports a fixed known violation, delete or lower its entry
+     in `KNOWN_VIOLATIONS`. Never add an entry to `KNOWN_VIOLATIONS`.
+5. **When the policy itself changes, update the policy doc and the contract
+   test in the same pull request.**
+   - The rules in `docs/contributing/env_vars.md`, the checks and allowlist in
+     `fastvideo/tests/contract/test_env_policy.py`, and this skill must agree.
+
+## Outputs
+
+- A registry entry in `fastvideo/envs.py` and call sites that use
+  `envs.NAME.get()`.
+- A regenerated table in `docs/contributing/env_vars.md`.
+- A passing `fastvideo/tests/contract/test_env_policy.py`.
+
+## Example Usage
+
+```
+Add a FASTVIDEO_DEBUG_MY_STAGE switch that logs MyStage inputs.
+```
+
+## References
+
+- `docs/contributing/env_vars.md`: the policy, the field types, and the
+  violation kinds that the contract test reports.
+- `fastvideo/envs.py`: the registry.
+- `fastvideo/tests/contract/test_env_policy.py`: the contract test,
+  `EXTERNAL_ALLOWLIST`, and `KNOWN_VIOLATIONS`.

.agents/skills/reseed-performance-baseline/SKILL.md

@@ -1,20 +1,23 @@
 ---
 name: reseed-performance-baseline
-description: Re-seed the HF performance-tracking baseline for an intentional runtime, dependency, or environment-caused benchmark shift using one or more reviewed normalized performance JSONs. Use when performance CI fails because metrics such as latency, throughput, component time, or peak memory changed for an accepted reason and the rolling median baseline in FastVideo/performance-tracking must be advanced from a consistent batch of reviewed source results. The workflow backs up existing history under /tmp, validates all source JSONs for the same (model_id, gpu_type), rejects internally inconsistent source batches, uploads one success=true reseed record per accepted source JSON, and offers to clean local temp state after a successful upload.
+description: Re-seed the HF performance-tracking baseline for an intentional runtime, dependency, environment-caused benchmark shift, or reviewed v2 calibration using one or more reviewed normalized performance JSONs. Use when performance CI fails because metrics such as latency, throughput, component time, or peak memory changed for an accepted reason and the rolling median baseline in FastVideo/performance-tracking must be advanced, or when a new v2 exact comparable identity needs its first approved baseline. The workflow backs up existing history under /tmp, validates all source JSONs for the same legacy (model_id, gpu_type) target or the same v2 exact identity, rejects internally inconsistent source batches, uploads one success=true baseline record per accepted source JSON, and offers to clean local temp state after a successful upload.
 ---
 
 # Re-seed Performance Baseline
 
 ## Purpose
 
-Replace or advance the rolling performance baseline for a single
-`(model_id, gpu_type)` pair in the HF dataset
-`FastVideo/performance-tracking`.
+Replace or advance the rolling performance baseline in the HF dataset
+`FastVideo/performance-tracking`. Legacy targets are scoped by
+`(model_id, gpu_type)`. V2 targets are scoped by exact comparable identity:
+`workload_id`, `variant_id`, `benchmark_version`, `hardware_profile_id`,
+`software_profile_id`, and `recipe_fingerprint`.
 
-Performance comparison uses the median of up to the last 5 successful records
-for the same model and GPU. Failed records are useful audit history, but they
-do not move the future baseline because `compare_baseline.py` loads records
-with `successful_only=True`.
+Performance comparison uses the median of up to the last 5 successful,
+baseline-eligible records for the same target. Failed or calibration-only
+records are useful audit history, but they do not move the future baseline
+because `compare_baseline.py` loads records with `successful_only=True` and
+`baseline_eligible_only=True`.
 
 This skill now reseeds from a reviewed batch of one or more source performance
 JSONs. It uploads one new `success=true` record per accepted source JSON; it
@@ -22,11 +25,13 @@ does not blindly replicate one measurement into 3 or 5 records. The effective
 reseed size is therefore dynamic and equals the number of provided, validated,
 internally consistent source JSONs.
 
-If the operator provides fewer than 3 records, call out that the last-5 rolling
-median may not move immediately. If the operator provides 3 consistent shifted
-records, the rolling median usually moves immediately. If the operator provides
-5 consistent shifted records, the last-5 window is effectively reset to the new
-runtime profile.
+For baseline shifts with existing history, if the operator provides fewer than
+3 records, call out that the last-5 rolling median may not move immediately. If
+the operator provides 3 consistent shifted records, the rolling median usually
+moves immediately. If the operator provides 5 consistent shifted records, the
+last-5 window is effectively reset to the new runtime profile. For the first
+approved v2 baseline of a new exact identity, one reviewed calibration seed is
+enough for the next comparable run to leave `CALIBRATION_NEEDED`.
 
 These records are intentional operator-approved baseline resets, not ordinary
 independent main-branch persistence. Mark them clearly with provenance fields
@@ -66,8 +71,8 @@ approval, then upload reviewed accepted baseline records.
 
 | Parameter | Required | Description |
 |-----------|----------|-------------|
-| `model_id` | Yes | Benchmark id, e.g. `wan-t2v-1.3b-2gpu`. This maps to the HF subdirectory after `sanitize(model_id)`. |
-| `gpu_type` | Yes | Exact GPU device string from the performance record, e.g. the L40S device name emitted by CI. Baselines are GPU-specific. |
+| `model_id` | Legacy required; v2 inferred | Benchmark id, e.g. `wan-t2v-1.3b-2gpu`. This maps to the HF subdirectory after `sanitize(model_id)`. For v2 records, use the `model_id` from each source artifact only as the upload directory; comparison is by exact identity. |
+| `gpu_type` | Legacy required; v2 inferred | Exact GPU device string from the performance record, e.g. the L40S device name emitted by CI. V2 hardware matching uses `hardware_profile_id`; preserve `gpu_type` as display metadata. |
 | `source_results` | Yes | One or more local paths or Buildkite artifact URLs for accepted shifted performance JSONs. Prefer normalized `normalized_perf_*.json` artifacts emitted by `compare_baseline.py`. Accept `source_result` as an alias only for a single JSON. |
 | `max_intra_batch_regression` | No | Maximum allowed regression of any source JSON against the source batch median. Default: `0.05` (5%). |
 | `intent_rationale` | Yes | One-line explanation for why the baseline shift is legitimate. This is written into provenance and should be reused in the PR. |
@@ -78,10 +83,14 @@ Hardcoded defaults:
   supported by the code, but use the default unless the user explicitly asks).
 - Local sync root: `/tmp/perf-tracking` (`PERFORMANCE_TRACKING_ROOT` override
   is supported).
+- Prepared-record staging root: `/tmp/performance_reseed_prepared`
+  (`PERFORMANCE_RESEED_STAGING_ROOT` override is supported). Keep it separate
+  and non-nested from the sync root.
 - Backup root: `/tmp/performance_reseed_backup`.
 - Download scratch root for source artifact URLs: `/tmp/performance_reseed_source`.
-- Baseline window: last 5 `success=true` records for the same
-  `(model_id, gpu_type)`.
+- Baseline window: last 5 `success=true`, `baseline_eligible=true` records
+  for the same legacy `(model_id, gpu_type)` target or the same v2 exact
+  comparable identity.
 - Reseed count: dynamic. Upload exactly one accepted seed record per validated
   source JSON.
 
@@ -115,12 +124,24 @@ with open(source_result, encoding="utf-8") as f:
     record = json.load(f)
 ```
 
-Stop if any normalized record's `model_id` or `gpu_type` does not match the
-requested `model_id` and `gpu_type`.
+Classify the source batch before continuing:
+
+- **Legacy source records** have no v2 exact identity fields. Stop if any
+  normalized record's `model_id` or `gpu_type` does not match the requested
+  `model_id` and `gpu_type`.
+- **V2 source records** have exact identity fields. Stop unless every source
+  record has all six comparable identity fields and they are identical across
+  the batch: `workload_id`, `variant_id`, `benchmark_version`,
+  `hardware_profile_id`, `software_profile_id`, and `recipe_fingerprint`.
+  Do not fall back to legacy `(model_id, gpu_type)` matching for v2 records.
 
 The source records may have `success: false` when they came from failed
 rolling baseline comparisons. That is expected; only the reviewed reseed
 records become new `success: true` baseline records after explicit approval.
+For a first v2 baseline seed, the source records must instead be successful
+scheduled-main full-suite `CALIBRATION_NEEDED` normalized artifacts. Reject PR,
+local, direct-run, non-main-branch, or non-full-suite calibration artifacts as
+seed sources.
 
 Sort validated source records by their original `timestamp` ascending before
 preparing the seed records. If a source timestamp is missing or unparsable,
@@ -194,7 +215,7 @@ export HF_REPO_ID="${HF_REPO_ID:-FastVideo/performance-tracking}"
 python -c 'from fastvideo.performance.hf_store import sync_from_hf; import os; sync_from_hf(os.environ["PERFORMANCE_TRACKING_ROOT"], strict=True)'
 ```
 
-Then back up only the sanitized model directory under `/tmp`:
+For legacy records, back up the sanitized model directory under `/tmp`:
 
 ```bash
 SHORT_COMMIT=$(git rev-parse --short=12 HEAD)
@@ -209,6 +230,16 @@ mkdir -p "$BACKUP_DIR"
 cp -R "${PERFORMANCE_TRACKING_ROOT}/${MODEL_SAFE}" "$BACKUP_DIR/" 2>/dev/null || true
 ```
 
+For v2 records, back up the full local tracking root after sync. Exact identity
+lookup scans across model directories, so a benchmark rename may have relevant
+history outside the current source artifact's `model_id` directory:
+
+```bash
+BACKUP_DIR="/tmp/performance_reseed_backup/${TIMESTAMP}_${SHORT_COMMIT}_v2_exact_identity"
+mkdir -p "$BACKUP_DIR"
+cp -R "${PERFORMANCE_TRACKING_ROOT}" "$BACKUP_DIR/tracking-root"
+```
+
 Write provenance next to the backup:
 
 ```bash
@@ -231,7 +262,9 @@ first baseline seed. Continue, but report that baseline history was empty.
 
 ### 3. Compute old baseline and candidate shift
 
-Load the last 5 successful records for the target:
+Load the last 5 successful baseline records for the target.
+
+For legacy targets:
 
 ```python
 from fastvideo.performance.hf_store import load_records_for_model
@@ -242,6 +275,28 @@ records = load_records_for_model(
     "<gpu_type>",
     last_n=5,
     successful_only=True,
+    baseline_eligible_only=True,
+)
+```
+
+For v2 exact-identity targets:
+
+```python
+from fastvideo.performance.hf_store import load_records_for_identity
+
+records = load_records_for_identity(
+    "/tmp/perf-tracking",
+    {
+        "workload_id": "<workload_id>",
+        "variant_id": "<variant_id>",
+        "benchmark_version": "<benchmark_version>",
+        "hardware_profile_id": "<hardware_profile_id>",
+        "software_profile_id": "<software_profile_id>",
+        "recipe_fingerprint": "<recipe_fingerprint>",
+    },
+    last_n=5,
+    successful_only=True,
+    baseline_eligible_only=True,
 )
 ```
 
@@ -257,7 +312,8 @@ medians after appending the proposed seed records, and source batch spread for:
 
 Also print how many successful old records exist. Make clear:
 
-- 1 seed record usually does not move a last-5 median by itself.
+- 1 seed record usually does not move an existing last-5 median by itself, but
+  it is enough to establish the first v2 baseline for a new exact identity.
 - 3 consistent seed records usually move the last-5 median immediately.
 - 5 consistent seed records effectively reset the last-5 window.
 - The records are intentional approved baseline resets and must be labeled
@@ -267,10 +323,10 @@ Also print how many successful old records exist. Make clear:
 
 Require an explicit confirmation phrase before preparing the upload:
 
-> About to RE-SEED performance baseline for `<model_id>` on `<gpu_type>`.
+> About to RE-SEED performance baseline for `<target description>`.
 > This will upload `<N>` new `success=true` records to
-> `FastVideo/performance-tracking/<sanitize(model_id)>/`, one per accepted
-> source JSON.
+> `FastVideo/performance-tracking/<sanitize(model_id)>/` or the source
+> artifact's v2 model directory, one per accepted source JSON.
 >
 > Reason: `<intent_rationale>`
 > Source results: `<source_results>`
@@ -288,25 +344,63 @@ Do not continue unless the user types exactly `confirm performance reseed`.
 
 ### 5. Create the accepted seed records
 
-Create one seed record from each normalized source result. Do not copy the
+Create one seed record from each normalized source result.
+
+For first v2 baseline seeds, use the scoped utility. It validates exact
+identity, requires successful scheduled-main full-suite `CALIBRATION_NEEDED`
+source artifacts, preserves the normalized v2 identity and metadata fields,
+and writes seed records with `success=true`, `baseline_eligible=true`, and
+`comparison_status=PASS`:
+
+```bash
+python fastvideo/tests/performance/seed_baseline.py \
+  --source-result <normalized_perf_1.json> \
+  --source-result <normalized_perf_2.json> \
+  --intent-rationale "<intent_rationale>" \
+  --max-intra-batch-regression 0.05 \
+  --tracking-root "${PERFORMANCE_TRACKING_ROOT}" \
+  --staging-root "${PERFORMANCE_RESEED_STAGING_ROOT:-/tmp/performance_reseed_prepared}"
+```
+
+The utility is prepare-only and intentionally has no upload option. Upload the
+scoped records only after the separate confirmation in step 6.
+
+The utility validates against an isolated fresh HF snapshot and leaves
+`PERFORMANCE_TRACKING_ROOT` untouched; that argument only proves the staging
+root is separate from the operator's tracking mirror. Before writing, it stops
+if the exact identity already has a successful baseline-eligible record or if
+the workload/variant/version already trusts another recipe. It atomically
+reserves the exact identity and writes a digest-protected upload manifest bound
+to the current HF endpoint, repository id, and repository type. Keep the
+prepared records, manifest, source files, and reservation unchanged until the
+operation is uploaded or explicitly cleaned up.
+
+If the prepared seed records look correct, upload only those scoped records in
+step 7. Do not rerun the utility with a different source list after approval.
+
+For legacy reseeds or accepted v2 baseline shifts from regression artifacts,
+create one seed record from each normalized source result. Do not copy the
 source JSON wholesale.
 
 Infer the baseline field allowlist from all existing HF records for the target
-`(model_id, gpu_type)` after syncing, including both `success=true` and
-`success=false` records. Use the union of non-provenance keys present in those
-target records, preserving only fields that also exist in the normalized
-source record or are explicitly set by the reseed workflow. Always include
-`model_id`, `timestamp`, and `success` because the upload path and baseline
-loader depend on them. Always set `timestamp` to a fresh reseed timestamp and
-`success` to `true`. Do not include unrelated source-only fields that are
-absent from existing HF records.
+after syncing, including both `success=true` and `success=false` records. For
+legacy targets the target is `(model_id, gpu_type)`. For v2 baseline-shift
+reseeds the target is the exact comparable identity. Use the union of
+non-provenance keys present in those target records, preserving only fields
+that also exist in the normalized source record or are explicitly set by the
+reseed workflow. Always include `model_id`, `timestamp`, `success`,
+`baseline_eligible`, and `comparison_status` because the upload path and
+baseline loader depend on them. Always set `timestamp` to a fresh reseed
+timestamp, `success` to `true`, `baseline_eligible` to `true`, and
+`comparison_status` to `PASS`. Do not include unrelated source-only fields
+that are absent from existing HF records.
 
 Exclude existing provenance or operator metadata from the inferred baseline
 field allowlist. At minimum, exclude keys prefixed with `baseline_reseed` and
 any fields known to be local-only audit metadata.
 
-If there are no previous HF records for the target model/GPU, fall back to this
-default baseline field list:
+If there are no previous HF records for the target, fall back to this default
+baseline field list:
 
 - `model_id`
 - `timestamp`
@@ -319,6 +413,22 @@ default baseline field list:
 - `dit_time_s`
 - `vae_decode_time_s`
 - `success`
+- `baseline_eligible`
+- `comparison_status`
+
+For v2 baseline-shift reseeds with no previous HF records for the exact
+identity, also preserve:
+
+- `workload_id`
+- `variant_id`
+- `benchmark_version`
+- `recipe_fingerprint`
+- `hardware_profile_id`
+- `software_profile_id`
+- `recipe`
+- `hardware_profile`
+- `software_profile`
+- `software_comparison_profile`
 
 Do not upload extra fields from the source artifact.
 
@@ -334,6 +444,22 @@ Optional provenance fields are allowed and useful:
 - `baseline_reseed_operator`
 - `baseline_reseed_max_intra_batch_regression`
 
+The v2 calibration seed utility writes analogous first-seed provenance:
+
+- `baseline_seed: true`
+- `baseline_seed_reason`
+- `baseline_seed_source_result`
+- `baseline_seed_source_status`
+- `baseline_seed_source_timestamp`
+- `baseline_seed_source_success`
+- `baseline_seed_source_run_source`
+- `baseline_seed_source_branch`
+- `baseline_seed_source_test_scope`
+- `baseline_seed_source_pr_number`
+- `baseline_seed_batch_size`
+- `baseline_seed_batch_index`
+- `baseline_seed_operator`
+
 Use a fresh reseed timestamp for each seed record, not the original source
 result timestamp. This is required because
 `load_records_for_model(..., last_n=5)` keeps the last records after loading
@@ -356,7 +482,8 @@ Prefer uploading new accepted seed records so failed history remains visible.
 Print:
 
 - Backup directory path under `/tmp`.
-- Prepared local record paths under `PERFORMANCE_TRACKING_ROOT`.
+- Prepared local record paths under `PERFORMANCE_RESEED_STAGING_ROOT`.
+- Prepared upload-manifest path under the identity reservation.
 - HF paths that will receive the new records.
 - Old rolling medians.
 - Source batch medians, source batch spread, reseed count, and candidate
@@ -368,22 +495,36 @@ prepared records plus backup on disk.
 
 ### 7. Upload only the scoped records
 
-Use the shared storage helper so the path and repo type match CI:
+For a first v2 calibration seed, use the manifest uploader after the user
+replies exactly `upload`:
 
-```python
-from fastvideo.performance.hf_store import upload_record
-
-upload_record("<local_record_path>", record, strict=True)
+```bash
+python -c 'from fastvideo.tests.performance.seed_baseline import upload_prepared_seed_manifest; print(upload_prepared_seed_manifest("<prepared_manifest>"))'
 ```
 
-Run it once per prepared record. Each upload goes to:
+The uploader verifies the source and prepared-record digests, pins and scans
+the current HF revision, rechecks exact-identity and recipe-cohort conflicts,
+and writes the entire batch in one commit whose `parent_commit` must still be
+current. A concurrent Hub update makes the commit fail. Do not retry
+automatically: preserve staging, refresh/review remote state, and request a new
+explicit `upload` after the conflict is understood. Each record goes to:
 
 ```text
 FastVideo/performance-tracking/<sanitize(model_id)>/<record_filename>.json
 ```
 
-Never bulk upload the whole tracking root. Never modify another model's
-directory in the same operation.
+Never call `upload_record()` once per first-seed record: that can partially
+land the batch and has no compare-and-swap guard.
+
+For a legacy reseed or an accepted v2 baseline shift, the first-seed manifest
+validator does not apply because an eligible baseline already exists. Upload
+only the individually reviewed records prepared in step 5 with the shared
+`upload_record(local_path, record, strict=True)` helper. Stop on the first
+failure and report exactly which records reached HF; do not silently rerun or
+replicate the remainder.
+
+Never bulk upload the tracking or staging root, and never modify another
+model's directory in the same operation.
 
 ### 8. Report outcome and offer cleanup
 
@@ -405,9 +546,14 @@ distinguish an accepted baseline shift from a hidden regression.
 After the upload is verified, ask whether the user wants to clear temporary
 local state. Explain what each directory is for:
 
-- `PERFORMANCE_TRACKING_ROOT`, usually `/tmp/perf-tracking`: local synced
-  mirror of `FastVideo/performance-tracking` plus the prepared local seed
-  records used for scoped upload.
+- `PERFORMANCE_TRACKING_ROOT`, usually `/tmp/perf-tracking`: read-only local
+  synced mirror used for operator review and reporting. First-v2 preparation
+  independently proves remote state from a fresh temporary HF snapshot.
+- `PERFORMANCE_RESEED_STAGING_ROOT`, usually
+  `/tmp/performance_reseed_prepared`: prepared local seed records used for the
+  scoped upload, plus the identity reservation and digest manifest. Keeping
+  this separate prevents aborted preparations from appearing in later
+  baseline reads.
 - `/tmp/performance_reseed_backup/<...>`: local backup of the target model's
   pre-reseed HF history plus `PROVENANCE.txt`, kept so a bad reseed can be
   audited or corrected.
@@ -417,14 +563,18 @@ local state. Explain what each directory is for:
 Ask:
 
 > Reseed succeeded. Do you want me to delete the local temp tracking mirror,
-> source downloads, and reseed backup under `/tmp`? These files are local
-> safety/audit artifacts only; HF already has the uploaded records.
+> this reseed's prepared staging records, source downloads, and reseed backup
+> under `/tmp`? These files are local safety/audit artifacts only; HF already
+> has the uploaded records.
 >
 > Reply `cleanup reseed temp` to delete them, anything else to keep them.
 
 Do not delete anything unless the user replies exactly
 `cleanup reseed temp`. If cleanup is requested, remove only the specific
-directories created for this reseed. Never remove unrelated `/tmp` contents.
+directories and prepared record paths created for this reseed. Do not remove
+the shared staging root when it contains other records. Remove this operation's
+identity reservation only with its prepared records and manifest, and never
+remove unrelated `/tmp` contents.
 
 ## Failure modes and handling
 
@@ -436,19 +586,34 @@ directories created for this reseed. Never remove unrelated `/tmp` contents.
   against the source batch median by more than `max_intra_batch_regression`.
   Ask for cleaner sources or a reviewed explanation before continuing.
 - **Too few source records to move the median.** Continue only after making
-  clear that one or two records may not immediately move the last-5 median.
+  clear that one or two records may not immediately move an existing last-5
+  median. This warning does not block a first v2 calibration seed for an exact
+  identity with no eligible baseline yet.
 - **The source results are noisy or suspicious.** Stop. Reseeding amplifies
   those measurements into the baseline, so they must be reviewed first.
 - **HF sync fails.** Stop for destructive reseeds. A stale or empty sync can
   make the old baseline look missing.
+- **The exact v2 identity already has an eligible baseline.** Stop. The
+  `CALIBRATION_NEEDED` artifact is stale; use the reviewed baseline-shift path
+  instead of the first-seed utility.
+- **The workload/variant/version trusts another recipe.** Stop. The source is
+  stale relative to the current recipe cohort and must not bypass
+  `RECIPE_MISMATCH` by creating a second trusted recipe.
+- **The staging root already has a prepared seed for the exact identity.**
+  Stop and reuse, upload, or explicitly clean that preparation. Do not prepare
+  another copy of the same measurement.
+- **The conditional Hub commit loses its parent race.** Stop without retrying.
+  Keep the preparation, refresh and review the new remote state, then request
+  a new explicit `upload` only if the seed is still valid.
 - **Candidate still violates fixed thresholds.** Report that this skill only
   handles the rolling HF baseline; update benchmark JSON thresholds in code
   review if maintainers accept the new absolute limit.
 - **The user aborts at either confirmation.** Leave the backup and prepared
   records on disk. Nothing should be uploaded.
-- **The user declines cleanup.** Keep `/tmp/perf-tracking`, the source
-  download directory if any, and `/tmp/performance_reseed_backup/<...>` in
-  place for audit/debugging.
+- **The user declines cleanup.** Keep `/tmp/perf-tracking`, the prepared seed
+  records under `/tmp/performance_reseed_prepared`, the source download
+  directory if any, and `/tmp/performance_reseed_backup/<...>` in place for
+  audit/debugging.
 - **A bad seed was uploaded.** Use the backup and HF history to identify the
   uploaded file, then remove or supersede it with an explicitly reviewed
   corrective record. Do not silently rewrite unrelated history.
@@ -459,8 +624,9 @@ directories created for this reseed. Never remove unrelated `/tmp` contents.
   intentional baseline replacement.
 - `fastvideo/tests/performance/compare_baseline.py` — normalization, rolling
   median comparison, and persistence rules.
-- `fastvideo/performance/hf_store.py` — HF sync, record loading,
-  `sanitize()`, and `upload_record()`.
+- `fastvideo/performance/hf_store.py` — HF sync and record loading helpers.
+- `fastvideo/tests/performance/seed_baseline.py` — first-seed preparation,
+  staging reservation, manifest validation, and conditional batch upload.
 - `fastvideo/tests/performance/test_inference_performance.py` — source result
   JSON schema.
 - `.buildkite/performance-benchmarks/tests/*.json` — fixed absolute benchmark
@@ -473,3 +639,4 @@ directories created for this reseed. Never remove unrelated `/tmp` contents.
 | 2026-05-03 | Initial version. Sister workflow to `reseed-ssim-references`, scoped to one performance `(model_id, gpu_type)` baseline seed with backup, confirmation, provenance, and `success=true` upload. |
 | 2026-05-03 | Previous policy: replicate one approved shifted source result into 3 success records by default, or 5 only when explicitly requested. Add provenance marker for replicated-source reseeds. Superseded by the 2026-05-08 dynamic multi-source policy. |
 | 2026-05-08 | Replace fixed 3/5 replication with dynamic multi-source reseeding: upload one seed record per reviewed source JSON, validate intra-batch consistency, move backup/source scratch under `/tmp`, and ask whether to clean temp state after successful upload. |
+| 2026-07-13 | Keep first-v2-seed preparation outside the canonical mirror, reserve staging identities atomically, reject stale or replayed calibration seeds, and upload reviewed manifests with a single parent-guarded Hub commit. |

.agents/skills/reseed-ssim-references/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: reseed-ssim-references
-description: Re-seed HF reference videos for a single existing SSIM test on Modal L40S. Always backs up current refs locally first, regenerates on Modal, pauses for the user to eyeball before-vs-after quality, then overwrites the targeted `<model_id>` subtree on `FastVideo/ssim-reference-videos` with `--force`. Use when an intentional code change (model port fix, attention backend swap, kernel upgrade, hyperparameter change) has invalidated existing refs and they need to be regenerated. Pairs with `seed-ssim-references`, which is for first-time seeding only.
+description: Re-seed HF reference videos for a single existing SSIM test on Modal L40S. Always backs up current refs locally first, regenerates on Modal, pauses for the user to eyeball before-vs-after quality, then overwrites the targeted model subtree on `FastVideo/ssim-reference-videos` with `--force`. Use when an intentional code change (model port fix, attention backend swap, kernel upgrade, hyperparameter change) has invalidated existing refs and they need to be regenerated. Pairs with `seed-ssim-references`, which is for first-time seeding only.
 ---
 
 # Re-seed SSIM Reference Videos
@@ -13,7 +13,7 @@ on HF — the old refs are overwritten — so the skill always:
 
 1. Confirms intent with a one-liner the user has to type.
 2. Downloads the existing refs as a local, timestamped backup.
-3. Regenerates on Modal L40S (same code path that CI uses).
+3. Regenerates through the manual legacy Modal L40S maintenance path.
 4. Pauses for a side-by-side eyeball of backup vs new mp4s.
 5. Uploads with `--force`, scoped to the single `--model-id`.
 6. Reminds the user to keep the backup until the PR lands.
@@ -51,12 +51,13 @@ harder to recover from than failing closed.
 
 Hardcoded:
 
-- Modal GPU: **L40S** (matches CI; re-seeding from another SKU produces refs
-  that L40S CI cannot match).
+- Modal GPU: **L40S**. This is a manual reference-maintenance target, not the
+  active Slurm CI compute path; changing the SKU also changes the historical
+  `L40S_reference_videos` contract.
 - Quality tier: **`default`**. `full_quality` is a separate, deliberate
   operation.
 - HF repo: `FastVideo/ssim-reference-videos` (override via
-  `FASTVIDEO_SSIM_REFERENCE_HF_REPO`).
+  `FASTVIDEO_TEST_SSIM_REFERENCE_HF_REPO`).
 - Device folder: `L40S_reference_videos`.
 
 ## Prerequisites

.agents/skills/seed-ssim-references/SKILL.md

@@ -35,7 +35,8 @@ The skill is run **manually**, once per new test. Before invoking it, the user
 has already sanity-tested the new test locally — it launches `VideoGenerator`
 and writes an artefact without crashing (the missing-reference assertion at
 the end is expected). The skill does not re-test locally; it goes straight
-to Modal L40S (which is what CI uses).
+to the manual legacy Modal L40S reference-maintenance target. Active CI runs
+on the Slinky Slurm cluster and only consumes the resulting references.
 
 ## When to use
 
@@ -61,7 +62,8 @@ Prompt the user for it if they didn't supply it.
 
 Everything else is fixed:
 
-- Modal runner GPU: **L40S** (hardcoded in `fastvideo/tests/modal/ssim_test.py`).
+- Modal maintenance GPU: **L40S** (hardcoded in
+  `fastvideo/tests/modal/ssim_test.py`; this is not the active CI compute path).
 - Device folder: `L40S_reference_videos`.
 - Quality tier: `default` (the tier CI runs). The `full_quality` tier is not
   seeded by this skill.

.buildkite/performance-benchmarks/tests/wan-t2v-1.3b-gb10-1gpu.json

@@ -0,0 +1,51 @@
+{
+  "benchmark_id": "wan-t2v-1.3b-1gpu-gb10",
+  "config_schema_version": 2,
+  "workload_id": "wan-t2v",
+  "variant_id": "1.3b-sp1",
+  "benchmark_version": 3,
+  "description": "Wan2.1 T2V 1.3B single-GPU inference performance on NVIDIA DGX Spark (GB10). Single-GPU variant of wan-t2v-1.3b (same workload_id for dashboard comparability). Gated to the GB10 via run_config.gpu_types so it does not run on the shared H100/L40S lanes.",
+  "model": {
+    "model_path": "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
+    "model_short_name": "Wan2.1-T2V-1.3B"
+  },
+  "init_kwargs": {
+    "num_gpus": 1,
+    "flow_shift": 7.0,
+    "sp_size": 1,
+    "tp_size": 1,
+    "vae_sp": false,
+    "vae_tiling": true,
+    "text_encoder_precisions": ["fp32"]
+  },
+  "generation_kwargs": {
+    "height": 480,
+    "width": 832,
+    "num_frames": 45,
+    "num_inference_steps": 4,
+    "guidance_scale": 3,
+    "embedded_cfg_scale": 6,
+    "seed": 1024,
+    "fps": 24,
+    "neg_prompt": "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"
+  },
+  "test_prompts": [
+    "Will Smith casually eats noodles, his relaxed demeanor contrasting with the energetic background of a bustling street food market. The scene captures a mix of humor and authenticity. Mid-shot framing, vibrant lighting."
+  ],
+  "run_config": {
+    "num_warmup_runs": 2,
+    "num_measurement_runs": 5,
+    "required_gpus": 1,
+    "gpu_types": ["GB10"]
+  },
+  "thresholds": {
+    "GB10": {
+      "max_generation_time_s": 55.0,
+      "max_peak_memory_mb": 12000.0
+    },
+    "default": {
+      "max_generation_time_s": 120.0,
+      "max_peak_memory_mb": 40000.0
+    }
+  }
+}

.buildkite/performance-benchmarks/tests/wan-t2v-1.3b.json

@@ -3,7 +3,7 @@
   "config_schema_version": 2,
   "workload_id": "wan-t2v",
   "variant_id": "1.3b-sp2",
-  "benchmark_version": 2,
+  "benchmark_version": 3,
   "description": "Wan2.1 T2V 1.3B inference performance",
   "model": {
     "model_path": "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",

.buildkite/pipeline.yml

@@ -1,6 +1,9 @@
 env:
   IMAGE_VERSION: "py3.12-latest"
   BUILDKITE_CLEAN_CHECKOUT: true
+  # Slurm workers clone the immutable commit and initialize submodules inside
+  # their isolated container. The Buildkite login-plane checkout is a no-op.
+  BUILDKITE_GIT_SUBMODULES: false
 
 notify:
   - github_commit_status:
@@ -8,527 +11,474 @@ notify:
     if: build.env("TEST_SCOPE") == "fastcheck" || build.env("TEST_SCOPE") == null
   - github_commit_status:
       context: "full-suite-passed"
-    if: build.env("TEST_SCOPE") == "full"
+    if: build.env("TEST_SCOPE") == "full" || build.env("TEST_SCOPE") == "merge"
   - github_commit_status:
       context: "direct-test-completed"
     if: build.env("TEST_SCOPE") == "direct"
+  - github_commit_status:
+      context: "scheduled-ssim-passed"
+    if: build.env("TEST_SCOPE") == "scheduled"
 
+# This is the complete active GPU CI surface. Every command is a trusted host
+# dispatcher, and every test payload executes inside the Slinky Slurm tray.
+# fastvideo/tests/modal remains available only for an explicit manual rollback;
+# no active pipeline or slash-command route invokes it.
+# Buildkite hands jobs to free agents in the order they appear here. Golden-gate comes first
+# because every later merge lane waits for it; the fastcheck lanes follow from longest to
+# shortest measured runtime, so the longest lane never starts last and stretches the build.
 steps:
-  # ============================================================
-  # Direct test: triggered by /test <name> slash command.
-  # Labels match fastcheck/full-suite counterparts so the GitHub
-  # check status overwrites the original failed check.
-  # Only ONE step executes per build (gated by TEST_TYPE).
-  # ============================================================
+  - label: ":test_tube: Golden-Gate Tests"
+    key: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,golden-gate,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "golden_gate" || build.env("TEST_TYPE") == "golden_gate_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "golden_gate_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
 
-  # --- Fastcheck-scope direct tests ---
-    - label: ":microscope: Encoder Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "encoder"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":microscope: VAE Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "vae"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":microscope: Transformer Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "transformer"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":microscope: Kernel Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "kernel_tests"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":microscope: Unit Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "unit_test"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":microscope: DreamVerse App Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "dreamverse_app"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
+  - label: ":microscope: Unit Tests"
+    key: "unit"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,unit,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "unit_test" || build.env("TEST_TYPE") == "unit_test_ci"))
+    command: "/opt/fastvideo-ci-runner/run-unit"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "unit_test_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
 
-  # --- Full-suite-scope direct tests ---
-    - label: ":bar_chart: SSIM Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "ssim"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-          - exit_status: 1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: LoRA Inference Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "inference_lora"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: LoRA Extraction Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "lora_extraction"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Training Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "training"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Distillation DMD Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "distillation_dmd"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Self-Forcing Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "self_forcing"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: LoRA Training Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "training_lora"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-          - exit_status: 1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Training Tests VSA"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "training_vsa"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-          - exit_status: 1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Inference Tests VMoBA"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "inference_vmoba"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Performance Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "performance"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: API Server Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "api_server"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Train Framework Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "train_framework"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
-    - label: ":test_tube: Eval Metrics Tests"
-      if: build.env("TEST_SCOPE") == "direct" && build.env("TEST_TYPE") == "eval"
-      command: "timeout 90m .buildkite/scripts/pr_test.sh"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      agents:
-        queue: "default"
+  - label: ":microscope: Kernel Tests"
+    key: "kernel-tests"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,kernel-tests,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "kernel_tests" || build.env("TEST_TYPE") == "kernel_tests_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "kernel_tests_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
 
-  # ============================================================
-  # Fastcheck: Runs on every PR (~10-15 min parallel)
-  # Core component validation: encoders, VAEs, transformers,
-  # CUDA kernels, and unit tests.
-  # ============================================================
-    - label: "Trigger Fastcheck"
-      if: build.env("TEST_SCOPE") == "fastcheck" || build.env("TEST_SCOPE") == null
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      plugins:
-        - monorepo-diff#v1.4.0:
-            diff: 'git fetch origin "${BUILDKITE_PULL_REQUEST_BASE_BRANCH:-main}" && git diff --name-only "origin/${BUILDKITE_PULL_REQUEST_BASE_BRANCH:-main}...HEAD"'
-            watch:
-            - path:
-                - "fastvideo/models/encoders/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/encoders/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 20m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: Encoder Tests"
-                env:
-                  - TEST_TYPE=encoder
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/models/vaes/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/vaes/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 20m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: VAE Tests"
-                env:
-                  - TEST_TYPE=vae
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/models/dits/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/transformers/**"
-                - "fastvideo/layers/**"
-                - "fastvideo/attention/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: Transformer Tests"
-                env:
-                  - TEST_TYPE=transformer
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo-kernel/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: Kernel Tests"
-                env:
-                  - TEST_TYPE=kernel_tests
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - ".buildkite/**"
-                - ".github/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: Unit Tests"
-                env:
-                  - TEST_TYPE=unit_test
-                agents:
-                  queue: "default"
-            - path:
-                - "apps/dreamverse/**"
-                - "pyproject.toml"
-              config:
-                command: "timeout 30m .buildkite/scripts/pr_test.sh"
-                label: ":microscope: DreamVerse App Tests"
-                env:
-                  - TEST_TYPE=dreamverse_app
-                agents:
-                  queue: "default"
+  - label: ":microscope: DreamVerse App Tests"
+    key: "dreamverse"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,dreamverse,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "dreamverse_app" || build.env("TEST_TYPE") == "dreamverse_app_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "dreamverse_app_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
 
-  # ============================================================
-  # Full Suite: Runs when TEST_SCOPE=full
-  # Triggered by adding the 'ready' label (via ci-trigger-full-suite.yml)
-  # or on-demand via /test full slash command.
-  # Includes integration tests, SSIM regression, training pipelines,
-  # and performance benchmarks.
-  # ============================================================
-    - label: "Trigger Full Suite"
-      if: build.env("TEST_SCOPE") == "full"
-      retry:
-        automatic:
-          - exit_status: 128
-            limit: 3
-          - exit_status: -1
-            limit: 2
-      plugins:
-        - monorepo-diff#v1.4.0:
-            diff: 'git fetch origin "${BUILDKITE_PULL_REQUEST_BASE_BRANCH:-main}" && git diff --name-only "origin/${BUILDKITE_PULL_REQUEST_BASE_BRANCH:-main}...HEAD"'
-            watch:
-            - path:
-                - "fastvideo/**/*.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 90m .buildkite/scripts/pr_test.sh"
-                label: ":bar_chart: SSIM Tests"
-                env:
-                  - TEST_TYPE=ssim
-                retry:
-                  automatic:
-                    - exit_status: 1
-                      limit: 2
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/tests/lora/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/transformers/**"
-                - "fastvideo/pipelines/**"
-                - "fastvideo/layers/lora/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 20m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: LoRA Inference Tests"
-                env:
-                  - TEST_TYPE=inference_lora
-                agents:
-                  queue: "default"
-            - path:
-                - "scripts/lora_extraction/**"
-                - "fastvideo/tests/lora_extraction/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/training/training_utils.py"
-                - "fastvideo/layers/lora/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 90m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: LoRA Extraction Tests"
-                env:
-                  - TEST_TYPE=lora_extraction
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Training Tests"
-                env:
-                  - TEST_TYPE=training
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/training/*distillation_pipeline.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Distillation DMD Tests"
-                env:
-                  - TEST_TYPE=distillation_dmd
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/training/*self_forcing_distillation_pipeline.py"
-                - "fastvideo/tests/training/self-forcing/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 30m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Self-Forcing Tests"
-                env:
-                  - TEST_TYPE=self_forcing
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: LoRA Training Tests"
-                env:
-                  - TEST_TYPE=training_lora
-                retry:
-                  automatic:
-                    - exit_status: 1
-                      limit: 2
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "fastvideo-kernel/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Training Tests VSA"
-                env:
-                  - TEST_TYPE=training_vsa
-                retry:
-                  automatic:
-                    - exit_status: 1
-                      limit: 2
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo-kernel/**"
-                - "fastvideo/attention/backends/vmoba.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Inference Tests VMoBA"
-                env:
-                  - TEST_TYPE=inference_vmoba
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/models/dits/**"
-                - "fastvideo/pipelines/**"
-                - "fastvideo/attention/**"
-                - "fastvideo/layers/**"
-                - "fastvideo/worker/**"
-                - "fastvideo/entrypoints/**"
-                - "fastvideo/tests/performance/**"
-                - ".buildkite/performance-benchmarks/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 30m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Performance Tests"
-                env:
-                  - TEST_TYPE=performance
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/entrypoints/openai/**"
-                - "fastvideo/entrypoints/cli/serve.py"
-                - "fastvideo/tests/entrypoints/test_openai_api_integration.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 30m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: API Server Tests"
-                env:
-                  - TEST_TYPE=api_server
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/train/**"
-                - "fastvideo/tests/train/models/**"
-                - "fastvideo/tests/train/fixtures/**"
-                - "fastvideo/models/dits/**"
-                - "fastvideo/models/loader/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 30m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Train Framework Tests"
-                env:
-                  - TEST_TYPE=train_framework
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/eval/**"
-                - "fastvideo/tests/eval/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile"
-              config:
-                command: "timeout 90m .buildkite/scripts/pr_test.sh"
-                label: ":test_tube: Eval Metrics Tests"
-                env:
-                  - TEST_TYPE=eval
-                agents:
-                  queue: "default"
+  - label: ":microscope: Encoder Tests"
+    key: "encoder"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,encoder,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "encoder" || build.env("TEST_TYPE") == "encoder_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "encoder_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":microscope: VAE Tests"
+    key: "vae"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,vae,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "vae" || build.env("TEST_TYPE") == "vae_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "vae_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":microscope: Transformer Tests"
+    key: "transformer"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,transformer,/) ||
+      build.env("TEST_SCOPE") == "fastcheck" ||
+      build.env("TEST_SCOPE") == null ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "transformer" || build.env("TEST_TYPE") == "transformer_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "transformer_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":bar_chart: SSIM Tests"
+    key: "ssim"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      build.env("TEST_SCOPE") == "scheduled" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,ssim,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "ssim" || build.env("TEST_TYPE") == "ssim_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    concurrency: 1
+    concurrency_group: "fastvideo/slinky/whole-tray"
+    env:
+      TEST_TYPE: "ssim_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+        - exit_status: 1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: LoRA Inference Tests"
+    key: "lora-inference"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,lora-inference,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "inference_lora" || build.env("TEST_TYPE") == "inference_lora_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "inference_lora_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: LoRA Extraction Tests"
+    key: "lora-extraction"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,lora-extraction,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "lora_extraction" || build.env("TEST_TYPE") == "lora_extraction_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "lora_extraction_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Training Tests"
+    key: "training"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,training,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "training" || build.env("TEST_TYPE") == "training_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    concurrency: 1
+    concurrency_group: "fastvideo/slinky/whole-tray"
+    env:
+      TEST_TYPE: "training_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Distillation DMD Tests"
+    key: "distillation"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,distillation,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "distillation_dmd" || build.env("TEST_TYPE") == "distillation_dmd_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "distillation_dmd_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Self-Forcing Tests"
+    key: "self-forcing"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,self-forcing,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "self_forcing" || build.env("TEST_TYPE") == "self_forcing_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "self_forcing_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: LoRA Training Tests"
+    key: "lora-training"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,lora-training,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "training_lora" || build.env("TEST_TYPE") == "training_lora_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "training_lora_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+        - exit_status: 1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Training Tests VSA"
+    key: "training-vsa"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,training-vsa,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "training_vsa" || build.env("TEST_TYPE") == "training_vsa_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "training_vsa_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+        - exit_status: 1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Inference Tests VMoBA"
+    key: "inference-vmoba"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,inference-vmoba,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "inference_vmoba" || build.env("TEST_TYPE") == "inference_vmoba_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "inference_vmoba_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Performance Tests"
+    key: "performance"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,performance,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "performance" || build.env("TEST_TYPE") == "performance_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "performance_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: API Server Tests"
+    key: "api-server"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,api-server,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "api_server" || build.env("TEST_TYPE") == "api_server_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "api_server_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Train Framework Tests"
+    key: "train-framework"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,train-framework,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "train_framework" || build.env("TEST_TYPE") == "train_framework_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "train_framework_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"
+
+  - label: ":test_tube: Eval Metrics Tests"
+    key: "eval"
+    depends_on: "golden-gate"
+    if: |
+      build.env("TEST_SCOPE") == "full" ||
+      (build.env("TEST_SCOPE") == "merge" &&
+       build.env("MERGE_TEST_PLAN") =~ /,eval,/) ||
+      (build.env("TEST_SCOPE") == "direct" &&
+       (build.env("TEST_TYPE") == "eval" || build.env("TEST_TYPE") == "eval_ci"))
+    command: "/opt/fastvideo-ci-runner/run-ci"
+    timeout_in_minutes: 90
+    env:
+      TEST_TYPE: "eval_ci"
+    retry:
+      automatic:
+        - exit_status: 128
+          limit: 3
+        - exit_status: -1
+          limit: 2
+    agents:
+      queue: "ci-runner"

.buildkite/scripts/lanes/api_server.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the OpenAI-compatible API lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/entrypoints/test_openai_api_integration.py -vs

.buildkite/scripts/lanes/distillation_dmd.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the distillation-DMD lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/training/distill/test_distill_dmd.py -vs

.buildkite/scripts/lanes/dreamverse.sh

@@ -0,0 +1,87 @@
+#!/usr/bin/env bash
+# DreamVerse needs a GPU for import-time device resolution, but it does not
+# build or exercise fastvideo-kernel. A checksummed Node archive is installed
+# in the disposable Slurm container because the shared CI image is
+# Python/CUDA focused.
+set -euo pipefail
+
+node_version=v22.23.2
+case $(uname -m) in
+  aarch64 | arm64)
+    node_arch=arm64
+    node_archive_sha256=013b59cfd2819703a6f4a14ab891fc46fc2a4e3f5bcd92de3fb4929b43e35b30
+    ;;
+  x86_64 | amd64)
+    node_arch=x64
+    node_archive_sha256=b294a556e639d64338823920e5866c21c02741742d2e1529ee1a225c1ec9252a
+    ;;
+  *)
+    echo "Unsupported architecture for DreamVerse Node runtime: $(uname -m)" >&2
+    exit 2
+    ;;
+esac
+node_archive="node-${node_version}-linux-${node_arch}.tar.gz"
+node_runtime_root=$(mktemp -d -t fastvideo-node.XXXXXX)
+node_archive_path="${node_runtime_root}/${node_archive}"
+node_install_dir="${node_runtime_root}/${node_archive%.tar.gz}"
+curl --proto '=https' --tlsv1.2 --retry 5 --retry-all-errors \
+  --location --fail --silent --show-error \
+  "https://nodejs.org/dist/${node_version}/${node_archive}" \
+  --output "$node_archive_path"
+printf '%s  %s\n' "$node_archive_sha256" "$node_archive_path" | sha256sum --check --status
+tar -xzf "$node_archive_path" -C "$node_runtime_root"
+export PATH="${node_install_dir}/bin:${PATH}"
+node --version
+npm --version
+
+export PYTHONPATH="$(pwd)/apps/dreamverse${PYTHONPATH:+:$PYTHONPATH}"
+pytest apps/dreamverse/dreamverse/tests -q
+
+cd apps/dreamverse/web
+npm ci
+npm run typecheck
+npm test
+machine_arch=$(uname -m)
+if [[ $machine_arch =~ ^(aarch64|arm64)$ ]]; then
+  npx playwright install --with-deps chromium firefox
+else
+  npx playwright install --with-deps chromium webkit firefox
+fi
+
+master_port=${MASTER_PORT:-7959}
+BACKEND_PORT=${BACKEND_PORT:-$((master_port + 50))}
+python -m uvicorn dreamverse.mock_server:app --host 127.0.0.1 --port "$BACKEND_PORT" &
+mock_server_pid=$!
+cleanup() {
+  kill "$mock_server_pid" 2>/dev/null || true
+  wait "$mock_server_pid" 2>/dev/null || true
+}
+trap cleanup EXIT INT TERM
+
+for _ in {1..30}; do
+  curl -fsS "http://127.0.0.1:$BACKEND_PORT/healthz" && break
+  sleep 1
+done
+curl -fsS "http://127.0.0.1:$BACKEND_PORT/healthz"
+
+if [[ $machine_arch =~ ^(aarch64|arm64)$ ]]; then
+  # Playwright WebKit traps before opening a page on Linux ARM64, and its
+  # bundled Chromium lacks the H.264/AAC codecs used by the fMP4 assertions.
+  # Firefox covers every flow, including streaming. Chromium and its mobile
+  # profile still cover all codec-independent UI behavior on GB200.
+  BACKEND_HOST=127.0.0.1 BACKEND_PORT="$BACKEND_PORT" CI=1 \
+    npm run e2e -- --project=firefox
+  BACKEND_HOST=127.0.0.1 BACKEND_PORT="$BACKEND_PORT" CI=1 \
+    npm run e2e -- \
+      --project=chromium \
+      --project=mobile-chromium \
+      --grep-invert='streams, plays, and surfaces a downloadable clip|starts a new project and switches back to the prior session|saved projects persist across a page reload'
+else
+  BACKEND_HOST=127.0.0.1 BACKEND_PORT="$BACKEND_PORT" CI=1 \
+    npm run e2e -- \
+      --project=chromium \
+      --project=webkit \
+      --project=firefox \
+      --project=mobile-safari \
+      --project=mobile-chromium
+fi

.buildkite/scripts/lanes/encoder.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the encoder lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/encoders -vs

.buildkite/scripts/lanes/eval.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the evaluation lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/eval -vs

.buildkite/scripts/lanes/golden_gate.sh

@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the golden-gate lane. Environment (HF_HOME
+# and authentication) is the runner's responsibility.
+set -euo pipefail
+
+golden_root=./fastvideo/tests/golden_gate
+selected=${FASTVIDEO_GOLDEN_TEST_FILES-}
+if [ -z "$selected" ]; then
+  if [ "${TEST_SCOPE:-}" = merge ]; then
+    echo "Missing FASTVIDEO_GOLDEN_TEST_FILES for merge scope" >&2
+    exit 2
+  fi
+  selected=all
+fi
+if [ "$selected" = all ]; then
+  exec pytest "$golden_root" -xvs
+fi
+
+[[ $selected =~ ^test_[a-z0-9_]+\.py(,test_[a-z0-9_]+\.py)*$ ]] || {
+  echo "Invalid FASTVIDEO_GOLDEN_TEST_FILES selection" >&2
+  exit 2
+}
+
+IFS=, read -r -a golden_files <<< "$selected"
+golden_paths=()
+for golden_file in "${golden_files[@]}"; do
+  golden_path="$golden_root/$golden_file"
+  [ -f "$golden_path" ] || {
+    echo "Selected golden test does not exist: $golden_file" >&2
+    exit 2
+  }
+  golden_paths+=("$golden_path")
+done
+
+exec pytest "${golden_paths[@]}" -xvs

.buildkite/scripts/lanes/inference_lora.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the LoRA-inference lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/inference/lora/test_lora_inference_similarity.py -vs

.buildkite/scripts/lanes/inference_vmoba.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the VMoBA-inference lane.
+set -euo pipefail
+
+exec python fastvideo/tests/inference/vmoba/test_vmoba_inference.py

.buildkite/scripts/lanes/kernel_tests.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the custom-kernel lane.
+set -euo pipefail
+
+exec pytest fastvideo-kernel/tests/ -vs

.buildkite/scripts/lanes/lora_extraction.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the LoRA-extraction lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/lora_extraction/ -vs

.buildkite/scripts/lanes/performance.sh

@@ -0,0 +1,64 @@
+#!/usr/bin/env bash
+# Canonical Slurm performance lane. Reports are written outside the checkout
+# so the trusted host driver can upload them after untrusted code exits.
+set -uo pipefail
+
+export PERFORMANCE_TRACKING_ROOT=/tmp/perf-tracking
+export PERF_REPORTS_DIR=/workspace/artifacts/performance
+mkdir -p "$PERF_REPORTS_DIR"
+
+if [[ ${BUILDKITE_PULL_REQUEST:-false} =~ ^[1-9][0-9]*$ ]]; then
+  export PERF_RUN_SOURCE=pr
+  export PERF_UPLOAD_POLICY=pass
+elif [ "${BUILDKITE_BRANCH:-}" = main ] \
+  && { [ "${BUILDKITE_SOURCE:-}" = schedule ] || [ "${TEST_SCOPE:-}" = full ]; }; then
+  export PERF_RUN_SOURCE=scheduled_main
+  export PERF_UPLOAD_POLICY=always
+elif [ "${TEST_SCOPE:-}" = direct ]; then
+  export PERF_RUN_SOURCE=unknown
+  export PERF_UPLOAD_POLICY=pass
+else
+  export PERF_RUN_SOURCE=unknown
+  export PERF_UPLOAD_POLICY=never
+fi
+
+nvidia-smi \
+  --query-gpu=index,timestamp,clocks.sm,clocks.max.sm,power.draw,power.limit,temperature.gpu \
+  --format=csv -l 10 > "$PERF_REPORTS_DIR/gpu_telemetry.csv" 2>/dev/null &
+telemetry_pid=$!
+cleanup() {
+  kill "$telemetry_pid" 2>/dev/null || true
+  wait "$telemetry_pid" 2>/dev/null || true
+}
+trap cleanup EXIT INT TERM
+
+pytest ./fastvideo/tests/performance/test_inference_performance.py -vs
+pytest_rc=$?
+compare_rc=0
+if [ "$pytest_rc" -eq 0 ] || [ "$PERF_UPLOAD_POLICY" = always ]; then
+  PERF_PYTEST_RC=$pytest_rc python ./fastvideo/tests/performance/compare_baseline.py
+  compare_rc=$?
+fi
+python ./fastvideo/tests/performance/dashboard.py || true
+cp -f fastvideo/tests/performance/results/*.json "$PERF_REPORTS_DIR/" 2>/dev/null || true
+# The trusted host relays only .md/.html/.json/.csv from PERF_REPORTS_DIR, so
+# mirror each captured worker log with an allowlisted extension.
+for worker_log in fastvideo/tests/performance/results/worker_logs/*.log; do
+  [ -f "$worker_log" ] || continue
+  base=$(basename "${worker_log%.log}")
+  # WorkerLogCapture keeps a .log.1 backup after rollover, and read_log_tail
+  # includes it; mirror that retained history too so the artifact is complete.
+  if [ -f "$worker_log.1" ]; then
+    cp -f "$worker_log.1" "$PERF_REPORTS_DIR/${base}.1.md" 2>/dev/null || true
+  fi
+  cp -f "$worker_log" "$PERF_REPORTS_DIR/${base}.md" 2>/dev/null || true
+done
+
+echo "--- GPU telemetry (clocks.sm vs clocks.max.sm reveals capped hosts) ---"
+cat "$PERF_REPORTS_DIR/gpu_telemetry.csv" || true
+
+final_rc=$pytest_rc
+if [ "$final_rc" -eq 0 ]; then
+  final_rc=$compare_rc
+fi
+exit "$final_rc"

.buildkite/scripts/lanes/self_forcing.sh

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the self-forcing lane.
+set -euo pipefail
+
+export WANDB_MODE=offline
+exec pytest ./fastvideo/tests/training/self-forcing/test_self_forcing.py -vs

.buildkite/scripts/lanes/ssim.sh

@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+# Canonical four-GPU SSIM lane for the Slinky Slurm worker.
+set -euo pipefail
+
+args=()
+if [ "${FASTVIDEO_SSIM_BOOTSTRAP_MODE:-0}" = 1 ]; then
+  args+=(--bootstrap-mode)
+fi
+selected=${FASTVIDEO_SSIM_TEST_FILES-}
+if [ -z "$selected" ]; then
+  if [ "${TEST_SCOPE:-}" = merge ]; then
+    echo "Missing FASTVIDEO_SSIM_TEST_FILES for merge scope" >&2
+    exit 2
+  fi
+  selected=all
+fi
+if [ "$selected" != all ]; then
+  [[ $selected =~ ^test_[a-z0-9_]+\.py(,test_[a-z0-9_]+\.py)*$ ]] || {
+    echo "Invalid FASTVIDEO_SSIM_TEST_FILES selection" >&2
+    exit 2
+  }
+  IFS=, read -r -a ssim_files <<< "$selected"
+  for ssim_file in "${ssim_files[@]}"; do
+    args+=(--test-file "$ssim_file")
+  done
+fi
+
+# MoGe's utils3d dependency builds glcontext from source on ARM64. The current
+# runner image predates the baked-in X11 headers below, so keep this guarded
+# bootstrap until every deployed image digest contains libx11-dev.
+if [ ! -f /usr/include/X11/Xlib.h ]; then
+  apt-get -o Acquire::Retries=5 update
+  apt-get -o Acquire::Retries=5 install -y --no-install-recommends libx11-dev
+  rm -rf /var/lib/apt/lists/*
+fi
+
+uv pip install git+https://github.com/microsoft/MoGe.git
+uv pip install k_diffusion einops_exts alias_free_torch torchsde
+
+exec python fastvideo/tests/ssim/ci_runner.py "${args[@]}"

.buildkite/scripts/lanes/train_framework.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the modular training-framework lane.
+set -euo pipefail
+
+exec pytest ./fastvideo/tests/train/models ./fastvideo/tests/train/methods -vs

.buildkite/scripts/lanes/training.sh

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the legacy vanilla-training lane.
+set -euo pipefail
+
+export WANDB_MODE=offline
+exec pytest ./fastvideo/tests/training/Vanilla -srP

.buildkite/scripts/lanes/training_lora.sh

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the legacy LoRA-training lane.
+set -euo pipefail
+
+export WANDB_MODE=offline
+exec pytest ./fastvideo/tests/training/lora/test_lora_training.py -srP

.buildkite/scripts/lanes/training_vsa.sh

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the legacy VSA-training lane.
+set -euo pipefail
+
+export WANDB_MODE=offline
+exec pytest ./fastvideo/tests/training/VSA -srP

.buildkite/scripts/lanes/transformer.sh

@@ -0,0 +1,9 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the transformer lane.
+set -euo pipefail
+
+# The existing block reference records an absent FASTVIDEO_FA4 (FA2). Keep
+# that reference identity; the component lane also selects FA2 explicitly.
+env -u FASTVIDEO_FA4 pytest ./fastvideo/tests/golden_gate/test_wan_t2v.py -xvs
+pytest ./fastvideo/tests/golden_gate/test_wan_causal.py -xvs
+exec pytest ./fastvideo/tests/transformers -vs

.buildkite/scripts/lanes/vae.sh

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+# Canonical Slurm CI selection for the VAE lane.
+set -euo pipefail
+
+pytest ./fastvideo/tests/golden_gate/test_wan_vae.py -xvs
+exec pytest ./fastvideo/tests/vaes -vs

.buildkite/scripts/pr_test.sh

@@ -1,6 +1,19 @@
 #!/bin/bash
 set -uo pipefail
 
+# DORMANT ROLLBACK ONLY. Active CI is Slurm-only and pipeline.yml never calls
+# this launcher. Refuse every Buildkite invocation even if a stale step or
+# operator typo reaches this file; local rollback experiments require an
+# explicit opt-in.
+if [ -n "${BUILDKITE:-}" ]; then
+  echo "Legacy Modal CI is disabled; use the Slinky Slurm runner." >&2
+  exit 2
+fi
+if [ "${FASTVIDEO_ENABLE_LEGACY_MODAL_CI:-0}" != 1 ]; then
+  echo "Legacy Modal CI is dormant. Set FASTVIDEO_ENABLE_LEGACY_MODAL_CI=1 only for a manual rollback test." >&2
+  exit 2
+fi
+
 log() {
   echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1"
 }
@@ -76,7 +89,7 @@ EFFECTIVE_PR=${BUILDKITE_PULL_REQUEST:-false}
 if [ "$EFFECTIVE_PR" = "false" ] && [ -n "${PR_NUMBER:-}" ]; then
     EFFECTIVE_PR=$PR_NUMBER
 fi
-MODAL_ENV="BUILDKITE_REPO=$BUILDKITE_REPO BUILDKITE_COMMIT=$BUILDKITE_COMMIT BUILDKITE_PULL_REQUEST=$EFFECTIVE_PR BUILDKITE_BRANCH=${BUILDKITE_BRANCH:-} TEST_SCOPE=${TEST_SCOPE:-} BUILDKITE_BUILD_URL=${BUILDKITE_BUILD_URL:-} BUILDKITE_BUILD_ID=${BUILDKITE_BUILD_ID:-} BUILDKITE_JOB_ID=${BUILDKITE_JOB_ID:-} IMAGE_VERSION=$IMAGE_VERSION"
+MODAL_ENV="BUILDKITE_REPO=$BUILDKITE_REPO BUILDKITE_COMMIT=$BUILDKITE_COMMIT BUILDKITE_PULL_REQUEST=$EFFECTIVE_PR BUILDKITE_BRANCH=${BUILDKITE_BRANCH:-} BUILDKITE_SOURCE=${BUILDKITE_SOURCE:-} TEST_SCOPE=${TEST_SCOPE:-} BUILDKITE_BUILD_URL=${BUILDKITE_BUILD_URL:-} BUILDKITE_BUILD_ID=${BUILDKITE_BUILD_ID:-} BUILDKITE_JOB_ID=${BUILDKITE_JOB_ID:-} IMAGE_VERSION=$IMAGE_VERSION"
 
 POST_RUN_HOOK=""
 
@@ -187,6 +200,10 @@ case "$TEST_TYPE" in
         log "Running transformer tests..."
         MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_transformer_tests"
         ;;
+    "golden_gate")
+        log "Running golden-gate tests..."
+        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_golden_gate_tests"
+        ;;
     "ssim")
         log "Running SSIM tests..."
         SSIM_BOOTSTRAP_ARGS=$(ssim_bootstrap_args)

.buildkite/scripts/unit_test.sh

@@ -0,0 +1,41 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Collect the whole attention directory so new files cannot land uncovered.
+# Its FA2/FA3 regression files skip when FA4 is selected (the Modal image
+# enables FA4 by default), so pin FA4 off for the directory to be real
+# coverage on every runner rather than a nominal collection.
+export FASTVIDEO_FA4=0
+
+# The livestream app's tests are CPU-only; its single gpu-marked module is
+# deselected, and DreamVerse's GPU tests have their own lane.
+exec pytest \
+  ./apps/infinite_livestream/infinite_livestream/tests \
+  ./fastvideo/tests/api/ \
+  ./fastvideo/tests/contract/ \
+  ./fastvideo/tests/dataset/ \
+  ./fastvideo/tests/workflow/ \
+  ./fastvideo/tests/entrypoints/ \
+  ./fastvideo/tests/loader/ \
+  ./fastvideo/tests/pipelines/ \
+  ./fastvideo/tests/platforms/ \
+  ./fastvideo/tests/schedulers/ \
+  ./fastvideo/tests/train/ \
+  ./fastvideo/tests/stages/ \
+  ./fastvideo/tests/ops/ \
+  ./fastvideo/tests/worker/ \
+  ./fastvideo/tests/training/test_runner.py \
+  ./fastvideo/tests/training/test_trackers.py \
+  ./fastvideo/tests/inference/test_basic_fasth3_omniref_pdd.py \
+  ./fastvideo/tests/inference/test_inference_regional_compile.py \
+  ./fastvideo/tests/attention/ \
+  ./fastvideo/tests/layers/test_pdd_linear.py \
+  ./fastvideo/tests/layers/test_triton_fused_norm.py \
+  ./fastvideo/tests/modal/test_kernel_build_cache.py \
+  ./fastvideo/tests/modal/test_pr_test.py \
+  ./fastvideo/tests/modal/test_ssim_test.py \
+  --ignore=./fastvideo/tests/entrypoints/test_openai_api_integration.py \
+  --ignore=./fastvideo/tests/train/models \
+  --ignore=./fastvideo/tests/train/methods \
+  -m "not gpu" \
+  -vs

.github/PULL_REQUEST_TEMPLATE.md

@@ -8,10 +8,10 @@ PR TITLE: Must start with a type tag, e.g.:
 MERGE WORKFLOW:
   1. Ensure pre-commit passes and you have at least 1 approval
   2. Comment /merge (or add the "ready" label) to enter the Merge Queue
-  3. Full Test Suite runs automatically on a staging branch → auto-merge on success
+  3. A path-aware merge gate runs only relevant integration tests → auto-merge on success
 
 ON-DEMAND TESTING (write access required):
-  /test full       — Full Test Suite        /test ssim        — SSIM regression
+  /test full       — Explicit all-lane run  /test ssim        — Full SSIM regression
   /test training   — Training pipeline      /test encoder     — Encoder tests
   /test transformer — Transformer tests     /test vae         — VAE tests
   /test kernel     — CUDA kernel tests      /test unit        — Unit tests

.github/scripts/gate_full_suite.sh

@@ -0,0 +1,133 @@
+#!/usr/bin/env bash
+# Gate the path-aware Buildkite merge plan on the cheap GitHub checks.
+#
+# Polls the workflow runs for the PR head commit and only exits 0 once the
+# watched cheap workflows (pre-commit, docs build) have succeeded, so the
+# 'ready' label cannot burn path-selected GPU lanes on a head that a cheap
+# check has already doomed.
+#
+# Semantics:
+#   - watched run completed with a bad conclusion  -> exit 1 (fail CLOSED:
+#     no merge gate; the next push re-arms via the 'synchronize' trigger)
+#   - watched run cancelled                         -> still pending: the docs
+#     workflow's repo-global 'pages' concurrency group cancels runs superseded
+#     by unrelated pushes, so 'cancelled' is not a verdict on this PR
+#   - watched runs pending                          -> poll until done
+#   - docs run absent                               -> not applicable after a
+#     short grace period ('Deploy Documentation' is path-filtered on PRs)
+#   - pre-commit run absent                         -> keep polling: pre-commit
+#     is never path-filtered, so its absence is always anomalous
+#   - 'ready' label removed while waiting           -> exit 1 (fail CLOSED:
+#     un-labeling is a deliberate maintainer action)
+#   - GitHub API unreachable or timeout             -> exit 0 (fail OPEN,
+#     loud warning: never brick CI on a GitHub outage)
+#
+# Required env: PR_SHA (PR head commit), PR_NUMBER, GITHUB_REPOSITORY, GH_TOKEN.
+set -euo pipefail
+
+: "${PR_SHA:?PR_SHA (PR head commit) is required}"
+: "${PR_NUMBER:?PR_NUMBER (pull request number) is required}"
+: "${GITHUB_REPOSITORY:?GITHUB_REPOSITORY is required}"
+
+# Workflow-level `name:` values that must be green before the merge gate
+# may start. "Deploy Documentation" is path-filtered on PRs, so its run may
+# legitimately never exist; pre-commit always runs, so it must appear.
+WATCHED_NAMES='["pre-commit", "Deploy Documentation"]'
+WATCHED_REGEX='^(pre-commit|Deploy Documentation)$'
+POLL_SECS="${POLL_SECS:-20}"
+GRACE_SECS="${GRACE_SECS:-60}"
+MAX_WAIT_SECS="${MAX_WAIT_SECS:-1500}"
+
+# Bound each API call so a hung connection hits the 3-strike fail-open path
+# instead of pinning the loop until the job timeout (which would fail closed
+# on exactly the GitHub-outage case this script is meant to survive).
+if command -v timeout >/dev/null 2>&1; then
+  gh_api() { timeout 30 gh api "$@"; }
+else
+  gh_api() { gh api "$@"; }  # macOS dev boxes; CI always has coreutils timeout
+fi
+
+# The workflow checked the label before starting the gate, but the wait can
+# last ~25 min: re-check once before any exit 0 and fail closed if 'ready'
+# was removed in the meantime. An API error here proceeds (the label was
+# present when the gate started; never brick CI on an outage).
+recheck_ready_label() {
+  local pr_json
+  if pr_json=$(gh_api "repos/${GITHUB_REPOSITORY}/pulls/${PR_NUMBER}" 2>/dev/null); then
+    if ! jq -e '[.labels[]?.name] | index("ready")' <<<"$pr_json" >/dev/null 2>&1; then
+      echo "::error::PR #${PR_NUMBER} no longer has the 'ready' label —" \
+        "NOT triggering the Buildkite merge gate. Re-add the label to re-arm."
+      exit 1
+    fi
+  else
+    echo "::warning::Could not re-check the 'ready' label on PR #${PR_NUMBER}; proceeding (it was present when the gate started)."
+  fi
+}
+
+start=$(date +%s)
+api_fails=0
+missing=""
+
+while true; do
+  elapsed=$(( $(date +%s) - start ))
+
+  if runs_json=$(gh_api "repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=${PR_SHA}&per_page=100" 2>/dev/null) \
+     && state=$(jq --arg re "$WATCHED_REGEX" '
+          [.workflow_runs[]? | select(.name // "" | test($re))]
+          | group_by(.name) | map(max_by(.id))
+          | map({name, status, conclusion})' <<<"$runs_json" 2>/dev/null); then
+    api_fails=0
+    echo "t+${elapsed}s watched checks: $(jq -c . <<<"$state")"
+
+    failed=$(jq -r '[.[] | select(.status == "completed"
+          and (.conclusion | IN("success", "skipped", "neutral", "cancelled") | not))]
+        | map(.name) | join(", ")' <<<"$state")
+    if [ -n "$failed" ]; then
+      echo "::error::Cheap check(s) failed on ${PR_SHA}: ${failed}." \
+        "NOT triggering the Buildkite merge gate. Push a fix (the 'ready'" \
+        "label re-arms on every push), or re-run the failed check and then" \
+        "re-run this workflow."
+      exit 1
+    fi
+
+    # 'cancelled' counts as pending: wait for a re-run to reach a real verdict
+    # (bounded by MAX_WAIT, then the fail-open below).
+    pending=$(jq '[.[] | select(.status != "completed" or .conclusion == "cancelled")] | length' <<<"$state")
+    missing=$(jq -r --argjson watched "$WATCHED_NAMES" '($watched - map(.name)) | join(", ")' <<<"$state")
+    if [ "$pending" -eq 0 ]; then
+      if [ -z "$missing" ]; then
+        recheck_ready_label
+        echo "All watched cheap checks are green — merge gate may proceed."
+        exit 0
+      fi
+      case "$missing" in
+        *pre-commit*)
+          echo "pre-commit run not found for ${PR_SHA} yet; waiting (pre-commit is never path-filtered, so its absence is anomalous)."
+          ;;
+        *)
+          if [ "$elapsed" -ge "$GRACE_SECS" ]; then
+            recheck_ready_label
+            echo "::warning::Watched run(s) never appeared for ${PR_SHA}: ${missing} (path-filtered, likely not applicable). Proceeding on the checks that did run."
+            exit 0
+          fi
+          echo "Waiting up to ${GRACE_SECS}s grace for path-filtered run(s) to appear: ${missing}."
+          ;;
+      esac
+    fi
+  else
+    api_fails=$(( api_fails + 1 ))
+    echo "::warning::GitHub API error querying workflow runs for ${PR_SHA} (attempt ${api_fails}/3)."
+    if [ "$api_fails" -ge 3 ]; then
+      recheck_ready_label
+      echo "::warning::FAILING OPEN: cannot query GitHub check status — triggering the merge gate WITHOUT the cheap-check gate."
+      exit 0
+    fi
+  fi
+
+  if [ "$elapsed" -ge "$MAX_WAIT_SECS" ]; then
+    recheck_ready_label
+    echo "::warning::FAILING OPEN: watched checks still pending after $(( MAX_WAIT_SECS / 60 )) min${missing:+ (never appeared: ${missing})} — triggering the merge gate anyway."
+    exit 0
+  fi
+  sleep "$POLL_SECS"
+done

.github/scripts/plan_merge_ci.py

@@ -0,0 +1,592 @@
+#!/usr/bin/env python3
+"""Select the additive GPU integration lanes needed by a PR diff.
+
+Fastcheck is the universal six-lane baseline and is intentionally not repeated
+here.  This planner selects only the more expensive merge-gate lanes.  Unknown
+source/build paths fail closed to the complete integration set, while explicit
+documentation and repository-metadata paths require no additional GPU work.
+"""
+
+from __future__ import annotations
+
+import argparse
+import fnmatch
+import re
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import TextIO
+
+MERGE_LANES = (
+    "golden-gate",
+    "ssim",
+    "lora-inference",
+    "lora-extraction",
+    "training",
+    "distillation",
+    "self-forcing",
+    "lora-training",
+    "training-vsa",
+    "inference-vmoba",
+    "performance",
+    "api-server",
+    "train-framework",
+    "eval",
+)
+
+LANE_SCRIPT_TO_KEY = {
+    "api_server.sh": "api-server",
+    "distillation_dmd.sh": "distillation",
+    "eval.sh": "eval",
+    "golden_gate.sh": "golden-gate",
+    "inference_lora.sh": "lora-inference",
+    "inference_vmoba.sh": "inference-vmoba",
+    "lora_extraction.sh": "lora-extraction",
+    "performance.sh": "performance",
+    "self_forcing.sh": "self-forcing",
+    "ssim.sh": "ssim",
+    "train_framework.sh": "train-framework",
+    "training.sh": "training",
+    "training_lora.sh": "lora-training",
+    "training_vsa.sh": "training-vsa",
+}
+
+FASTCHECK_LANE_SCRIPTS = {
+    "dreamverse.sh",
+    "encoder.sh",
+    "kernel_tests.sh",
+    "transformer.sh",
+    "vae.sh",
+}
+
+LEGACY_TRAINING_LANES = (
+    "training",
+    "distillation",
+    "self-forcing",
+    "lora-training",
+    "training-vsa",
+)
+
+ALL_TRAINING_LANES = (*LEGACY_TRAINING_LANES, "train-framework")
+
+SSIM_SMOKE_TESTS = (
+    "test_flux_t2i_similarity.py",
+    "test_wan_t2v_similarity.py",
+)
+
+SAFE_PATTERNS = (
+    "*.md",
+    "*.rst",
+    ".agents/**",
+    ".claude/**",
+    ".codex/**",
+    ".github/ISSUE_TEMPLATE/**",
+    ".github/PULL_REQUEST_TEMPLATE.md",
+    ".github/dependabot.yml",
+    ".github/mergify.yml",
+    ".github/scripts/**",
+    ".github/workflows/**",
+    ".buildkite/scripts/pre_commit.sh",
+    ".git-blame-ignore-revs",
+    ".gitattributes",
+    ".gitignore",
+    ".pre-commit-config.yaml",
+    "AGENTS.md",
+    "CITATION.cff",
+    "CODE_OF_CONDUCT.md",
+    "CONTRIBUTING.md",
+    "LICENSE",
+    "NOTICE",
+    "__init__.py",
+    "collect_env.py",
+    "SECURITY.md",
+    "assets/**",
+    "comfyui/**",
+    "docs/**",
+    "examples/**",
+    "mkdocs.yml",
+    "requirements-mkdocs.in",
+    "requirements-mkdocs.txt",
+    "scripts/**",
+    "tests/__init__.py",
+    "tests/local_tests/**",
+)
+
+ALL_IMPACT_PATTERNS = (
+    ".buildkite/pipeline.yml",
+    "docker/**",
+    "pyproject.toml",
+    "requirements*.txt",
+    "setup.cfg",
+    "setup.py",
+    "uv.lock",
+)
+
+
+@dataclass(frozen=True)
+class FamilyCoverage:
+    pattern: re.Pattern[str]
+    golden_tests: tuple[str, ...]
+    ssim_tests: tuple[str, ...]
+
+
+FAMILY_COVERAGE = (
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])dreamx(_world)?([/_.-]|$)"),
+        ("test_dreamx.py", ),
+        ("test_dreamx_world_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])flux[_-]?2([/_.-]|$)"),
+        ("test_flux2_klein.py", ),
+        ("test_flux2_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])flux(?![_-]?2)([/_.-]|$)"),
+        ("test_flux.py", ),
+        ("test_flux_t2i_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])(hunyuan)?gamecraft([/_.-]|$)"),
+        ("test_gamecraft.py", ),
+        ("test_gamecraft_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])gen3c([/_.-]|$)"),
+        ("test_gen3c.py", ),
+        ("test_gen3c_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])glm[_-]?image([/_.-]|$)"),
+        ("test_glm_image.py", ),
+        ("test_glm_image_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])hunyuan(video)?15([a-z0-9_-]*)([/_.-]|$)"),
+        (),
+        ("test_hunyuan15_i2v_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])kandinsky[_-]?5([/_.-]|$)"),
+        ("test_kandinsky5.py", ),
+        ("test_kandinsky5_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])lingbot([a-z0-9_-]*)([/_.-]|$)"),
+        ("test_lingbot.py", ),
+        ("test_lingbot_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])longcat([/_.-]|$)"),
+        ("test_longcat.py", ),
+        ("test_longcat_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])ltx[_-]?2([/_.-]|$)"),
+        ("test_ltx2.py", ),
+        ("test_ltx2_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])matrixgame[_-]?2([/_.-]|$)"),
+        ("test_matrixgame.py", ),
+        ("test_matrixgame2_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])matrixgame[_-]?3([/_.-]|$)"),
+        ("test_matrixgame.py", ),
+        ("test_matrixgame3_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])minimax[_-]?h3([/_.-]|$)"),
+        ("test_minimax_h3_t2v.py", ),
+        ("test_minimax_h3_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])sd[_-]?3([._-]?5)?([/_.-]|$)"),
+        ("test_sd35.py", ),
+        ("test_sd35_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])stable[_-]?audio([/_.-]|$)"),
+        ("test_stable_audio.py", ),
+        ("test_stable_audio_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])turbo(diffusion)?([/_.-]|$)"),
+        (),
+        ("test_turbodiffusion_similarity.py", ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])wan(video|vae)?([/_.-]|$)"),
+        ("test_wan_t2v.py", "test_wan_vae.py", "test_wan_causal.py", "test_wan_denoising.py"),
+        (
+            "test_causal_similarity.py",
+            "test_wan_i2v_similarity.py",
+            "test_wan_t2v_similarity.py",
+            "test_wan_ti2v_similarity.py",
+        ),
+    ),
+    FamilyCoverage(
+        re.compile(r"(^|[/_.-])z[_-]?image([/_.-]|$)"),
+        ("test_zimage.py", ),
+        ("test_zimage_similarity.py", ),
+    ),
+)
+
+
+@dataclass
+class MergePlan:
+    lanes: set[str] = field(default_factory=set)
+    golden_tests: set[str] = field(default_factory=set)
+    ssim_tests: set[str] = field(default_factory=set)
+    golden_all: bool = False
+    ssim_all: bool = False
+    reasons: list[str] = field(default_factory=list)
+
+    def add_lanes(self, *lanes: str, reason: str) -> None:
+        unknown = set(lanes) - set(MERGE_LANES)
+        if unknown:
+            raise ValueError(f"Unknown merge lanes: {sorted(unknown)}")
+        self.lanes.update(lanes)
+        self.reasons.append(reason)
+
+    def add_golden(self, tests: tuple[str, ...], reason: str) -> None:
+        self.add_lanes("golden-gate", reason=reason)
+        self.golden_tests.update(tests)
+
+    def add_ssim(self, tests: tuple[str, ...], reason: str) -> None:
+        self.add_lanes("ssim", reason=reason)
+        self.ssim_tests.update(tests)
+
+    def require_all(self, reason: str) -> None:
+        self.lanes.update(MERGE_LANES)
+        self.golden_all = True
+        self.ssim_all = True
+        self.reasons.append(reason)
+
+    def ordered_lanes(self) -> tuple[str, ...]:
+        return tuple(lane for lane in MERGE_LANES if lane in self.lanes)
+
+    def encoded_lanes(self) -> str:
+        lanes = self.ordered_lanes()
+        return "," + ",".join(lanes or ("none", )) + ","
+
+    def encoded_golden_tests(self) -> str:
+        if "golden-gate" not in self.lanes:
+            return "none"
+        if self.golden_all or not self.golden_tests:
+            return "all"
+        return ",".join(sorted(self.golden_tests))
+
+    def encoded_ssim_tests(self) -> str:
+        if "ssim" not in self.lanes:
+            return "none"
+        if self.ssim_all or not self.ssim_tests:
+            return "all"
+        return ",".join(sorted(self.ssim_tests))
+
+
+def _matches_any(path: str, patterns: tuple[str, ...]) -> bool:
+    return any(fnmatch.fnmatchcase(path, pattern) for pattern in patterns)
+
+
+def _family_coverage(path: str) -> tuple[set[str], set[str]]:
+    normalized = path.lower()
+    golden: set[str] = set()
+    ssim: set[str] = set()
+    for family in FAMILY_COVERAGE:
+        if family.pattern.search(normalized):
+            golden.update(family.golden_tests)
+            ssim.update(family.ssim_tests)
+    # Select the component actually touched, including compatibility paths.
+    # Family configs/pipeline wiring can affect all four Wan gates.
+    if re.search(r"(^|[/_.-])wan(video|vae)?([/_.-]|$)", normalized):
+        if (normalized.endswith(("/wan/vae.py", "/wan/vae_config.py", "/vaes/wanvae.py"))
+                or normalized.endswith("/wan/stages/conditioning.py")):
+            golden = {"test_wan_vae.py"}
+        elif normalized.endswith(("/wan/causal_transformer.py", "/dits/causal_wanvideo.py",
+                                  "/wan/stages/causal_denoising.py")):
+            golden = {"test_wan_causal.py"}
+        elif (normalized == "fastvideo/models/dits/wanvideo.py"
+              or normalized.endswith(("/wan/transformer.py", "/wan/stages/denoising.py", "/wan/stages/dmd.py"))):
+            golden = {"test_wan_t2v.py", "test_wan_denoising.py"}
+    return golden, ssim
+
+
+def _select_output_coverage(plan: MergePlan, path: str) -> None:
+    golden, ssim = _family_coverage(path)
+    if golden:
+        plan.add_golden(tuple(sorted(golden)), reason=f"model-family golden coverage: {path}")
+    else:
+        plan.golden_all = True
+        plan.add_lanes("golden-gate", reason=f"shared output golden coverage: {path}")
+    if ssim:
+        plan.add_ssim(tuple(sorted(ssim)), reason=f"model-family SSIM coverage: {path}")
+    else:
+        plan.add_ssim(SSIM_SMOKE_TESTS, reason=f"shared output SSIM smoke coverage: {path}")
+
+
+def classify_paths(paths: list[str]) -> MergePlan:
+    plan = MergePlan()
+    normalized_paths: list[str] = []
+    for raw_path in paths:
+        path = raw_path.strip()
+        while path.startswith("./"):
+            path = path[2:]
+        if path:
+            normalized_paths.append(path)
+    normalized_paths = sorted(set(normalized_paths))
+    if not normalized_paths:
+        plan.require_all("changed-file list was empty; failing closed")
+        return plan
+
+    for path in normalized_paths:
+        if path == "__FASTVIDEO_CI_PLAN_ALL__":
+            plan.require_all("changed-file API failed; failing closed")
+            continue
+
+        if path in {"requirements-mkdocs.in", "requirements-mkdocs.txt"}:
+            plan.reasons.append(f"documentation dependencies need no GPU integration: {path}")
+            continue
+
+        if _matches_any(path, ALL_IMPACT_PATTERNS):
+            plan.require_all(f"cross-cutting build/runtime surface: {path}")
+            continue
+
+        lane_script_prefix = ".buildkite/scripts/lanes/"
+        if path.startswith(lane_script_prefix):
+            script_name = Path(path).name
+            lane = LANE_SCRIPT_TO_KEY.get(script_name)
+            if lane is None:
+                if script_name in FASTCHECK_LANE_SCRIPTS:
+                    plan.reasons.append(f"covered by automatic Fastcheck lane: {path}")
+                else:
+                    plan.require_all(f"unknown lane script: {path}")
+            elif lane == "golden-gate":
+                plan.golden_all = True
+                plan.add_lanes(lane, reason=f"golden lane implementation: {path}")
+            elif lane == "ssim":
+                plan.ssim_all = True
+                plan.add_lanes(lane, reason=f"SSIM lane implementation: {path}")
+            else:
+                plan.add_lanes(lane, reason=f"lane implementation: {path}")
+            continue
+
+        if path.startswith("fastvideo/tests/golden_gate/"):
+            name = Path(path).name
+            if name.startswith("test_") and name.endswith(".py"):
+                plan.add_golden((name, ), reason=f"changed golden test: {path}")
+            elif name in {"AGENTS.md", "README.md"}:
+                plan.reasons.append(f"golden documentation only: {path}")
+            else:
+                plan.golden_all = True
+                plan.add_lanes("golden-gate", reason=f"shared golden harness/reference: {path}")
+            continue
+
+        if path.startswith("fastvideo/tests/ssim/"):
+            name = Path(path).name
+            if name.startswith("test_") and name.endswith(".py"):
+                plan.add_ssim((name, ), reason=f"changed SSIM test: {path}")
+            elif path.endswith((".py", ".json", ".pt", ".png", ".mp4")):
+                plan.ssim_all = True
+                plan.add_lanes("ssim", reason=f"shared SSIM harness/reference: {path}")
+            continue
+
+        if path.startswith("fastvideo/tests/performance/") or path.startswith(".buildkite/performance-benchmarks/"):
+            plan.add_lanes("performance", reason=f"performance coverage: {path}")
+            continue
+        if path.startswith(("fastvideo/performance/", "fastvideo/performance_dashboard/",
+                            "apps/performance_dashboard/")):
+            plan.add_lanes("performance", reason=f"performance implementation: {path}")
+            continue
+        if path.startswith("fastvideo/benchmarks/"):
+            if "/mlx_" in path or Path(path).name.startswith("mlx_"):
+                plan.reasons.append(f"covered by the path-filtered macOS MLX workflow: {path}")
+            else:
+                plan.add_lanes("performance", reason=f"benchmark implementation: {path}")
+            continue
+        if path.startswith("fastvideo/tests/eval/") or path.startswith("fastvideo/eval/"):
+            plan.add_lanes("eval", reason=f"evaluation coverage: {path}")
+            continue
+        if path.startswith("fastvideo/third_party/eval/"):
+            plan.add_lanes("eval", reason=f"vendored evaluation implementation: {path}")
+            continue
+        if path.startswith("fastvideo/tests/lora_extraction/") or path.startswith("scripts/lora_extraction/"):
+            plan.add_lanes("lora-extraction", reason=f"LoRA extraction coverage: {path}")
+            continue
+        if path.startswith("fastvideo/tests/inference/lora/"):
+            plan.add_lanes("lora-inference", reason=f"LoRA inference coverage: {path}")
+            continue
+        if path.startswith("fastvideo/tests/inference/vmoba/"):
+            plan.add_lanes("inference-vmoba", reason=f"VMoBA inference coverage: {path}")
+            continue
+        if path.startswith(("fastvideo/dataset/", "fastvideo/workflow/", "fastvideo/pipelines/preprocess/",
+                            "fastvideo/pipelines/training/")):
+            plan.add_lanes(*ALL_TRAINING_LANES, reason=f"shared data/training input surface: {path}")
+            continue
+        if path.startswith("fastvideo/tests/train/") or path.startswith("fastvideo/train/"):
+            plan.add_lanes("train-framework", reason=f"modular training coverage: {path}")
+            continue
+
+        if path.startswith("fastvideo/tests/training/"):
+            lowered = path.lower()
+            if "/vanilla/" in lowered:
+                plan.add_lanes("training", reason=f"vanilla training coverage: {path}")
+            elif "/distill/" in lowered:
+                plan.add_lanes("distillation", reason=f"distillation coverage: {path}")
+            elif "/self-forcing/" in lowered:
+                plan.add_lanes("self-forcing", reason=f"self-forcing coverage: {path}")
+            elif "/lora/" in lowered:
+                plan.add_lanes("lora-training", reason=f"LoRA training coverage: {path}")
+            elif "/vsa/" in lowered:
+                plan.add_lanes("training-vsa", reason=f"VSA training coverage: {path}")
+            else:
+                plan.add_lanes(*LEGACY_TRAINING_LANES, reason=f"shared legacy training coverage: {path}")
+            continue
+
+        if path.startswith("fastvideo/training/"):
+            lowered = path.lower()
+            if "self_forcing" in lowered:
+                plan.add_lanes("self-forcing", reason=f"self-forcing implementation: {path}")
+            elif "distill" in lowered:
+                plan.add_lanes("distillation", reason=f"distillation implementation: {path}")
+            elif "lora" in lowered:
+                plan.add_lanes("lora-training", reason=f"LoRA training implementation: {path}")
+            else:
+                plan.add_lanes(*LEGACY_TRAINING_LANES, reason=f"shared legacy training implementation: {path}")
+            continue
+
+        lowered = path.lower()
+        if "vmoba" in lowered and path.startswith(("fastvideo/", ".buildkite/")):
+            plan.add_lanes("inference-vmoba", reason=f"VMoBA implementation: {path}")
+            plan.add_golden(("test_wan_t2v.py", ), reason=f"VMoBA end-to-end coverage: {path}")
+            continue
+        if "lora" in lowered and path.startswith("fastvideo/"):
+            plan.add_lanes(
+                "lora-inference",
+                "lora-extraction",
+                "lora-training",
+                reason=f"shared LoRA implementation: {path}",
+            )
+            _select_output_coverage(plan, path)
+            continue
+
+        if path.startswith("fastvideo/entrypoints/") or path.startswith("fastvideo/api/"):
+            plan.add_lanes("api-server", reason=f"API/entrypoint integration: {path}")
+            if "openai" not in lowered and "/cli/" not in lowered:
+                _select_output_coverage(plan, path)
+            continue
+        if path.startswith("fastvideo/worker/"):
+            plan.add_lanes("api-server", reason=f"worker/API integration: {path}")
+            _select_output_coverage(plan, path)
+            continue
+        if path.startswith("fastvideo/distributed/"):
+            plan.add_lanes(
+                "training",
+                "train-framework",
+                reason=f"distributed runtime integration: {path}",
+            )
+            _select_output_coverage(plan, path)
+            continue
+        if path.startswith(("fastvideo/hooks/", "fastvideo/platforms/", "fastvideo/third_party/")):
+            _select_output_coverage(plan, path)
+            continue
+        if path.startswith(("fastvideo/models/", "fastvideo/pipelines/", "fastvideo/configs/",
+                            "fastvideo/layers/", "fastvideo/attention/")):
+            _select_output_coverage(plan, path)
+            continue
+        if path in {
+                "fastvideo/fastvideo_args.py",
+                "fastvideo/forward_context.py",
+                "fastvideo/image_processor.py",
+                "fastvideo/registry.py",
+                "fastvideo/utils.py",
+        }:
+            _select_output_coverage(plan, path)
+            continue
+        if path.startswith("fastvideo/mlx_runtime/"):
+            plan.reasons.append(f"covered by the path-filtered macOS MLX workflow: {path}")
+            continue
+        if path.startswith("fastvideo/logging_utils/") or path in {
+                "fastvideo/__init__.py",
+                "fastvideo/envs.py",
+                "fastvideo/logger.py",
+                "fastvideo/profiler.py",
+                "fastvideo/version.py",
+        }:
+            plan.reasons.append(f"covered by automatic Fastcheck: {path}")
+            continue
+        if path.startswith(("fastvideo-kernel/", "csrc/")):
+            plan.add_golden(("test_wan_t2v.py", ), reason=f"kernel integration smoke: {path}")
+            plan.add_ssim(("test_wan_t2v_similarity.py", ), reason=f"kernel numerical smoke: {path}")
+            continue
+
+        if path.startswith("apps/dreamverse/"):
+            # DreamVerse is already one of the six automatic Fastcheck lanes.
+            plan.reasons.append(f"covered by automatic DreamVerse Fastcheck: {path}")
+            continue
+        if path.startswith("apps/infinite_livestream/"):
+            # The app's CPU-only tests run in the automatic unit Fastcheck lane.
+            plan.reasons.append(f"covered by automatic unit Fastcheck: {path}")
+            continue
+        if path.startswith("fastvideo/tests/"):
+            # The automatic unit/component Fastcheck lanes own the remaining
+            # package tests. Domain-specific expensive test roots were handled
+            # above.
+            plan.reasons.append(f"covered by automatic Fastcheck: {path}")
+            continue
+        if path in {".buildkite/scripts/unit_test.sh", ".buildkite/scripts/pr_test.sh"}:
+            plan.reasons.append(f"covered by automatic unit Fastcheck: {path}")
+            continue
+        if _matches_any(path, SAFE_PATTERNS):
+            plan.reasons.append(f"no additional GPU integration needed: {path}")
+            continue
+
+        plan.require_all(f"unclassified path; failing closed: {path}")
+
+    return plan
+
+
+def _write_github_output(output: TextIO, plan: MergePlan) -> None:
+    output.write(f"merge_test_plan={plan.encoded_lanes()}\n")
+    output.write(f"merge_golden_tests={plan.encoded_golden_tests()}\n")
+    output.write(f"merge_ssim_tests={plan.encoded_ssim_tests()}\n")
+    output.write(f"merge_plan_label={','.join(plan.ordered_lanes()) or 'none'}\n")
+
+
+def _write_summary(output: TextIO, plan: MergePlan) -> None:
+    output.write("## Change-aware merge test plan\n\n")
+    output.write("| Selection | Value |\n|---|---|\n")
+    output.write(f"| Additional Slurm lanes | `{','.join(plan.ordered_lanes()) or 'none'}` |\n")
+    output.write(f"| Golden tests | `{plan.encoded_golden_tests()}` |\n")
+    output.write(f"| SSIM tests | `{plan.encoded_ssim_tests()}` |\n\n")
+    output.write("Fastcheck remains the universal six-lane baseline.\n")
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--paths-file", type=Path, required=True)
+    parser.add_argument("--github-output", type=Path)
+    parser.add_argument("--summary-file", type=Path)
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    paths = args.paths_file.read_text(encoding="utf-8").splitlines()
+    plan = classify_paths(paths)
+    print(f"MERGE_TEST_PLAN={plan.encoded_lanes()}")
+    print(f"MERGE_GOLDEN_TESTS={plan.encoded_golden_tests()}")
+    print(f"MERGE_SSIM_TESTS={plan.encoded_ssim_tests()}")
+    for reason in plan.reasons:
+        print(f"- {reason}")
+    if args.github_output:
+        with args.github_output.open("a", encoding="utf-8") as output:
+            _write_github_output(output, plan)
+    if args.summary_file:
+        with args.summary_file.open("a", encoding="utf-8") as output:
+            _write_summary(output, plan)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

.github/scripts/test_gate_full_suite.sh

@@ -0,0 +1,122 @@
+#!/usr/bin/env bash
+# Self-test for gate_full_suite.sh using a mocked `gh`. No network, runs on
+# any dev box: bash .github/scripts/test_gate_full_suite.sh
+set -u
+here=$(cd "$(dirname "$0")" && pwd)
+tmp=$(mktemp -d)
+trap 'rm -rf "$tmp"' EXIT
+
+# Mock gh. Asserts the exact endpoint (including head_sha) it is called
+# with — an endpoint typo in the gate script fails the test rather than
+# silently serving canned data. On the runs endpoint it serves
+# $MOCK_DIR/response_<call#>.json, sticking on the highest existing file,
+# and exits 1 if none exist (simulates a GitHub API outage). On the pulls
+# endpoint it serves $MOCK_DIR/pr.json, defaulting to a 'ready'-labeled PR.
+cat > "$tmp/gh" <<'EOF'
+#!/usr/bin/env bash
+if [ "${1:-}" != "api" ]; then
+  echo "unexpected gh invocation: $*" >> "$MOCK_DIR/endpoint_error"
+  exit 2
+fi
+case "${2:-}" in
+  "repos/o/r/actions/runs?head_sha=deadbeef&per_page=100")
+    n=$(( $(cat "$MOCK_DIR/count" 2>/dev/null || echo 0) + 1 ))
+    echo "$n" > "$MOCK_DIR/count"
+    while [ "$n" -gt 0 ]; do
+      if [ -f "$MOCK_DIR/response_$n.json" ]; then
+        cat "$MOCK_DIR/response_$n.json"
+        exit 0
+      fi
+      n=$(( n - 1 ))
+    done
+    echo "api outage" >&2
+    exit 1
+    ;;
+  "repos/o/r/pulls/42")
+    if [ -f "$MOCK_DIR/pr.json" ]; then
+      cat "$MOCK_DIR/pr.json"
+    else
+      echo '{"labels": [{"name": "ready"}]}'
+    fi
+    ;;
+  *)
+    echo "unexpected gh endpoint: $2" >> "$MOCK_DIR/endpoint_error"
+    exit 2
+    ;;
+esac
+EOF
+chmod +x "$tmp/gh"
+
+PC_OK='{"name": "pre-commit", "id": 1, "status": "completed", "conclusion": "success"}'
+PC_BAD='{"name": "pre-commit", "id": 1, "status": "completed", "conclusion": "failure"}'
+PC_PENDING='{"name": "pre-commit", "id": 1, "status": "in_progress", "conclusion": null}'
+DOCS_OK='{"name": "Deploy Documentation", "id": 2, "status": "completed", "conclusion": "success"}'
+DOCS_BAD='{"name": "Deploy Documentation", "id": 2, "status": "completed", "conclusion": "failure"}'
+DOCS_CANCELLED='{"name": "Deploy Documentation", "id": 2, "status": "completed", "conclusion": "cancelled"}'
+OTHER='{"name": "Trigger Merge Gate", "id": 3, "status": "in_progress", "conclusion": null}'
+NULL_NAME='{"name": null, "id": 4, "status": "completed", "conclusion": "failure"}'
+PC_OK_RERUN='{"name": "pre-commit", "id": 5, "status": "completed", "conclusion": "success"}'
+
+fails=0
+want_log=""  # optional: expect() also greps out.log for this regex, then resets
+pr_json=""   # optional: served for the pulls (label re-check) endpoint, then resets
+raw_body=""  # optional: serve responses verbatim instead of wrapping in workflow_runs
+expect() { # <name> <expected-exit> <response json>...
+  local name=$1 want=$2 dir i=1
+  shift 2
+  dir=$(mktemp -d "$tmp/test_XXXXXX")
+  for body in "$@"; do
+    if [ -n "$raw_body" ]; then
+      printf '%s' "$body" > "$dir/response_$i.json"
+    else
+      printf '{"workflow_runs": [%s]}' "$body" > "$dir/response_$i.json"
+    fi
+    i=$(( i + 1 ))
+  done
+  [ -n "$pr_json" ] && printf '%s' "$pr_json" > "$dir/pr.json"
+  ( export PATH="$tmp:$PATH" MOCK_DIR="$dir" PR_SHA=deadbeef PR_NUMBER=42 \
+      GITHUB_REPOSITORY=o/r POLL_SECS=0 GRACE_SECS=1 MAX_WAIT_SECS=3
+    bash "$here/gate_full_suite.sh" > "$dir/out.log" 2>&1 )
+  local rc=$?
+  if [ "$rc" -ne "$want" ]; then
+    echo "FAIL: $name (exit $rc, want $want)"
+    cat "$dir/out.log"
+    fails=1
+  elif [ -f "$dir/endpoint_error" ]; then
+    echo "FAIL: $name (mock gh got an unexpected call)"
+    cat "$dir/endpoint_error"
+    fails=1
+  elif [ -n "$want_log" ] && ! grep -Eq "$want_log" "$dir/out.log"; then
+    echo "FAIL: $name (log does not match: $want_log)"
+    cat "$dir/out.log"
+    fails=1
+  else
+    echo "ok: $name"
+  fi
+  want_log="" pr_json="" raw_body=""
+}
+
+expect "both green -> proceed" 0 "$PC_OK, $DOCS_OK, $OTHER, $NULL_NAME"
+expect "docs build failed -> blocked" 1 "$PC_OK, $DOCS_BAD"
+expect "pre-commit failed -> blocked" 1 "$PC_BAD"
+expect "pending then green -> proceed" 0 "$PC_PENDING" "$PC_OK, $DOCS_OK"
+want_log="never appeared.*Deploy Documentation"
+expect "docs run absent (path-filtered) -> proceed after grace" 0 "$PC_OK"
+expect "API outage -> fail open" 0
+want_log="FAILING OPEN"
+expect "pending past MAX_WAIT -> fail open" 0 "$PC_PENDING"
+want_log="FAILING OPEN"
+expect "unrelated runs only -> no grace, fail open at MAX_WAIT" 0 "$OTHER"
+expect "cancelled docs then green -> proceed" 0 \
+  "$PC_OK, $DOCS_CANCELLED" "$PC_OK, $DOCS_OK"
+want_log="FAILING OPEN"
+expect "cancelled docs forever -> fail open at MAX_WAIT" 0 "$PC_OK, $DOCS_CANCELLED"
+want_log="FAILING OPEN"
+expect "pre-commit absent -> no grace, fail open at MAX_WAIT" 0 "$DOCS_OK"
+expect "duplicate run names -> latest wins" 0 "$PC_BAD, $PC_OK_RERUN, $DOCS_OK"
+raw_body=1
+expect "garbage response body -> fail open" 0 "this is not json"
+pr_json='{"labels": [{"name": "other"}]}'
+expect "ready label removed mid-gate -> blocked" 1 "$PC_OK, $DOCS_OK"
+
+exit "$fails"

.github/workflows/_template-build-image.yml

@@ -190,6 +190,7 @@ jobs:
         if: ${{ !inputs.push_by_digest }}
         run: |
           echo "✅ Python ${{ inputs.python_version }} image successfully built and pushed to ${{ steps.image.outputs.name }}:${{ inputs.tag_suffix }}-sha-${GITHUB_SHA::7}"
+          echo "Digest: ${{ steps.build-push.outputs.digest }}"
           echo "To run tests with this image, manually trigger the 'Run Tests' workflow."
 
       - name: Digest success message

.github/workflows/ci-aggregate-status.yml

@@ -26,29 +26,54 @@ jobs:
               per_page: 100,
             });
 
-            const bkStatuses = data.statuses.filter(
-              s => s.context.startsWith('buildkite/ci/')
-            );
-
-            const FASTCHECK_PREFIX = 'buildkite/ci/microscope-';
+            // Buildkite derives the GitHub context prefix from the label emoji.
+            // Keep hard Full Suite lanes in test-tube/bar-chart namespaces and
+            // Fastcheck lanes in microscope so targeted reruns cannot clear the
+            // wrong aggregate status. Automatic PR jobs use pr-fastcheck while
+            // slash-command and Full Suite jobs use ci; normalize the suffix
+            // and keep the newest status for each logical lane.
+            const FASTCHECK_PREFIXES = [
+              'buildkite/pr-fastcheck/microscope-',
+              'buildkite/ci/microscope-',
+            ];
             const FULL_SUITE_PREFIXES = [
               'buildkite/ci/test-tube-',
               'buildkite/ci/bar-chart-',
             ];
 
-            const fastcheck = bkStatuses.filter(
-              s => s.context.startsWith(FASTCHECK_PREFIX)
-            );
-            const fullSuite = bkStatuses.filter(
-              s => FULL_SUITE_PREFIXES.some(p => s.context.startsWith(p))
+            function newestByLane(prefixes) {
+              const statuses = new Map();
+              for (const status of data.statuses) {
+                const prefix = prefixes.find(p => status.context.startsWith(p));
+                if (!prefix) continue;
+                const lane = status.context.slice(prefix.length);
+                const previous = statuses.get(lane);
+                if (!previous || Date.parse(status.updated_at) > Date.parse(previous.updated_at)) {
+                  statuses.set(lane, status);
+                }
+              }
+              return statuses;
+            }
+
+            const fastcheck = newestByLane(FASTCHECK_PREFIXES);
+            const fullSuiteOnly = newestByLane(FULL_SUITE_PREFIXES);
+            const fastcheckPassed =
+              fastcheck.size === 6
+              && [...fastcheck.values()].every(s => s.state === 'success');
+            const fullSuitePassed =
+              fastcheckPassed
+              && fullSuiteOnly.size === 14
+              && [...fullSuiteOnly.values()].every(s => s.state === 'success');
+
+            // Direct reruns may repair a failed suite, never create a gate for
+            // a suite that did not run.
+            const failedAggregate = context => data.statuses.some(
+              s => s.context === context && s.state === 'failure'
             );
 
-            if (
-              fastcheck.length > 0
-              && fastcheck.every(s => s.state === 'success')
-            ) {
+            if (failedAggregate('fastcheck-passed') && fastcheckPassed) {
               core.info(
-                `All ${fastcheck.length} fastcheck tests passed — updating fastcheck-passed`
+                `All ${fastcheck.size} fastcheck tests passed — updating fastcheck-passed`
               );
               await github.rest.repos.createCommitStatus({
                 owner: context.repo.owner,
@@ -56,17 +81,13 @@ jobs:
                 sha,
                 state: 'success',
                 context: 'fastcheck-passed',
-                description:
-                  `All ${fastcheck.length} fastcheck tests passed`,
+                description: `All ${fastcheck.size} fastcheck tests passed`,
               });
             }
 
-            if (
-              fullSuite.length > 0
-              && fullSuite.every(s => s.state === 'success')
-            ) {
+            if (failedAggregate('full-suite-passed') && fullSuitePassed) {
               core.info(
-                `All ${fullSuite.length} full suite tests passed — updating full-suite-passed`
+                'All 20 full suite tests passed — updating full-suite-passed'
               );
               await github.rest.repos.createCommitStatus({
                 owner: context.repo.owner,
@@ -74,7 +95,6 @@ jobs:
                 sha,
                 state: 'success',
                 context: 'full-suite-passed',
-                description:
-                  `All ${fullSuite.length} full suite tests passed`,
+                description: 'All 20 full suite tests passed',
               });
             }

.github/workflows/ci-macos-mlx.yml

@@ -0,0 +1,170 @@
+name: macOS MLX Smoke
+
+on:
+  pull_request:
+    branches: [main]
+    paths:
+      - ".github/workflows/ci-macos-mlx.yml"
+      - "fastvideo/mlx_runtime/**"
+      - "fastvideo/tests/mlx/**"
+      - "fastvideo/tests/platforms/test_mps_vsa_error.py"
+      - "fastvideo/tests/platforms/test_cpu_sdpa.py"
+      - "fastvideo/platforms/cpu.py"
+      - "fastvideo/platforms/mps.py"
+      - "fastvideo/platforms/__init__.py"
+      - "fastvideo/__init__.py"
+      - "examples/inference/basic/mlx_*.py"
+      - "fastvideo/benchmarks/mlx_*.py"
+      - "pyproject.toml"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: macos-mlx-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  mlx-smoke:
+    if: github.event_name == 'workflow_dispatch' || github.event.pull_request.draft != true
+    runs-on: macos-15
+    timeout-minutes: 25
+    env:
+      FASTVIDEO_ATTENTION_BACKEND: TORCH_SDPA
+      TOKENIZERS_PARALLELISM: "false"
+      MASTER_ADDR: "127.0.0.1"
+      MASTER_PORT: "29513"
+      GLOO_SOCKET_IFNAME: lo0
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - uses: astral-sh/setup-uv@v3
+
+      - name: Install lightweight MLX smoke dependencies
+        run: |
+          uv pip install --system \
+            --index-url https://download.pytorch.org/whl/cpu \
+            torch==2.12.0 torchvision torchaudio
+          uv pip install --system \
+            pytest pytest-timeout numpy scipy pillow imageio einops cloudpickle filelock \
+            PyYAML diffusers huggingface_hub remote-pdb safetensors loguru mlx \
+            "ftfy>=6.3.1" "opencv-python>=4.10.0.84" psutil "transformers>=5.0.0"
+
+      - name: Show Apple runtime
+        run: |
+          python - <<'PY'
+          import platform
+          import mlx.core as mx
+          import torch
+
+          print("machine:", platform.machine())
+          print("processor:", platform.processor())
+          print("mlx default device:", mx.default_device())
+          device_info = mx.metal.device_info() if mx.metal.is_available() else "metal unavailable"
+          print("mlx device_info:", device_info)
+          print("torch:", torch.__version__)
+          print("torch mps available:", torch.backends.mps.is_available())
+          PY
+
+      - name: Run MLX smoke tests
+        run: |
+          python -m pytest \
+            fastvideo/mlx_runtime/tests/ \
+            fastvideo/tests/mlx/test_dmd_sampling.py \
+            fastvideo/tests/mlx/test_memory_limits.py \
+            fastvideo/tests/mlx/test_quant_capability.py \
+            fastvideo/tests/mlx/test_mlx_dit_parity.py \
+            fastvideo/tests/mlx/test_mlx_compile_parity.py \
+            fastvideo/tests/mlx/test_mlx_checkpoint.py \
+            fastvideo/tests/mlx/test_mlx_checkpoint_compat.py \
+            fastvideo/tests/mlx/test_mlx_affine_dq_gemm.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_parity.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_vsa.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_vsa_regressions.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_fast_mode.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_fast_spatial.py \
+            fastvideo/tests/mlx/test_mlx_fastwan_benchmark.py \
+            fastvideo/tests/mlx/test_taehv_decode.py \
+            fastvideo/tests/mlx/test_frame_upsample.py \
+            fastvideo/tests/mlx/test_mlx_fast_spatial.py \
+            fastvideo/tests/mlx/test_mlx_refine.py \
+            fastvideo/tests/mlx/test_mlx_prompt_enhance.py \
+            fastvideo/tests/mlx/test_mlx_prompt_to_video_decode.py \
+            fastvideo/tests/mlx/test_mlx_wan22_prompt_cache_fingerprint.py \
+            fastvideo/tests/mlx/test_wan22_sample.py \
+            fastvideo/tests/mlx/test_windowed_attention.py \
+            fastvideo/tests/mlx/test_mlx_rife_interpolation.py::test_rife_download_unavailable_has_specific_error \
+            fastvideo/tests/mlx/test_mlx_rife_interpolation.py::test_rife_backend_regression_is_not_skip_eligible \
+            fastvideo/tests/platforms/test_mps_vsa_error.py \
+            fastvideo/tests/platforms/test_cpu_sdpa.py \
+            -v -s --timeout=120 -o faulthandler_timeout=120
+
+  # Same tests on MLX's CPU backend. Hosted macOS runners are scarce and
+  # slower to schedule; this Linux job gives fast PR signal on the identical
+  # graph (the parity tests were designed to be backend-agnostic), while the
+  # macOS job above stays the source of truth for Metal behavior.
+  mlx-smoke-linux-cpu:
+    if: github.event_name == 'workflow_dispatch' || github.event.pull_request.draft != true
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    env:
+      FASTVIDEO_ATTENTION_BACKEND: TORCH_SDPA
+      TOKENIZERS_PARALLELISM: "false"
+      MASTER_ADDR: localhost
+      MASTER_PORT: "29513"
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - uses: astral-sh/setup-uv@v3
+
+      - name: Install lightweight MLX smoke dependencies (CPU backend)
+        run: |
+          uv pip install --system \
+            --index-url https://download.pytorch.org/whl/cpu \
+            torch==2.12.0 torchvision torchaudio
+          uv pip install --system \
+            pytest pytest-timeout numpy scipy pillow imageio einops cloudpickle filelock \
+            PyYAML diffusers huggingface_hub remote-pdb safetensors loguru "mlx[cpu]" \
+            "ftfy>=6.3.1" "opencv-python>=4.10.0.84" psutil "transformers>=5.0.0"
+
+      - name: Run MLX smoke tests (CPU backend)
+        run: |
+          python -m pytest \
+            fastvideo/mlx_runtime/tests/ \
+            fastvideo/tests/mlx/test_dmd_sampling.py \
+            fastvideo/tests/mlx/test_memory_limits.py \
+            fastvideo/tests/mlx/test_quant_capability.py \
+            fastvideo/tests/mlx/test_mlx_dit_parity.py \
+            fastvideo/tests/mlx/test_mlx_compile_parity.py \
+            fastvideo/tests/mlx/test_mlx_checkpoint.py \
+            fastvideo/tests/mlx/test_mlx_checkpoint_compat.py \
+            fastvideo/tests/mlx/test_mlx_affine_dq_gemm.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_parity.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_vsa.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_vsa_regressions.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_fast_mode.py \
+            fastvideo/tests/mlx/test_mlx_minimax_h3_fast_spatial.py \
+            fastvideo/tests/mlx/test_mlx_fastwan_benchmark.py \
+            fastvideo/tests/mlx/test_taehv_decode.py \
+            fastvideo/tests/mlx/test_frame_upsample.py \
+            fastvideo/tests/mlx/test_mlx_fast_spatial.py \
+            fastvideo/tests/mlx/test_mlx_refine.py \
+            fastvideo/tests/mlx/test_mlx_prompt_enhance.py \
+            fastvideo/tests/mlx/test_mlx_prompt_to_video_decode.py \
+            fastvideo/tests/mlx/test_mlx_wan22_prompt_cache_fingerprint.py \
+            fastvideo/tests/mlx/test_wan22_sample.py \
+            fastvideo/tests/mlx/test_windowed_attention.py \
+            fastvideo/tests/mlx/test_mlx_rife_interpolation.py::test_rife_download_unavailable_has_specific_error \
+            fastvideo/tests/mlx/test_mlx_rife_interpolation.py::test_rife_backend_regression_is_not_skip_eligible \
+            fastvideo/tests/platforms/test_mps_vsa_error.py \
+            fastvideo/tests/platforms/test_cpu_sdpa.py \
+            -v -s --timeout=120 -o faulthandler_timeout=120

.github/workflows/ci-precommit.yml

@@ -27,14 +27,25 @@ jobs:
         ref: ${{ inputs.ref || '' }}
     # For PR events, lint the PR head — but keep the hook definitions from
     # the base branch so an untrusted PR cannot alter what gets executed.
-    - name: Save trusted hook config
+    # The gate scripts are saved too: the self-test step below executes them,
+    # so it must run the base-branch copies, not the PR head's.
+    - name: Save trusted hook config and gate scripts
       if: github.event_name == 'pull_request_target'
-      run: cp .pre-commit-config.yaml "$RUNNER_TEMP/trusted-pre-commit-config.yaml"
-    - uses: actions/checkout@v4
+      run: |
+        cp .pre-commit-config.yaml "$RUNNER_TEMP/trusted-pre-commit-config.yaml"
+        cp -a .github/scripts "$RUNNER_TEMP/trusted-scripts"
+        echo "GATE_SCRIPTS_DIR=$RUNNER_TEMP/trusted-scripts" >> "$GITHUB_ENV"
+    # allow-unsafe-pr-checkout acknowledges checkout's pull_request_target
+    # guard: the head is data for the trusted hooks to lint; nothing from it
+    # is executed (config and gate scripts are pinned to the base branch
+    # above) and credentials are not persisted. SHA-pinned to v4.4.0 because
+    # actionlint's action schema does not know the new input yet.
+    - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4.4.0
       if: github.event_name == 'pull_request_target'
       with:
         ref: ${{ github.event.pull_request.head.sha }}
         persist-credentials: false
+        allow-unsafe-pr-checkout: true
     - name: Restore trusted hook config
       if: github.event_name == 'pull_request_target'
       run: cp "$RUNNER_TEMP/trusted-pre-commit-config.yaml" .pre-commit-config.yaml
@@ -47,3 +58,8 @@ jobs:
     - uses: pre-commit/action@v3.0.1
       with:
         extra_args: --all-files --hook-stage manual
+    # After pre-commit so a self-test failure cannot mask lint failures.
+    # GATE_SCRIPTS_DIR points at the base-branch copy on fork PRs (set above);
+    # push / workflow_call runs use the checked-out tree directly.
+    - name: Full-suite gate self-test
+      run: bash "${GATE_SCRIPTS_DIR:-.github/scripts}/test_gate_full_suite.sh"

.github/workflows/ci-scheduled-ssim.yml

@@ -0,0 +1,44 @@
+name: Scheduled Full SSIM
+
+on:
+  schedule:
+    - cron: "0 5 * * 0"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  trigger:
+    if: github.repository == 'hao-ai-lab/FastVideo'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Trigger weekly full SSIM on Slinky Slurm
+        env:
+          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
+          SOURCE_SHA: ${{ github.sha }}
+          SOURCE_BRANCH: ${{ github.event.repository.default_branch }}
+          BK_ORG: ${{ vars.BUILDKITE_ORG_SLUG }}
+          BK_PIPELINE: ${{ vars.BUILDKITE_PIPELINE_SLUG }}
+        run: |
+          set -euo pipefail
+          curl -sS --fail-with-body -X POST \
+            "https://api.buildkite.com/v2/organizations/${BK_ORG}/pipelines/${BK_PIPELINE}/builds" \
+            -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
+            -H "Content-Type: application/json" \
+            --data-raw "$(jq -n \
+              --arg commit "$SOURCE_SHA" \
+              --arg branch "$SOURCE_BRANCH" \
+              '{
+                commit: $commit,
+                branch: $branch,
+                message: "Weekly full SSIM on Slinky Slurm",
+                ignore_pipeline_branch_filters: true,
+                env: {
+                  TEST_SCOPE: "scheduled",
+                  FULL_SUITE: "false",
+                  TEST_TYPE: "ssim",
+                  PR_NUMBER: "false",
+                  PR_TITLE: "Scheduled full SSIM"
+                }
+              }')"

.github/workflows/ci-slash-commands.yml

@@ -32,8 +32,7 @@ jobs:
             }
             core.setOutput('has_write', String(hasWrite));
 
-      - name: Add ready label and react
-        id: label
+      - name: Add ready label
         if: steps.perm.outputs.has_write == 'true'
         uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
         with:
@@ -41,54 +40,33 @@ jobs:
             const owner = context.repo.owner;
             const repo = context.repo.repo;
             const prNumber = context.payload.issue.number;
-            try { await github.rest.issues.removeLabel({ owner, repo, issue_number: prNumber, name: 'ready' }); } catch {}
             await github.rest.issues.addLabels({ owner, repo, issue_number: prNumber, labels: ['ready'] });
+
+      - name: React to comment
+        if: steps.perm.outputs.has_write == 'true'
+        continue-on-error: true
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
+        with:
+          script: |
             await github.rest.reactions.createForIssueComment({
-              owner, repo,
+              owner: context.repo.owner,
+              repo: context.repo.repo,
               comment_id: context.payload.comment.id,
               content: 'rocket',
             });
-            const { data: pr } = await github.rest.pulls.get({ owner, repo, pull_number: prNumber });
-            core.setOutput('pr_sha', pr.head.sha);
-            core.setOutput('pr_branch', pr.head.ref);
-            core.setOutput('pr_number', String(prNumber));
-            core.setOutput('pr_title', pr.title);
 
-      - name: Trigger Full Suite
-        if: steps.perm.outputs.has_write == 'true'
-        env:
-          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
-          PR_SHA: ${{ steps.label.outputs.pr_sha }}
-          PR_BRANCH: ${{ steps.label.outputs.pr_branch }}
-          PR_NUMBER: ${{ steps.label.outputs.pr_number }}
-          PR_TITLE: ${{ steps.label.outputs.pr_title }}
-          BK_ORG: ${{ vars.BUILDKITE_ORG_SLUG }}
-          BK_PIPELINE: ${{ vars.BUILDKITE_PIPELINE_SLUG }}
-        run: |
-          curl -sS --fail-with-body -X POST \
-            "https://api.buildkite.com/v2/organizations/${BK_ORG}/pipelines/${BK_PIPELINE}/builds" \
-            -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
-            -H "Content-Type: application/json" \
-            --data-raw "$(jq -n \
-              --arg commit "$PR_SHA" \
-              --arg branch "$PR_BRANCH" \
-              --arg message "Full Suite for PR #${PR_NUMBER} (via /merge)" \
-              --arg pr_title "$PR_TITLE" \
-              --argjson pr_id "$PR_NUMBER" \
-              '{
-                commit: $commit,
-                branch: $branch,
-                message: $message,
-                ignore_pipeline_branch_filters: true,
-                pull_request_id: $pr_id,
-                pull_request_base_branch: "main",
-                env: {
-                    TEST_SCOPE: "full",
-                    FULL_SUITE: "true",
-                    PR_NUMBER: ($pr_id | tostring),
-                    PR_TITLE: $pr_title
-                  }
-                }')"
+  trigger-merge-gate:
+    needs: handle-merge
+    if: needs.handle-merge.result == 'success'
+    permissions:
+      actions: read
+      contents: read
+      pull-requests: read
+    uses: ./.github/workflows/ci-trigger-full-suite.yml
+    with:
+      pr_number: ${{ github.event.issue.number }}
+    secrets:
+      BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
 
   parse-command:
     if: >-
@@ -129,7 +107,7 @@ jobs:
           set -euo pipefail
           TEST_NAME=$(echo "$COMMENT" | grep -oP '(?<=/test\s)\S+' | head -1 || true)
 
-          VALID="encoder vae transformer kernel unit dreamverse ssim training lora-inference lora-training lora-extraction distillation self-forcing vsa vmoba performance api train-framework eval full fastcheck pre-commit"
+          VALID="encoder vae transformer kernel unit dreamverse ssim golden-gate training lora-inference lora-training lora-extraction distillation self-forcing vsa vmoba performance api train-framework eval unit-ci kernel-ci dreamverse-ci ssim-ci golden-gate-ci encoder-ci vae-ci transformer-ci lora-inference-ci lora-training-ci lora-extraction-ci training-ci distillation-ci self-forcing-ci vsa-ci vmoba-ci performance-ci api-ci train-framework-ci eval-ci full fastcheck pre-commit"
           if [ -z "$TEST_NAME" ] || ! echo "$VALID" | grep -qw "$TEST_NAME"; then
             echo "Unknown test: '$TEST_NAME'. Valid: $VALID"
             exit 1
@@ -137,8 +115,18 @@ jobs:
 
           declare -A MAP=(
             [encoder]=encoder [vae]=vae [transformer]=transformer
-            [kernel]=kernel_tests [unit]=unit_test [dreamverse]=dreamverse_app
-            [ssim]=ssim [training]=training
+            [kernel]=kernel_tests [unit]=unit_test [unit-ci]=unit_test_ci
+            [kernel-ci]=kernel_tests_ci [dreamverse-ci]=dreamverse_app_ci
+            [ssim-ci]=ssim_ci [vmoba-ci]=inference_vmoba_ci
+            [golden-gate-ci]=golden_gate_ci [training-ci]=training_ci
+            [encoder-ci]=encoder_ci [vae-ci]=vae_ci [transformer-ci]=transformer_ci
+            [lora-inference-ci]=inference_lora_ci [lora-training-ci]=training_lora_ci
+            [lora-extraction-ci]=lora_extraction_ci [distillation-ci]=distillation_dmd_ci
+            [self-forcing-ci]=self_forcing_ci [vsa-ci]=training_vsa_ci
+            [performance-ci]=performance_ci [api-ci]=api_server_ci
+            [train-framework-ci]=train_framework_ci [eval-ci]=eval_ci
+            [dreamverse]=dreamverse_app
+            [ssim]=ssim [golden-gate]=golden_gate [training]=training
             [lora-inference]=inference_lora [lora-training]=training_lora
             [lora-extraction]=lora_extraction
             [distillation]=distillation_dmd [self-forcing]=self_forcing

.github/workflows/ci-trigger-full-suite.yml

@@ -1,64 +1,232 @@
-name: Trigger Full Suite
+name: Trigger Merge Gate
 
 on:
   pull_request_target:
     types: [labeled, synchronize]
+  workflow_call:
+    inputs:
+      pr_number:
+        description: Pull request number to enter into the merge gate
+        required: true
+        type: number
+    secrets:
+      BUILDKITE_API_TOKEN:
+        required: true
 
 permissions:
   contents: read
   pull-requests: read
-
-concurrency:
-  group: full-suite-${{ github.event.pull_request.number }}
-  cancel-in-progress: false
+  actions: read
 
 jobs:
   trigger:
     if: >-
-      (github.event.action == 'labeled' && github.event.label.name == 'ready')
+      inputs.pr_number > 0
+      || (github.event.action == 'labeled' && github.event.label.name == 'ready')
       || github.event.action == 'synchronize'
     runs-on: ubuntu-latest
+    # Job-level concurrency: only this guarded job acquires the group, so an
+    # unrelated `labeled` event (which skips the job) cannot cancel an in-flight
+    # gate and then skip its replacement. The newest real trigger (`ready`,
+    # push, or `/merge`) supersedes the in-flight run, whose Buildkite build the
+    # cancel step below replaces.
+    concurrency:
+      group: merge-gate-${{ inputs.pr_number || github.event.pull_request.number }}
+      cancel-in-progress: true
+    # Gate below may wait for cheap checks (up to MAX_WAIT_SECS = 25 min).
+    timeout-minutes: 35
     steps:
       - name: Check ready label
         id: check
         uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
+        env:
+          CALLED_PR_NUMBER: ${{ inputs.pr_number }}
         with:
           script: |
+            const eventPrNumber = context.payload.pull_request?.number;
+            const calledPrNumber = Number(process.env.CALLED_PR_NUMBER);
+            const prNumber = eventPrNumber ?? calledPrNumber;
+            if (!Number.isSafeInteger(prNumber) || prNumber <= 0) {
+              core.setFailed(`Invalid pull request number: ${process.env.CALLED_PR_NUMBER}`);
+              return;
+            }
             const { data: pr } = await github.rest.pulls.get({
               owner: context.repo.owner,
               repo: context.repo.repo,
-              pull_number: context.payload.pull_request.number,
+              pull_number: prNumber,
             });
+            if (pr.state !== 'open') {
+              core.setFailed(`PR #${prNumber} is not open.`);
+              return;
+            }
+            if (pr.base.repo.full_name !== context.payload.repository.full_name
+                || pr.base.ref !== context.payload.repository.default_branch) {
+              core.setFailed(`PR #${prNumber} does not target this repository's default branch.`);
+              return;
+            }
             const hasReady = pr.labels.some(l => l.name === 'ready');
             core.setOutput('has_ready', String(hasReady));
-            if (!hasReady) core.info('No ready label — skipping Full Suite trigger.');
+            core.setOutput('changed_files', String(pr.changed_files));
+            core.setOutput('pr_number', String(pr.number));
+            core.setOutput('head_sha', pr.head.sha);
+            core.setOutput('head_ref', pr.head.ref);
+            core.setOutput('base_sha', pr.base.sha);
+            core.setOutput('title', pr.title);
+            if (!hasReady) core.info('No ready label — skipping merge-gate trigger.');
 
       - name: Cancel previous Buildkite builds
+        # Cancelling stale builds only saves agent time. If it cannot run, the
+        # merge gate must still be triggered by the steps below, so a failure
+        # here is reported and stepped over rather than ending the job.
+        continue-on-error: true
+        timeout-minutes: 3
         if: steps.check.outputs.has_ready == 'true'
         env:
-          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
-          PR_BRANCH: ${{ github.event.pull_request.head.ref }}
-        run: |
-          # Find running builds for this branch with TEST_SCOPE=full and cancel them
-          builds=$(curl -sS -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
-            "https://api.buildkite.com/v2/organizations/${{ vars.BUILDKITE_ORG_SLUG }}/pipelines/${{ vars.BUILDKITE_PIPELINE_SLUG }}/builds?branch=${PR_BRANCH}&state=running,scheduled" \
-            | jq -r '.[] | select(try (.env.TEST_SCOPE == "full") catch false) | .number')
-          for build_num in $builds; do
-            echo "Cancelling Buildkite build #$build_num"
-            curl -sS -X PUT -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
-              "https://api.buildkite.com/v2/organizations/${{ vars.BUILDKITE_ORG_SLUG }}/pipelines/${{ vars.BUILDKITE_PIPELINE_SLUG }}/builds/${build_num}/cancel"
-          done
-
-      - name: Trigger Buildkite Full Suite
-        if: steps.check.outputs.has_ready == 'true'
-        env:
-          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
-          PR_SHA: ${{ github.event.pull_request.head.sha }}
-          PR_BRANCH: ${{ github.event.pull_request.head.ref }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-          PR_TITLE: ${{ github.event.pull_request.title }}
           BK_ORG: ${{ vars.BUILDKITE_ORG_SLUG }}
           BK_PIPELINE: ${{ vars.BUILDKITE_PIPELINE_SLUG }}
+          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
+          PR_BRANCH: ${{ steps.check.outputs.head_ref }}
+          PR_NUMBER: ${{ steps.check.outputs.pr_number }}
+        run: |
+          set -euo pipefail
+          response_file=$(mktemp)
+          builds_file=$(mktemp)
+          trap 'rm -f "$response_file" "$builds_file"' EXIT
+
+          if [[ ! "$PR_NUMBER" =~ ^[1-9][0-9]*$ ]]; then
+            echo "::warning::Invalid pull request number; stale Buildkite builds may continue."
+            exit 1
+          fi
+
+          if ! curl -sS --fail-with-body --connect-timeout 5 --max-time 20 --get \
+            -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
+            --data-urlencode "branch=$PR_BRANCH" \
+            --data-urlencode "state[]=running" \
+            --data-urlencode "state[]=scheduled" \
+            --data-urlencode "state[]=failing" \
+            --data-urlencode "exclude_jobs=true" \
+            --data-urlencode "exclude_pipeline=true" \
+            --output "$response_file" \
+            "https://api.buildkite.com/v2/organizations/${BK_ORG}/pipelines/${BK_PIPELINE}/builds"; then
+            echo "::warning::Could not list Buildkite builds; stale merge-gate builds may continue."
+            exit 1
+          fi
+
+          if ! jq -e '
+            if type != "array" then false
+            else all(.[];
+              if type != "object" then false
+              else
+                (.number | if type == "number" then . > 0 and floor == . else false end)
+                and (
+                  (.env? | if . == null then {} else . end) as $env
+                  | if ($env | type) != "object" then false
+                    else
+                      ($env.TEST_SCOPE? | . == null or type == "string")
+                      and ($env.PR_NUMBER? | . == null or type == "string")
+                    end
+                )
+              end
+            )
+            end
+          ' "$response_file" >/dev/null 2>&1; then
+            # Do not print the response body: it is remote data and may contain
+            # multiline values that would be interpreted as workflow commands.
+            echo "::warning::Buildkite returned an invalid build list; stale merge-gate builds may continue."
+            exit 1
+          fi
+
+          # Match both branch and PR number: forks can reuse the same branch name.
+          if ! jq -r --arg pr_number "$PR_NUMBER" '
+            .[]
+            | select((.env.TEST_SCOPE? == "merge") and (.env.PR_NUMBER? == $pr_number))
+            | .number
+          ' "$response_file" > "$builds_file"; then
+            echo "::warning::Could not select stale Buildkite builds; stale merge-gate builds may continue."
+            exit 1
+          fi
+
+          cancellation_failed=0
+          while IFS= read -r build_num; do
+            echo "Cancelling Buildkite build #$build_num"
+            if ! curl -sS --fail-with-body --connect-timeout 5 --max-time 20 -o /dev/null -X PUT \
+              -H "Authorization: Bearer $BUILDKITE_API_TOKEN" \
+              "https://api.buildkite.com/v2/organizations/${BK_ORG}/pipelines/${BK_PIPELINE}/builds/${build_num}/cancel"; then
+              echo "::warning::Could not cancel Buildkite build #$build_num; trying remaining builds."
+              cancellation_failed=1
+            fi
+          done < "$builds_file"
+
+          if (( cancellation_failed != 0 )); then
+            exit 1
+          fi
+
+      # Check out the immutable BASE SHA: neither pull_request_target nor the
+      # privileged slash-command call may run code from the untrusted PR head.
+      - name: Checkout trusted merge planner
+        if: steps.check.outputs.has_ready == 'true'
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          ref: ${{ steps.check.outputs.base_sha }}
+          persist-credentials: false
+
+      - name: Collect changed paths
+        if: steps.check.outputs.has_ready == 'true'
+        env:
+          GH_TOKEN: ${{ github.token }}
+          PR_NUMBER: ${{ steps.check.outputs.pr_number }}
+          EXPECTED_CHANGED_FILES: ${{ steps.check.outputs.changed_files }}
+        run: |
+          set -euo pipefail
+          changed_json="$RUNNER_TEMP/merge-changed-files.json"
+          changed_paths="$RUNNER_TEMP/merge-changed-paths.txt"
+          if gh api --paginate --slurp \
+              "repos/${GITHUB_REPOSITORY}/pulls/${PR_NUMBER}/files?per_page=100" \
+              > "$changed_json"; then
+            observed=$(jq '[.[][] | .filename] | unique | length' "$changed_json")
+            if [ "$observed" = "$EXPECTED_CHANGED_FILES" ]; then
+              jq -r '.[][] | .filename, (.previous_filename // empty)' "$changed_json" \
+                | sort -u > "$changed_paths"
+            else
+              echo "::warning::Changed-file API returned $observed of $EXPECTED_CHANGED_FILES paths; selecting all merge lanes."
+              echo '__FASTVIDEO_CI_PLAN_ALL__' > "$changed_paths"
+            fi
+          else
+            echo "::warning::Changed-file API failed; selecting all merge lanes."
+            echo '__FASTVIDEO_CI_PLAN_ALL__' > "$changed_paths"
+          fi
+
+      - name: Select minimal merge tests
+        id: plan
+        if: steps.check.outputs.has_ready == 'true'
+        run: |
+          python3 .github/scripts/plan_merge_ci.py \
+            --paths-file "$RUNNER_TEMP/merge-changed-paths.txt" \
+            --github-output "$GITHUB_OUTPUT" \
+            --summary-file "$GITHUB_STEP_SUMMARY"
+
+      - name: Wait for pre-commit and docs build
+        if: steps.check.outputs.has_ready == 'true'
+        env:
+          GH_TOKEN: ${{ github.token }}
+          PR_SHA: ${{ steps.check.outputs.head_sha }}
+          PR_NUMBER: ${{ steps.check.outputs.pr_number }}
+        run: bash .github/scripts/gate_full_suite.sh
+
+      - name: Trigger Buildkite merge gate
+        if: steps.check.outputs.has_ready == 'true'
+        env:
+          BUILDKITE_API_TOKEN: ${{ secrets.BUILDKITE_API_TOKEN }}
+          PR_SHA: ${{ steps.check.outputs.head_sha }}
+          PR_BRANCH: ${{ steps.check.outputs.head_ref }}
+          PR_NUMBER: ${{ steps.check.outputs.pr_number }}
+          PR_TITLE: ${{ steps.check.outputs.title }}
+          BK_ORG: ${{ vars.BUILDKITE_ORG_SLUG }}
+          BK_PIPELINE: ${{ vars.BUILDKITE_PIPELINE_SLUG }}
+          MERGE_TEST_PLAN: ${{ steps.plan.outputs.merge_test_plan }}
+          MERGE_GOLDEN_TESTS: ${{ steps.plan.outputs.merge_golden_tests }}
+          MERGE_SSIM_TESTS: ${{ steps.plan.outputs.merge_ssim_tests }}
+          MERGE_PLAN_LABEL: ${{ steps.plan.outputs.merge_plan_label }}
         run: |
           curl -sS --fail-with-body -X POST \
             "https://api.buildkite.com/v2/organizations/${BK_ORG}/pipelines/${BK_PIPELINE}/builds" \
@@ -67,8 +235,11 @@ jobs:
             --data-raw "$(jq -n \
               --arg commit "$PR_SHA" \
               --arg branch "$PR_BRANCH" \
-              --arg message "Full Suite for PR #${PR_NUMBER}" \
+              --arg message "Merge gate [${MERGE_PLAN_LABEL}] for PR #${PR_NUMBER}" \
               --arg pr_title "$PR_TITLE" \
+              --arg merge_test_plan "$MERGE_TEST_PLAN" \
+              --arg merge_golden_tests "$MERGE_GOLDEN_TESTS" \
+              --arg merge_ssim_tests "$MERGE_SSIM_TESTS" \
               --argjson pr_id "$PR_NUMBER" \
               '{
                 commit: $commit,
@@ -78,8 +249,11 @@ jobs:
                 pull_request_id: $pr_id,
                 pull_request_base_branch: "main",
                 env: {
-                  TEST_SCOPE: "full",
+                  TEST_SCOPE: "merge",
                   FULL_SUITE: "true",
+                  MERGE_TEST_PLAN: $merge_test_plan,
+                  MERGE_GOLDEN_TESTS: $merge_golden_tests,
+                  MERGE_SSIM_TESTS: $merge_ssim_tests,
                   PR_NUMBER: ($pr_id | tostring),
                   PR_TITLE: $pr_title
                 }

.github/workflows/community-welcome.yml

@@ -38,17 +38,17 @@ jobs:
 
             **How our CI works:**
 
-            PRs run a two-tier CI system:
+            PRs run a three-tier CI system:
             1. **Pre-commit** — formatting (yapf), linting (ruff), type checking (mypy). Runs immediately on every PR.
-            2. **Fastcheck** — core GPU tests (encoders, VAEs, transformers, kernels, unit tests). Runs automatically via Buildkite on relevant file changes (~10-15 min).
-            3. **Full Suite** — integration tests, training pipelines, SSIM regression. Runs only when a reviewer adds the `ready` label.
+            2. **Fastcheck** — six core GPU lanes run automatically via Buildkite (~10-15 min).
+            3. **Merge gate** — a reviewer adds `ready`; changed paths select only the relevant integration, training, golden, or SSIM coverage.
 
             **Before your PR is reviewed:**
             - [ ] `pre-commit run --all-files` passes locally
             - [ ] You've added or updated tests for your changes
             - [ ] The PR description explains what and why
 
-            If pre-commit fails, a bot comment will explain how to fix it. Fastcheck and Full Suite results appear in the Checks section below.
+            If pre-commit fails, a bot comment will explain how to fix it. Fastcheck and merge-gate results appear in the Checks section below.
 
             **Useful links:**
             - [Contributing Guide](https://hao-ai-lab.github.io/FastVideo/contributing/overview/)

.github/workflows/infra-build-image.yml

@@ -13,16 +13,28 @@ on:
         required: false
         default: false
         type: boolean
-  # Auto-rebuild the CUDA images when their Dockerfile changes on main. The CUDA
-  # matrix is the only lane that builds from docker/Dockerfile, so a path-scoped
-  # push trigger is a sufficient change detector on its own -- no separate
-  # detect-changes/paths-filter job is needed now that there is a single
-  # in-scope Dockerfile. Dreamverse (apps/dreamverse/docker/Dockerfile) and the
-  # rocm Dockerfile stay manual-dispatch only.
+      build_ci_runner_image:
+        description: 'Build the ARM64 CUDA 13 CI runner image (sm_100)'
+        required: false
+        default: false
+        type: boolean
+  # Auto-rebuild the CUDA images when a repository-controlled image input
+  # changes on main. This includes the trusted SM89 kernel artifact's source,
+  # metadata/key helper, ABI dependency metadata, and build orchestration.
+  # Dreamverse (apps/dreamverse/docker/Dockerfile) and the ROCm Dockerfile stay
+  # manual-dispatch only.
   push:
     branches: [main]
     paths:
+      - '.dockerignore'
+      - '.github/workflows/_template-build-image.yml'
+      - '.github/workflows/infra-build-image.yml'
+      - '.gitmodules'
       - 'docker/Dockerfile'
+      - 'docker/uv-excludes'
+      - 'fastvideo-kernel/**'
+      - 'fastvideo/tests/modal/kernel_build_cache.py'
+      - 'pyproject.toml'
 
 
 permissions:
@@ -50,7 +62,7 @@ jobs:
   # 2.8.3 comes from the architecture-specific prebuilt releases.
   build-cuda-images:
     # Runs on a manual dispatch when build_cuda_matrix is set, or automatically
-    # on a push that changed docker/Dockerfile (inputs are null on push). The
+    # on an in-scope main push (inputs are null on push). The
     # repository guard keeps fork syncs from auto-building; manual dispatch
     # still works in forks.
     if: ${{ (github.event_name == 'push' && github.repository == 'hao-ai-lab/FastVideo') || github.event.inputs.build_cuda_matrix == 'true' }}
@@ -191,6 +203,28 @@ jobs:
           docker buildx imagetools create "${TAG_ARGS[@]}" "${IMAGE_REFS[@]}"
           docker buildx imagetools inspect "${TAGS[0]}"
 
+  # The CI runner is ARM64 like DGX Spark, but targets sm_100 rather than sm_121.
+  # Publish a single-architecture variant so the self-hosted CI runner can reuse
+  # the exact prebuilt kernel instead of compiling it in every job.
+  build-ci-runner-image:
+    if: ${{ (github.event_name == 'push' && github.repository == 'hao-ai-lab/FastVideo') || github.event.inputs.build_ci_runner_image == 'true' }}
+    uses: ./.github/workflows/_template-build-image.yml
+    with:
+      python_version: '3.12'
+      dockerfile_path: docker/Dockerfile
+      tag_suffix: py3.12-cuda13.0.0-sm100
+      runner: ubuntu-24.04-arm
+      architecture: arm64
+      build_args: |
+        PYTHON_VERSION=3.12
+        CUDA_VERSION=13.0.0
+        UV_TORCH_BACKEND=cu130
+        TORCH_CUDA_ARCH_LIST=10.0
+        CMAKE_BUILD_PARALLEL_LEVEL=1
+        FLASH_ATTN_WHEEL_TAG=cu130torch2.12
+        FLASH_ATTN_WHEEL_RELEASE_ARM64=https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.9.22
+    secrets: inherit
+
   # Dreamverse matrix: {backend, UI} x {12.6.3, 13.0.0}, Python 3.12. Torch backend
   # matches the base CUDA (cu126 / cu130). Keep these images amd64-only until the
   # required FA4 dependency stack is available and validated on arm64.

.github/workflows/infra-docs.yml

@@ -6,6 +6,7 @@ on:
     paths:
       - 'docs/**'
       - 'examples/**'
+      - 'scripts/inference/**'
       - 'mkdocs.yml'
       - 'requirements-mkdocs.in'
       - 'requirements-mkdocs.txt'
@@ -16,6 +17,7 @@ on:
     paths:
       - 'docs/**'
       - 'examples/**'
+      - 'scripts/inference/**'
       - 'mkdocs.yml'
       - 'requirements-mkdocs.in'
       - 'requirements-mkdocs.txt'

.github/workflows/publish-fastvideo-kernel-hub.yml

@@ -1,95 +0,0 @@
-name: Publish FastVideo Kernel to Hugging Face Hub
-
-# Version-gated like the PyPI flow (publish-kernel.yml): a push to main only
-# publishes when the fastvideo-kernel version actually changes. Use
-# workflow_dispatch as the manual override.
-on:
-  push:
-    branches:
-      - main
-    paths:
-      - "fastvideo-kernel/pyproject.toml"
-  workflow_dispatch:
-    inputs:
-      branch:
-        description: "Hub branch to upload the build"
-        default: ""
-        required: false
-
-# A Hub build takes hours; never race two publishes to the same repo.
-concurrency:
-  group: ${{ github.workflow }}
-  cancel-in-progress: false
-
-jobs:
-  check-version-change:
-    runs-on: ubuntu-latest
-    outputs:
-      version-changed: ${{ steps.check-version.outputs.changed }}
-      new-version: ${{ steps.check-version.outputs.new-version }}
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 2
-
-      - name: Check if version changed
-        id: check-version
-        run: |
-          cd fastvideo-kernel
-          # Get current commit's version from pyproject.toml
-          # Use ^ to match start of line to avoid matching minimum-version
-          NEW_VERSION=$(grep -oP '^version\s*=\s*"\K[^"]+' pyproject.toml)
-          echo "New version: $NEW_VERSION"
-
-          # Get previous version from git history
-          # Note: git show expects path relative to repo root
-          OLD_VERSION=$(git show HEAD~1:fastvideo-kernel/pyproject.toml | grep -oP '^version\s*=\s*"\K[^"]+' || echo "0.0.0")
-          echo "Old version: $OLD_VERSION"
-
-          if [ "$NEW_VERSION" != "$OLD_VERSION" ]; then
-            echo "Version changed from $OLD_VERSION to $NEW_VERSION"
-            echo "changed=true" >> "$GITHUB_OUTPUT"
-            echo "new-version=$NEW_VERSION" >> "$GITHUB_OUTPUT"
-          else
-            echo "Version did not change"
-            echo "changed=false" >> "$GITHUB_OUTPUT"
-          fi
-
-  build:
-    needs: check-version-change
-    if: ${{ needs.check-version-change.outputs.version-changed == 'true' || github.event_name == 'workflow_dispatch' }}
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Build kernel via HF Jobs and upload to hao-ai-lab/fastvideo-kernel
-        # Pinned to a commit SHA: this third-party action receives HF_TOKEN.
-        uses: huggingface/kernel-builder-job@ee590cee1ff01e922310401f18d5554b8da448ac # main, 2026-06-03
-        with:
-          token: ${{ secrets.HF_TOKEN }}
-          namespace: hao-ai-lab
-          flavor: cpu-xl
-          timeout: "21600"
-          script: |
-            set +x
-            export HF_TOKEN="${{ secrets.HF_TOKEN }}"
-            export GIT_LFS_SKIP_SMUDGE=1
-
-            git clone "${{ github.server_url }}/${{ github.repository }}" FastVideo
-            cd FastVideo
-            git checkout "${{ github.sha }}"
-            git submodule update --init --recursive fastvideo-kernel/include/cutlass fastvideo-kernel/include/tk
-            cd fastvideo-kernel
-
-            UPLOAD_BRANCH="${{ github.event.inputs.branch }}"
-            BRANCH_ARGS=()
-            if [ -n "${UPLOAD_BRANCH}" ]; then
-              BRANCH_ARGS=(--branch "${UPLOAD_BRANCH}")
-            fi
-
-            nix run github:huggingface/kernels#kernel-builder -- build-and-upload \
-              --max-jobs 4 \
-              --cores 8 \
-              --repo-id hao-ai-lab/fastvideo-kernel \
-              "${BRANCH_ARGS[@]}"

.github/workflows/publish-kernel.yml

@@ -62,12 +62,13 @@ jobs:
               cuda-version: '13.0.0'
               torch-cuda-short: 'cu130'
           platform:
-            # x86_64 builds the full cu126 + cu130 set (cu130 ships the consumer
-            # Blackwell sm_120a FP4 kernels).
+            # x86_64 builds the full cu126 + cu130 set. cu130 ships the
+            # data-center Blackwell sm_100a/sm_103a VSA and consumer sm_120a FP4
+            # kernels.
             - os: ubuntu-22.04
               arch: x86_64
               wheel-plat: manylinux_2_35_x86_64
-          # aarch64 is Blackwell (GB200 sm_100a + DGX Spark / consumer sm_120a), not
+          # aarch64 is Blackwell (GB200 sm_100a/sm_103a + sm_120a + DGX Spark sm_121a), not
           # Hopper, and Blackwell needs CUDA >= 12.8 — so only the cu130 leg applies.
           # Added via include so x86 keeps cu126 + cu130 while aarch64 stays cu130-only.
           include:
@@ -124,7 +125,7 @@ jobs:
       - name: Install dependencies (GCC, Clang, CUDA Paths, Git)
         run: |
           sudo apt update
-          sudo apt install -y git patchelf gcc-11 g++-11 clang-11
+          sudo apt install -y git gcc-11 g++-11 clang-11
           sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11
 
           # Allow Git to Access Safe Directory
@@ -163,22 +164,24 @@ jobs:
           cd fastvideo-kernel
           git submodule update --init --recursive  # Ensure ThunderKittens submodule is initialized
           # Release builds run on GPU-less runners, so set kernels + arch explicitly:
-          #   * aarch64 = Blackwell (GB200 sm_100a + DGX Spark/consumer sm_120a), NOT
-          #     Hopper, so TK (sm_90a wgmma) is OFF. The C++ FP4 (attn_qat_infer, SM120)
-          #     covers sm_120a; turbodiffusion covers sm_100a+sm_120a. The sm_100 FP4
-          #     forward is the FA4 CuTe DSL path in the fastvideo package (PR #1221),
+          #   * aarch64 = Blackwell (GB200 sm_100a/sm_103a + sm_120a + DGX Spark sm_121a), NOT
+          #     Hopper, so TK (sm_90a wgmma) is OFF. The C++ FP4 (attn_qat_infer)
+          #     covers sm_120a+sm_121a; turbodiffusion covers every listed arch. The
+          #     sm_100 FP4 forward is the FA4 CuTe DSL path in the fastvideo package (PR #1221),
           #     JIT-compiled at runtime — not built into this wheel.
-          #   * x86_64 cu130 = Hopper TK + consumer Blackwell sm_120a FP4.
+          #   * x86_64 cu130 = Hopper TK + data-center Blackwell sm_100a/sm_103a VSA
+          #     + consumer Blackwell sm_120a FP4.
           #   * x86_64 cu126 = Hopper TK only (older drivers; CUDA < 12.8 has no FP4).
-          # The per-arch split in CMakeLists pins the FP4 targets to sm_120a and builds
-          # the main extension for the full arch list. CMAKE_BUILD_PARALLEL_LEVEL caps
+          # The per-arch split in CMakeLists pins the FP4 targets to requested
+          # sm_120a/sm_121a and builds the main extension for the full arch list.
+          # CMAKE_BUILD_PARALLEL_LEVEL caps
           # Ninja so heavy CUTLASS/TK template TUs don't OOM the 16 GB runner (exit 143).
           if [ "${{ matrix.platform.arch }}" = "aarch64" ]; then
-            export TORCH_CUDA_ARCH_LIST="10.0a;12.0a"
+            export TORCH_CUDA_ARCH_LIST="10.0a;10.3a;12.0a;12.1a"
             export CMAKE_ARGS="${CMAKE_ARGS:-} -DFASTVIDEO_KERNEL_BUILD_TK=OFF -DFASTVIDEO_KERNEL_BUILD_ATTN_QAT_INFER=ON"
             export CMAKE_BUILD_PARALLEL_LEVEL=1
           elif [ "${{ matrix.torch-cuda.torch-cuda-short }}" = "cu130" ]; then
-            export TORCH_CUDA_ARCH_LIST="9.0a;12.0a"
+            export TORCH_CUDA_ARCH_LIST="9.0a;10.0a;10.3a;12.0a"
             export CMAKE_ARGS="${CMAKE_ARGS:-} -DFASTVIDEO_KERNEL_BUILD_TK=ON -DFASTVIDEO_KERNEL_BUILD_ATTN_QAT_INFER=ON -DCMAKE_CUDA_ARCHITECTURES=90a"
             # A single FP4 TU (attn_qat_infer) can use ~8-12 GB on its own, so serialize.
             export CMAKE_BUILD_PARALLEL_LEVEL=1
@@ -194,7 +197,11 @@ jobs:
           python -m build --wheel --outdir dist
 
           # Fix the wheel to be manylinux compliant
-          uv pip install --system auditwheel
+          # Ubuntu 22.04 ships patchelf 0.14.3, while current auditwheel
+          # requires at least 0.14.5. Use the stable PyPI binary on both
+          # x86_64 and aarch64 release runners.
+          uv pip install --system auditwheel patchelf==0.17.2.4
+          patchelf --version
           # Point auditwheel at torch libs, but do not vendor them into the wheel.
           TORCH_LIB_DIR=$(python - <<'PY'
           import os
@@ -211,7 +218,8 @@ jobs:
             --exclude libtorch.so \
             --exclude libc10.so \
             --exclude libc10_cuda.so \
-            --exclude libtorch_python.so
+            --exclude libtorch_python.so \
+            --exclude libnccl.so.2
           # Move fixed wheels back to dist for upload consistency
           rm dist/*.whl
           mv fixed_dist/*.whl dist/
@@ -243,7 +251,8 @@ jobs:
       - name: Download PyPI wheels
         # Publish the cu130 (CUDA 13) wheels to PyPI for both architectures:
         #   x86_64  — Hopper sm_90a TK + consumer Blackwell sm_120a FP4
-        #   aarch64 — Blackwell: turbodiffusion (sm_100a/sm_120a) + C++ FP4 (sm_120a);
+        #   aarch64 — Blackwell: turbodiffusion (sm_100a/sm_103a/sm_120a/sm_121a)
+        #             + C++ FP4 (sm_120a/sm_121a);
         #             no TK (Hopper). sm_100 FP4 forward is the FA4 CuTe DSL path in the
         #             fastvideo package (#1221), shipped/JIT separately.
         # The x86_64 cu126 wheel stays available as a build artifact / GitHub-release asset.

.gitignore

@@ -6,6 +6,7 @@ results/
 wandb/
 *.ipynb
 *.jpg
+!examples/datasets/lingbotworld2/image.jpg
 *.safetensors
 *.mp4
 *.png
@@ -22,6 +23,7 @@ Miniconda3-latest-Linux-x86_64.sh
 *validation/
 data/
 outputs/
+outputs_audio/
 outputs_video
 checkpoints/
 sbatch.sh
@@ -34,6 +36,7 @@ env
 *.log
 weights/
 logs/
+/Z-Image/
 official_weights/
 converted_weights/
 
@@ -52,6 +55,8 @@ eggs/
 
 # MkDocs documentation
 site/
+docs/assets/cookbook-serving.json
+examples/serving/clients/node_modules/
 docs/getting_started/examples/
 docs/examples/
 docs/inference/examples/
@@ -72,8 +77,8 @@ docs/distillation/examples/
 # Python pickle files
 *.pkl
 
-# Reference videos
-!fastvideo/tests/ssim/reference_videos/**/*.mp4
+# Reference videos (negations must come after the catch-all on line below)
+!fastvideo/tests/nightly/reference_video_*.mp4
 
 # Static images
 !docs/assets/images/**/*.png
@@ -127,8 +132,18 @@ apps/dreamverse/web/.env.production.local
 .sisyphus/
 openspec/
 fastvideo/tests/ssim/reference_videos/**
+!fastvideo/tests/ssim/reference_videos/**/*.mp4
+!fastvideo/tests/ssim/reference_videos/**/*.png
+fastvideo/tests/ssim/.reference_videos_download.lock
+
+# Local H3 MLX kernel / exactness benches (JSON, logs, frames, videos)
+.kernel_bench/
 
 # Editor logs and local Python version pins (accidentally committed)
 *.nvimlog
 .nvimlog
 .python-version
+/LTX-2-Reference/
+/DFDReference/
+scripts/benchmarks/minimax_h3_pro6000/headline_results/
+fastvideo/tests/ssim/.reference_videos_download.lock

.gitmodules

@@ -7,3 +7,6 @@
 [submodule "fastvideo/third_party/eval/vbench"]
 	path = fastvideo/third_party/eval/vbench
 	url = https://github.com/Vchitect/VBench.git
+[submodule "fastvideo/third_party/eval/vqeval"]
+	path = fastvideo/third_party/eval/vqeval
+	url = https://github.com/JiusiServe/LongVideoSparseAttention.git

.pre-commit-config.yaml

@@ -9,7 +9,7 @@ exclude: |
     tests/.*|
     scripts/.*|
     fastvideo/dataset/.*|
-    fastvideo/models/.*|
+    fastvideo/models/(?!wan/(config|vae_config|pipeline_config|definition|__init__)\.py$).*|
     ^apps/dreamverse/web/.*|
     examples/.*|
     \.agents/.*|
@@ -22,6 +22,7 @@ repos:
   hooks:
   - id: yapf
     args: [--in-place, --verbose]
+    language_version: python3.12
     additional_dependencies: [toml] # TODO: Remove when yapf is upgraded
 - repo: https://github.com/astral-sh/ruff-pre-commit
   rev: v0.11.12

AGENTS.md

@@ -66,14 +66,18 @@ Local guidance lives next to the code. Read the in-scope file before editing:
 | `fastvideo/AGENTS.md` | Core package map, public API, registry-driven model dispatch |
 | `fastvideo/configs/AGENTS.md` | Arch + pipeline config dataclasses, `param_names_mapping` |
 | `fastvideo/models/AGENTS.md` | DiT / VAE / encoder / scheduler / loader layout (pre-commit excluded) |
+| `fastvideo/models/wan/AGENTS.md` | Wan family-local transformers, VAE, configs, and the SP sharding invariant |
 | `fastvideo/layers/AGENTS.md` | Tensor-parallel linear/attention layer rules for ports |
 | `fastvideo/attention/AGENTS.md` | Backend registry + env-var override |
 | `fastvideo/pipelines/AGENTS.md` | Stage ABC, `basic/<model>/`, `preprocess/`, presets |
+| `fastvideo/pipelines/basic/wan/AGENTS.md` | Wan sampling stages, first-frame conditioning, DMD/causal boundaries |
+| `fastvideo/pipelines/basic/magi_human/AGENTS.md` | MagiHuman umbrella repo, lazy-loaded components, packing invariants |
 | `fastvideo/training/AGENTS.md` | Legacy monolithic pipelines (frozen for existing models) |
 | `fastvideo/train/AGENTS.md` | New modular trainer (methods × models × callbacks, YAML) |
 | `fastvideo/tests/AGENTS.md` | Test taxonomy, conftest, pre-commit-excluded path |
 | `fastvideo/tests/ssim/AGENTS.md` | GPU SSIM regression authoring + reference video sync |
 | `scripts/checkpoint_conversion/AGENTS.md` | Adding a converter for a new HF/official checkpoint |
+| `apps/dreamverse/AGENTS.md` | DreamVerse app structure and conventions |
 
 ## Critical: Two Training Stacks Coexist
 

README.md

@@ -3,13 +3,19 @@
 </div>
 
 <p align="center">
-     | <a href="https://hao-ai-lab.github.io/FastVideo"><b>Documentation</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start/"><b> Quick Start</b></a> | <a href="https://github.com/hao-ai-lab/FastVideo/discussions/982"  target="_blank"><b>Weekly Dev Meeting</b></a>  | 🟣💬 <a href="https://join.slack.com/t/fastvideo/shared_invite/zt-3f4lao1uq-u~Ipx6Lt4J27AlD2y~IdLQ" target="_blank"> <b>Slack</b> </a> |  🟣💬 <a href="https://github.com/hao-ai-lab/FastVideo/discussions/1097" target="_blank"> <b> WeChat </b> </a> |
+     | <a href="https://hao-ai-lab.github.io/FastVideo"><b>Documentation</b></a> | <a href="https://haoailab.com/FastVideo/cookbook/"><b>Cookbook</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start/"><b> Quick Start</b></a> | <a href="https://github.com/hao-ai-lab/FastVideo/discussions/982"  target="_blank"><b>Weekly Dev Meeting</b></a>  | 🟣💬 <a href="https://join.slack.com/t/fastvideo/shared_invite/zt-3f4lao1uq-u~Ipx6Lt4J27AlD2y~IdLQ" target="_blank"> <b>Slack</b> </a> |  🟣💬 <a href="https://github.com/hao-ai-lab/FastVideo/discussions/1097" target="_blank"> <b> WeChat </b> </a> |
 </p>
 
 **FastVideo is a unified post-training and real-time inference framework for accelerated video generation.**
 
 ## NEWS
-- `2026/06/23`: Release FastWan-QAD: 5s of Video generated in 1.8s E2E. [FastWan-QAD models](https://huggingface.co/FastVideo/FastWan-QAD-FP8-1.3B), check out the [Blog](https://haoailab.com/blogs/fastwan-qad/).
+- `2026/10/06`: FastH3 V2 now runs on a single consumer machine: NVIDIA RTX 5090, RTX 4090 and RTX PRO 6000 GPUs, DGX Spark and Apple Silicon. We also release [FastH3 Trim](https://huggingface.co/FastVideo/FastVideo-FastH3-Trim-8-Step-NVFP4), an experimental pruned model that is 4.2× smaller than base H3 and runs in as little as 8 GB of GPU memory. Get the [models](https://huggingface.co/collections/FastVideo/fastvideo-fasth3) and read the [Blog](https://haoailab.com/blogs/fasth3-rtx/).
+- `2026/10/06`: FastVideo now supports [Kandinsky 6](https://x.com/kandinskylab_ai/status/2107374635218055345) from Kandinsky Lab: text- and image-to-video with synchronized audio (base and 10-step distilled pi-Flow checkpoints) plus video super-resolution up to 4x. See the [Kandinsky 6 recipes](https://haoailab.com/FastVideo/cookbook/kandinsky6/).
+- `2026/09/15`: Release [FastH3 8-Step V2](https://huggingface.co/FastVideo/FastVideo-FastH3-8-Step-V2), an eight-forward data-free DMD2 checkpoint distilled from MiniMax-H3 with 80% Video Sparse Attention. Run it with `examples/inference/basic/basic_fasth3_8step.py` or the [FastH3 8-Step V2 recipe](https://haoailab.com/FastVideo/cookbook/minimax-h3/).
+- `2026/09/01`: FastH3 now runs locally on Apple Silicon through MLX and on NVIDIA DGX Spark through CUDA 13, including two-Spark inference. Follow the [FastH3 recipes](https://haoailab.com/FastVideo/cookbook/minimax-h3/) and read the [Blog](https://haoailab.com/blogs/fasth3-local/).
+- `2026/08/27`: [FastH3 Preview v1](https://haoailab.com/blogs/fasth3-preview/) is an open-weight 4-step sparse-distilled MiniMax-H3 model for synchronized video-and-audio generation, developed in collaboration with [Nuva Lab](https://nuvalab.ai/) and the [NVIDIA FastGen team](https://github.com/NVlabs/FastGen). Download the recommended [VSA / Data-Free weights](https://huggingface.co/FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree), or see the [full FastH3 collection](https://huggingface.co/collections/FastVideo/fastvideo-fasth3).
+- `2026/08/19`: FastVideo now supports MLX on Apple Silicon with [FastMetal-QAD](https://huggingface.co/collections/FastVideo/fastmetal), a family of 1.3B, 5B, and 14B models optimized for Mac. Follow the [MLX install guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/mlx/) and read the [Blog](https://haoailab.com/blogs/fastmetal/).
+- `2026/06/23`: Release FastWan-QAD: 5s of Video generated in 1.8s E2E. See the [FastWan-QAD models](https://huggingface.co/FastVideo/FastWan-QAD-FP8-1.3B), [Attn-QAT training guide](https://haoailab.com/FastVideo/training/attn_qat/), and [blog](https://haoailab.com/blogs/fastwan-qad/).
 - `2026/03/17`: Release demo: Into the Dreamverse: Vibe Directing in FastVideo, check out the [Blog](https://haoailab.com/blogs/dreamverse/).
 - `2026/03/13`: Release demo: Create a 5s 1080p Video in 4.5s with FastVideo on a Single GPU, check out the [Blog](https://haoailab.com/blogs/fastvideo_realtime_1080p/).
 - `2025/11/19`: Release [CausalWan2.2 I2V A14B Preview](https://huggingface.co/FastVideo/CausalWan2.2-I2V-A14B-Preview-Diffusers) models, [Blog](https://hao-ai-lab.github.io/blogs/fastvideo_causalwan_preview/) and [Inference Code!](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_self_forcing_causal_wan2_2_i2v.py).
@@ -33,7 +39,7 @@ FastVideo has the following features:
   - [Sparse distillation](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/) to achieve >50x denoising speedup
   - Scalable training with FSDP2, sequence parallelism, and selective activation checkpointing.
   - Causal distillation through Self-Forcing
-  - See this [page](https://hao-ai-lab.github.io/FastVideo/training/overview/) for full list of supported models and recipes.
+  - See this [page](https://hao-ai-lab.github.io/FastVideo/training/overview/) for the supported training workflows, and the [support matrix](https://hao-ai-lab.github.io/FastVideo/inference/support_matrix/) for supported models.
 - State-of-the-art performance optimizations for inference
   - Sequence Parallelism for distributed inference
   - Multiple state-of-the-art attention backends
@@ -60,7 +66,12 @@ UV_TORCH_BACKEND=cu126 uv pip install fastvideo
 ```
 
 Use `UV_TORCH_BACKEND=cu130` on CUDA 13. Apple silicon users should follow the
-[MPS installation guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/mps/).
+[MLX install guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/mlx/).
+
+> **On an Apple Silicon Mac?** Install with `uv pip install -e '.[mlx]'` from
+> a clone, then pick a recipe in the
+> [cookbook](https://haoailab.com/FastVideo/cookbook/). See the
+> [MLX install guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/mlx/).
 
 Please see our [docs](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/) for more detailed installation instructions.
 
@@ -78,7 +89,7 @@ Install FastVideo (https://github.com/hao-ai-lab/FastVideo) into a fresh uv virt
    https://hao-ai-lab.github.io/FastVideo/getting_started/installation/):
      - NVIDIA GPU, x86_64                         -> docs/getting_started/installation/gpu.md
      - NVIDIA DGX Spark / GB10, aarch64, CUDA 13  -> docs/getting_started/installation/spark.md
-     - Apple Silicon, macOS                       -> docs/getting_started/installation/mps.md
+     - Apple Silicon, macOS                       -> docs/getting_started/installation/mlx.md
 3. Use uv for every step. If a command fails, debug it and tell me what you changed.
 4. Verify the result:
      python -c "import fastvideo, torch; print('cuda', torch.cuda.is_available())"
@@ -143,6 +154,8 @@ if __name__ == '__main__':
     main()
 ```
 
+`num_gpus=1` runs the worker in-process (weights load once, no extra Python process). On Colab/Kaggle-style machines with ~16GB host RAM, keep `num_gpus=1`; free-tier system memory does not grow with extra T4s, so `num_gpus>1` is likely to OOM.
+
 Run the script with:
 
 ```bash

apps/dreamverse/README.md

@@ -97,13 +97,33 @@ dreamverse-server --port 8009
 dreamverse-mock-server --port 8009
 ```
 
+### Run Dreamverse with FastH3
+
+Select the VSA data-free FastH3 Preview profile when you start the backend:
+
+```bash
+DREAMVERSE_MODEL_ID=fast-h3 dreamverse-server --port 8009
+```
+
+The `fast-h3` profile uses four visible GPUs by default. It loads the `MiniMaxAI/MiniMax-H3` base checkpoint and the
+`vsa-datafree/adapter_model.safetensors` adapter from
+`FastVideo/FastVideo-FastH3-4-step-Preview-v1-LoRA`. Each request generates a 124-frame, 768×1344 video with
+synchronized audio and five sigma-grid points. Dreamverse uses the last frame of each segment as first-frame
+conditioning for the following segment.
+
+Set `CUDA_VISIBLE_DEVICES` when you need to choose the four physical GPUs:
+
+```bash
+CUDA_VISIBLE_DEVICES=0,1,2,3 DREAMVERSE_MODEL_ID=fast-h3 dreamverse-server --port 8009
+```
+
 > **Expect a slow first boot.** With `torch.compile` and startup warmup enabled
 > (the default), the backend compiles the segment 1 and segment 2 inference
 > paths before it reports ready — this can take **tens of minutes on a cold
 > cache**, regardless of how you deploy (local, server, Docker, or Modal).
 > `/healthz` responds as soon as the process is up; `/readyz` stays `503` until
-> warmup finishes. For a faster, uncompiled startup while testing, set
-> `FASTVIDEO_ENABLE_STARTUP_WARMUP=0` before starting the backend.
+> warmup finishes. To defer compilation until the first generated request while
+> testing, set `FASTVIDEO_ENABLE_STARTUP_WARMUP=0` before starting the backend.
 
 ## Frontend Setup
 
@@ -138,6 +158,40 @@ dreamverse-server --host 0.0.0.0 --port 8009
 The Dreamverse backend defaults to `0.0.0.0:8009` and starts one GPU worker on
 the first visible GPU by default.
 
+### Cosmos Predict2.5 DFD continuation (experimental)
+
+Dreamverse can combine two converted Cosmos Predict2.5 2B packages: the
+distilled Text2World student creates an unconditioned first segment, then the
+Data-Forcing Distillation (DFD) Video2World student conditions each later
+segment on the prior terminal frame. Point the runtime at both local converted
+packages:
+
+```bash
+export DREAMVERSE_MODEL_ID=cosmos25-dfd
+export DREAMVERSE_MODEL_PATH=/path/to/Cosmos-Predict2.5-2B-Distilled-TrigFlow-FastVideo
+export DREAMVERSE_COSMOS25_DFD_MODEL_PATH=/path/to/Cosmos-Predict2.5-2B-DFD-FastVideo
+export ENABLE_TORCH_COMPILE=0
+dreamverse-server --host 0.0.0.0 --port 8009
+```
+
+The backend loads and warms both model roles before reporting ready. Both use
+BF16, Torch SDPA, 704x1280 output, 24 FPS, and four steps. Bootstrap segments
+contain 77 frames. DFD segments contain 81 decoded frames, but Dreamverse drops
+the repeated conditioning frame before streaming, leaving 80 new frames. An
+initial user image selects DFD immediately without treating that first frame as
+a cross-segment overlap.
+
+The profile uses a 30-minute session lease because sequential generation on
+GB10-class hardware can exceed Dreamverse's five-minute default while the GPU
+is still making progress. Deployments can override the lease with
+`FASTVIDEO_SESSION_TIMEOUT_SECONDS`.
+
+Cosmos does not produce audio, so the backend supplies duration-matched silent
+24 kHz audio for the existing browser streaming contract and trims 1,000 audio
+samples with each repeated DFD boundary frame. Runtime LoRA changes are not
+supported. Full segments take roughly 145 seconds on GB10, so this profile is a
+continuation-quality integration rather than a real-time configuration.
+
 ### Check Readiness
 
 In another shell, verify that the backend process is alive:
@@ -219,6 +273,7 @@ selection, and mock-server behavior:
 pytest apps/dreamverse/dreamverse/tests/test_config.py \
   apps/dreamverse/dreamverse/tests/test_entrypoints.py \
   apps/dreamverse/dreamverse/tests/test_gpu_pool.py \
+  apps/dreamverse/dreamverse/tests/test_minimax_h3_generation.py \
   apps/dreamverse/dreamverse/tests/test_mock_server.py -q
 ```
 

apps/dreamverse/arch.md

@@ -139,7 +139,18 @@ session.
 - startup warmup
 - user join/leave commands
 - `USER_STEP` execution for each segment
-- continuation state between segments
+- generation-command routing and stream-result delivery
+
+Model generation has a separate ownership boundary inside each GPU process:
+
+- `apps/dreamverse/dreamverse/generation_worker.py` selects the backend that the active model profile declares and owns
+  the backend lifecycle.
+- `apps/dreamverse/dreamverse/ltx2_generation.py` owns LTX-2 generator configuration, video and audio continuation, and
+  runtime LoRA application.
+- `apps/dreamverse/dreamverse/minimax_h3_generation.py` owns the VSA data-free FastH3 adapter, FastH3 generator and
+  request configuration, and last-frame continuation through MiniMax H3 first-frame conditioning.
+- `apps/dreamverse/dreamverse/generation_contracts.py` defines the decoded media and stream-trimming result that both
+  model backends return to `apps/dreamverse/dreamverse/gpu_pool.py`.
 
 `apps/dreamverse/dreamverse/prompt_enhancer.py` manages:
 

apps/dreamverse/dreamverse/assets.py

@@ -0,0 +1,184 @@
+"""Bounded, runtime-local media library shared by the HTTP and generation APIs."""
+
+from __future__ import annotations
+
+import json
+import math
+import os
+import re
+import shutil
+import subprocess
+import tempfile
+import threading
+import uuid
+from dataclasses import dataclass
+from pathlib import Path
+
+from PIL import Image, UnidentifiedImageError
+
+IMAGE_LIMIT = 15 * 1024 * 1024
+MEDIA_LIMIT = 100 * 1024 * 1024
+STORE_LIMIT = 2 * 1024 * 1024 * 1024
+ASSET_LIMIT = 100
+MAX_MEDIA_SECONDS = 30
+MIME_TYPES = {
+    "image/png": ("image", ".png"),
+    "image/jpeg": ("image", ".jpg"),
+    "image/webp": ("image", ".webp"),
+    "video/mp4": ("video", ".mp4"),
+    "video/quicktime": ("video", ".mov"),
+    "video/webm": ("video", ".webm"),
+    "audio/mpeg": ("audio", ".mp3"),
+    "audio/mp4": ("audio", ".m4a"),
+    "audio/x-m4a": ("audio", ".m4a"),
+    "audio/wav": ("audio", ".wav"),
+    "audio/x-wav": ("audio", ".wav"),
+    "audio/flac": ("audio", ".flac"),
+    "audio/x-flac": ("audio", ".flac"),
+    "audio/ogg": ("audio", ".ogg"),
+    "audio/webm": ("audio", ".webm"),
+}
+
+
+@dataclass(frozen=True)
+class StoredAsset:
+    asset_id: str
+    kind: str
+    path: str
+    name: str
+    mime_type: str
+    size: int
+
+    def public(self) -> dict:
+        return {
+            "asset_id": self.asset_id,
+            "kind": self.kind,
+            "name": self.name,
+            "mime_type": self.mime_type,
+            "size": self.size,
+            "url": f"/assets/{self.asset_id}",
+        }
+
+
+def validate_media(path: Path, mime_type: str) -> None:
+    """Inspect content, not filenames; refuse playlists and non-media uploads."""
+    kind = MIME_TYPES[mime_type][0]
+    if kind == "image":
+        try:
+            with Image.open(path) as img:
+                expected = {"image/png": "PNG", "image/jpeg": "JPEG", "image/webp": "WEBP"}[mime_type]
+                if img.format != expected:
+                    raise ValueError("The image content does not match its file type.")
+                if img.width * img.height > 16_777_216:
+                    raise ValueError("Images must contain at most 16 megapixels.")
+                if getattr(img, "is_animated", False):
+                    raise ValueError("Use a still image or upload the animation as a video.")
+                img.verify()
+        except (UnidentifiedImageError, OSError, Image.DecompressionBombError) as exc:
+            raise ValueError("The image could not be decoded. Use PNG, JPEG, or WebP.") from exc
+        return
+
+    probe = shutil.which(os.getenv("FASTVIDEO_FFPROBE_BIN", "ffprobe"))
+    if not probe:
+        raise ValueError("This runtime needs ffprobe installed to accept video and audio assets.")
+    try:
+        result = subprocess.run(
+            [
+                probe, "-v", "error", "-protocol_whitelist", "file,pipe", "-format_whitelist",
+                "mov,matroska,webm,mp3,wav,flac,ogg", "-show_format", "-show_streams", "-of", "json",
+                str(path)
+            ],
+            check=True,
+            capture_output=True,
+            timeout=15,
+        )
+        info = json.loads(result.stdout)
+        formats = set(info.get("format", {}).get("format_name", "").split(","))
+        if not formats.intersection({"mov", "mp4", "matroska", "webm", "mp3", "wav", "flac", "ogg"}):
+            raise ValueError("Upload a media file, not a playlist or external reference.")
+        streams = [stream for stream in info.get("streams", []) if stream.get("codec_type") == kind]
+        if not streams:
+            raise ValueError(f"The file contains no {kind} stream.")
+        for stream in info.get("streams", []):
+            if stream.get("codec_type") == "audio" and int(stream.get("channels", 0)) not in (1, 2):
+                raise ValueError("H3 references require mono or stereo audio, including video soundtracks.")
+        duration = float(info.get("format", {}).get("duration", "nan"))
+        if not math.isfinite(duration) or not 0 < duration <= MAX_MEDIA_SECONDS:
+            raise ValueError(f"Reference video and audio must be between 0 and {MAX_MEDIA_SECONDS} seconds long.")
+        for stream in streams:
+            if kind == "video" and int(stream.get("width", 0)) * int(stream.get("height", 0)) > 8_294_400:
+                raise ValueError("Reference videos must be 4K or smaller.")
+    except (subprocess.SubprocessError, json.JSONDecodeError, OSError) as exc:
+        raise ValueError("The media file could not be decoded. Check its format and try again.") from exc
+
+
+class AssetStore:
+    """Assets live until deletion or runtime exit; pinned generation inputs cannot be deleted."""
+
+    def __init__(self) -> None:
+        self._directory: tempfile.TemporaryDirectory | None = None
+        self._assets: dict[str, StoredAsset] = {}
+        self._pins: dict[str, int] = {}
+        self._lock = threading.RLock()
+
+    def staging_path(self, mime_type: str) -> Path:
+        with self._lock:
+            if mime_type not in MIME_TYPES:
+                raise ValueError("Unsupported media type. Use PNG/JPEG/WebP, MP4/WebM/MOV, or WAV/MP3/M4A/FLAC/OGG.")
+            if len(self._assets) >= ASSET_LIMIT or sum(item.size for item in self._assets.values()) >= STORE_LIMIT:
+                raise ValueError("The runtime asset library is full. Remove unused assets before uploading more.")
+            if self._directory is None:
+                self._directory = tempfile.TemporaryDirectory(prefix="dreamverse-assets-")
+            return Path(self._directory.name) / f"{uuid.uuid4().hex}{MIME_TYPES[mime_type][1]}"
+
+    def add(self, path: Path, name: str, mime_type: str) -> StoredAsset:
+        validate_media(path, mime_type)
+        size = path.stat().st_size
+        if size == 0 or size > (IMAGE_LIMIT if MIME_TYPES[mime_type][0] == "image" else MEDIA_LIMIT):
+            raise ValueError("The asset is empty or exceeds its upload size limit.")
+        with self._lock:
+            if len(self._assets) >= ASSET_LIMIT or size + sum(item.size
+                                                              for item in self._assets.values()) > STORE_LIMIT:
+                raise ValueError("The runtime asset library is full. Remove unused assets before uploading more.")
+            if self._directory is None or path.parent != Path(self._directory.name):
+                raise ValueError("The asset must be uploaded to this runtime.")
+            display_name = re.sub(r"[\x00-\x1f\x7f/\\]", "_", name).strip()[:200] or "Untitled asset"
+            asset = StoredAsset(path.stem, MIME_TYPES[mime_type][0], str(path), display_name, mime_type, size)
+            self._assets[asset.asset_id] = asset
+            return asset
+
+    def get(self, asset_id: str) -> StoredAsset:
+        with self._lock:
+            if not isinstance(asset_id, str) or not re.fullmatch(r"[a-f0-9]{32}", asset_id):
+                raise ValueError("Invalid asset ID. Upload or select an asset from the library.")
+            asset = self._assets.get(asset_id)
+            if asset is None or not Path(asset.path).is_file():
+                raise ValueError("An asset is no longer available. Upload it again and reselect it.")
+            return asset
+
+    def pin(self, asset_ids: list[str]) -> None:
+        with self._lock:
+            for asset_id in asset_ids:
+                self.get(asset_id)
+            for asset_id in asset_ids:
+                self._pins[asset_id] = self._pins.get(asset_id, 0) + 1
+
+    def release(self, asset_ids: list[str]) -> None:
+        with self._lock:
+            for asset_id in asset_ids:
+                count = self._pins.get(asset_id, 0)
+                if count > 1:
+                    self._pins[asset_id] = count - 1
+                else:
+                    self._pins.pop(asset_id, None)
+
+    def delete(self, asset_id: str) -> None:
+        with self._lock:
+            asset = self.get(asset_id)
+            if self._pins.get(asset_id, 0):
+                raise ValueError("This asset is in use by a generation session. End the session before deleting it.")
+            Path(asset.path).unlink(missing_ok=True)
+            del self._assets[asset_id]
+
+
+asset_store = AssetStore()

apps/dreamverse/dreamverse/benchmarks/benchmark_pipeline.py

@@ -1,6 +1,6 @@
 """Benchmark the LTX-2 generation pipeline driven by the dreamverse Python SDK path.
 
-Mirrors how ``apps/dreamverse/dreamverse/video_generation.py`` constructs
+Mirrors how ``apps/dreamverse/dreamverse/ltx2_generation.py`` constructs
 ``GeneratorConfig`` and calls ``VideoGenerator.generate()``, then
 captures per-stage timings via the ``FASTVIDEO_STAGE_LOGGING=1`` log
 hooks (same mechanism as ``FastVideo-internal/examples/inference/basic/
@@ -111,7 +111,13 @@ def _build_generator_config(model_path: str, enable_compile: bool, num_gpus: int
                                   mode="max-autotune-no-cudagraphs",
                                   dynamic=False),
             use_fsdp_inference=False,
-            quantization=QuantizationConfig(transformer_quant="NVFP4"),
+            # The bundled LTX2 model enables a refinement LoRA during the
+            # first request. NVFP4 otherwise purges the dense weights that
+            # FastVideo's LoRA merge path requires.
+            quantization=QuantizationConfig(
+                transformer_quant="NVFP4",
+                transformer_retain_original_weights=True,
+            ),
         ),
         pipeline=PipelineSelection(
             components=components,

apps/dreamverse/dreamverse/config.py

@@ -1,5 +1,6 @@
 import os
 from pathlib import Path
+from typing import cast
 
 _REPO_ROOT = Path(__file__).resolve().parents[1]
 _SERVER_ROOT = Path(__file__).resolve().parent
@@ -55,16 +56,65 @@ FRONTEND_STATIC_DIR_CANDIDATES = _resolve_frontend_static_dir_candidates()
 MODEL_REGISTRY = {
     "fast-ltx2": {
         "name": "FastLTX2",
+        "generation_backend": "ltx2",
+        "default_sp_size": 1,
         "model_path": "FastVideo/LTX2-Distilled-Diffusers",
         "config_model_path": "FastVideo/LTX2-Distilled-Diffusers",
         "lora_repo": "FastVideo/LTX2-OmniNFT-LoRA",
     },
     "fast-ltx23": {
         "name": "FastLTX23",
+        "generation_backend": "ltx2",
+        "default_sp_size": 1,
         "model_path": "FastVideo/LTX-2.3-Distilled-Diffusers",
         "config_model_path": "FastVideo/LTX-2.3-Distilled-Diffusers",
         "lora_repo": "FastVideo/LTX-2.3-OmniNFT-LoRA",
     },
+    "fast-h3": {
+        "name": "FastH3",
+        "generation_backend": "minimax_h3",
+        "default_sp_size": 4,
+        "model_path": "MiniMaxAI/MiniMax-H3",
+        "adapter_repo": "FastVideo/FastVideo-FastH3-4-step-Preview-v1-LoRA",
+        "adapter_filename": "vsa-datafree/adapter_model.safetensors",
+        "attention_backend": "VIDEO_SPARSE_ATTN_H3",
+        "height": 768,
+        "width": 1344,
+        "num_frames": 124,
+        "num_inference_steps": 5,
+        "seed": 1000,
+    },
+    "full-h3": {
+        "name": "MiniMax H3 (Full)",
+        "generation_backend": "minimax_h3",
+        "default_sp_size": 4,
+        "model_path": "MiniMaxAI/MiniMax-H3",
+        "attention_backend": "FLASH_ATTN",
+        "height": 768,
+        "width": 1344,
+        "num_frames": 124,
+        "num_inference_steps": 50,
+        "seed": 1000,
+        "full_checkpoint": True,
+    },
+    "cosmos25-dfd": {
+        "name": "Cosmos Predict2.5 DFD",
+        "generation_backend": "cosmos25_dfd",
+        "default_sp_size": 1,
+        "model_path": "FastVideo/Cosmos-Predict2.5-2B-Distilled-TrigFlow",
+        "continuation_model_path": "FastVideo/Cosmos-Predict2.5-2B-DFD",
+        "attention_backend": "TORCH_SDPA",
+        "height": 704,
+        "width": 1280,
+        "bootstrap_num_frames": 77,
+        "continuation_num_frames": 81,
+        "fps": 24,
+        "num_inference_steps": 4,
+        "seed": 42,
+        # Six sequential GB10 segments can exceed the legacy five-minute
+        # DreamVerse lease even though the GPU is making progress.
+        "session_timeout_seconds": 1800,
+    },
 }
 
 DEFAULT_MODEL_ID = "fast-ltx2"
@@ -76,22 +126,6 @@ if ACTIVE_MODEL_ID not in MODEL_REGISTRY:
 # Active model configuration
 MODEL_CONFIG = MODEL_REGISTRY[ACTIVE_MODEL_ID]
 
-# Generation limits
-SESSION_TIMEOUT_SECONDS = 300
-
-# Frame settings
-NUM_FRAMES = 121
-FRAME_HEIGHT = 1088
-FRAME_WIDTH = 1920
-NUM_INFERENCE_STEPS = 5
-JPEG_QUALITY = 100
-BATCH_SIZE = 3
-
-# Streaming mode:
-# - legacy_jpeg: send frame_batch JSON payloads with base64 JPEGs
-# - av_fmp4: send muxed fMP4 binary chunks over WebSocket
-STREAM_MODE = os.getenv("STREAM_MODE", "av_fmp4").strip().lower()
-
 
 def _env_int(name: str, default: int) -> int:
     value = os.getenv(name)
@@ -168,10 +202,41 @@ def _optional_env(*names: str) -> str | None:
     return None
 
 
+# Generation limits
+# Slower backends may own a longer default lease. A profile can set
+# ``session_timeout_seconds``; Full H3 loads and generates substantially longer
+# than the Preview adapter, which also covers a base/ref pipeline reload inside a
+# retained session. An explicit environment override remains available for
+# deployment policy: DREAMVERSE_SESSION_TIMEOUT_SECONDS, with
+# FASTVIDEO_SESSION_TIMEOUT_SECONDS accepted as an alias.
+# Values below 60 seconds are floored so a single segment cannot outlast the session.
+_DEFAULT_SESSION_TIMEOUT_SECONDS = cast(
+    int, MODEL_CONFIG.get("session_timeout_seconds", 7200 if ACTIVE_MODEL_ID == "full-h3" else 300))
+SESSION_TIMEOUT_SECONDS = max(
+    60,
+    _env_int(
+        "DREAMVERSE_SESSION_TIMEOUT_SECONDS",
+        _env_int("FASTVIDEO_SESSION_TIMEOUT_SECONDS", _DEFAULT_SESSION_TIMEOUT_SECONDS),
+    ),
+)
+
+# Frame settings
+NUM_FRAMES = 121
+FRAME_HEIGHT = 1088
+FRAME_WIDTH = 1920
+NUM_INFERENCE_STEPS = 5
+JPEG_QUALITY = 100
+BATCH_SIZE = 3
+
+# Streaming mode:
+# - legacy_jpeg: send frame_batch JSON payloads with base64 JPEGs
+# - av_fmp4: send muxed fMP4 binary chunks over WebSocket
+STREAM_MODE = os.getenv("STREAM_MODE", "av_fmp4").strip().lower()
+
 DEVTOOLS_ENABLED = _env_bool("FASTVIDEO_ENABLE_DEVTOOLS", False)
 PROMPT_SAFETY_ENABLED = _env_bool("FASTVIDEO_ENABLE_PROMPT_SAFETY", False)
 DREAMVERSE_MAX_AUTOTUNE = _env_bool("DREAMVERSE_MAX_AUTOTUNE", True)
-DREAMVERSE_SP_SIZE = max(1, _env_int("DREAMVERSE_SP_SIZE", 1))
+DREAMVERSE_SP_SIZE = max(1, _env_int("DREAMVERSE_SP_SIZE", cast(int, MODEL_CONFIG["default_sp_size"])))
 
 DREAMVERSE_MODEL_PATH = (os.getenv("DREAMVERSE_MODEL_PATH", "").strip() or None)
 if DREAMVERSE_MODEL_PATH:
@@ -181,6 +246,13 @@ if DREAMVERSE_MODEL_PATH:
         "config_model_path": DREAMVERSE_MODEL_PATH,
     }
 
+DREAMVERSE_COSMOS25_DFD_MODEL_PATH = (os.getenv("DREAMVERSE_COSMOS25_DFD_MODEL_PATH", "").strip() or None)
+if DREAMVERSE_COSMOS25_DFD_MODEL_PATH and MODEL_CONFIG.get("generation_backend") == "cosmos25_dfd":
+    MODEL_CONFIG = {
+        **MODEL_CONFIG,
+        "continuation_model_path": DREAMVERSE_COSMOS25_DFD_MODEL_PATH,
+    }
+
 AVAILABLE_LORAS = {
     "pixar": {
         "repo": "vrgamedevgirl84/LTX_2.3_Pixar_Toon_Style_LoRa",
@@ -213,7 +285,7 @@ def _resolve_lora_spec(spec: str) -> str | None:
     if not spec:
         return None
     if spec.lower() == "omninft":
-        return MODEL_CONFIG.get("lora_repo")
+        return cast(str | None, MODEL_CONFIG.get("lora_repo"))
     if spec.lower() in AVAILABLE_LORAS:
         return AVAILABLE_LORAS[spec.lower()]["repo"]
     return spec

apps/dreamverse/dreamverse/cosmos25_dfd_generation.py

@@ -0,0 +1,252 @@
+"""Cosmos Predict2.5 distilled bootstrap and DFD continuation for DreamVerse."""
+
+from __future__ import annotations
+
+import gc
+import os
+import time
+from typing import TYPE_CHECKING, Any
+
+import numpy as np
+import torch
+
+from dreamverse.generation_contracts import StepResult
+from dreamverse.generation_inputs import GenerationInputs
+
+if TYPE_CHECKING:
+    from PIL.Image import Image
+
+_SILENT_AUDIO_SAMPLE_RATE = 24_000
+
+
+def _required_config_str(model_config: dict, field_name: str) -> str:
+    value = model_config.get(field_name)
+    if not isinstance(value, str) or not value.strip():
+        raise ValueError(f"Cosmos Predict2.5 DFD model configuration requires `{field_name}`.")
+    return value.strip()
+
+
+class Cosmos25DFDGenerationBackend:
+    """Own complementary Cosmos T2W and one-frame-conditioned DFD generators."""
+
+    def __init__(self, gpu_id: int):
+        self.gpu_id = gpu_id
+        self.bootstrap_generator: Any | None = None
+        self.continuation_generator: Any | None = None
+        self.model_config: dict = {}
+        self.continuation_image: Image | None = None
+
+    def _gpu_mem(self) -> str:
+        allocated_gib = torch.cuda.memory_allocated() / 1024**3
+        reserved_gib = torch.cuda.memory_reserved() / 1024**3
+        return f"alloc={allocated_gib:.2f}GiB, reserved={reserved_gib:.2f}GiB"
+
+    @staticmethod
+    def _configure_environment(attention_backend: str) -> None:
+        os.environ["FASTVIDEO_ATTENTION_BACKEND"] = attention_backend
+        os.environ.pop("FASTVIDEO_INFERENCE_TORCH_COMPILE", None)
+
+    @staticmethod
+    def _load_generator(model_path: str):
+        from fastvideo import VideoGenerator
+
+        return VideoGenerator.from_pretrained(
+            model_path,
+            num_gpus=1,
+            use_fsdp_inference=False,
+            dit_cpu_offload=False,
+            vae_cpu_offload=False,
+            text_encoder_cpu_offload=True,
+            pin_cpu_memory=True,
+            enable_torch_compile=False,
+        )
+
+    def initialize(self, model_config: dict | None = None) -> None:
+        """Load both package roles so bootstrap and continuation are ready."""
+        if model_config is not None:
+            self.model_config = dict(model_config)
+        if not self.model_config:
+            raise ValueError("Cosmos Predict2.5 DFD initialization requires a model configuration.")
+
+        self.shutdown()
+        bootstrap_path = _required_config_str(self.model_config, "model_path")
+        continuation_path = _required_config_str(self.model_config, "continuation_model_path")
+        attention_backend = _required_config_str(self.model_config, "attention_backend")
+        self._configure_environment(attention_backend)
+
+        print(f"[GPU {self.gpu_id}] Loading Cosmos T2W bootstrap: {bootstrap_path}")
+        print(f"[GPU {self.gpu_id}] Before bootstrap load: {self._gpu_mem()}")
+        self.bootstrap_generator = self._load_generator(bootstrap_path)
+        print(f"[GPU {self.gpu_id}] Loading Cosmos DFD continuation: {continuation_path}")
+        self.continuation_generator = self._load_generator(continuation_path)
+        print(f"[GPU {self.gpu_id}] Cosmos T2W + DFD loaded: {self._gpu_mem()} (warmup pending)")
+
+    def shutdown(self) -> None:
+        """Release both FastVideo generators and the retained terminal frame."""
+        self.clear_conditioning()
+        for attr_name in ("bootstrap_generator", "continuation_generator"):
+            generator = getattr(self, attr_name)
+            if generator is not None:
+                try:
+                    generator.shutdown()
+                except Exception as exc:
+                    print(f"[GPU {self.gpu_id}] Cosmos generator shutdown warning: {exc}")
+                setattr(self, attr_name, None)
+        gc.collect()
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+
+    def clear_conditioning(self) -> None:
+        if self.continuation_image is not None:
+            self.continuation_image.close()
+            self.continuation_image = None
+
+    @staticmethod
+    def _load_rgb_image(image_path: str) -> Image:
+        from PIL import Image
+
+        with Image.open(image_path) as image:
+            return image.convert("RGB").copy()
+
+    def _select_conditioning_image(
+        self,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+    ) -> tuple[Image | None, bool]:
+        if reset_conditioning:
+            self.clear_conditioning()
+        if segment_idx > 1 and self.continuation_image is not None:
+            return self.continuation_image.copy(), True
+        if segment_idx > 1 and not reset_conditioning:
+            raise RuntimeError(f"Cosmos DFD segment {segment_idx} requires a retained continuation frame.")
+        if segment_idx == 1 and image_path:
+            return self._load_rgb_image(image_path), False
+        return None, False
+
+    def _sampling_param(self, *, conditioned: bool):
+        # ``num_cond_frames`` is not yet exposed by the typed SamplingConfig,
+        # so this backend uses the compatibility request until that field lands.
+        from fastvideo.api.sampling_param import SamplingParam
+
+        num_frames_key = "continuation_num_frames" if conditioned else "bootstrap_num_frames"
+        return SamplingParam(
+            negative_prompt="",
+            save_video=False,
+            return_frames=True,
+            height=int(self.model_config["height"]),
+            width=int(self.model_config["width"]),
+            num_frames=int(self.model_config[num_frames_key]),
+            fps=int(self.model_config["fps"]),
+            num_inference_steps=int(self.model_config["num_inference_steps"]),
+            guidance_scale=1.0,
+            seed=int(self.model_config["seed"]),
+            num_cond_frames=1 if conditioned else 0,
+        )
+
+    def _save_continuation_frame(self, frame: object) -> None:
+        from PIL import Image
+
+        self.clear_conditioning()
+        if isinstance(frame, Image.Image):
+            self.continuation_image = frame.convert("RGB").copy()
+            return
+        pixels = np.asarray(frame)
+        self.continuation_image = Image.fromarray(np.ascontiguousarray(pixels)).convert("RGB")
+
+    @staticmethod
+    def _silent_audio(frame_count: int, fps: int) -> torch.Tensor:
+        sample_count = max(1, int(round((frame_count / float(fps)) * _SILENT_AUDIO_SAMPLE_RATE)))
+        return torch.zeros(sample_count, dtype=torch.float32)
+
+    def generate_step(
+        self,
+        prompt: str,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+        generation_inputs: GenerationInputs | None = None,
+    ) -> StepResult:
+        """Generate a T2W start or DFD continuation and retain its last frame."""
+        if generation_inputs is not None and (generation_inputs.mode not in (None, "t2va") or generation_inputs.assets):
+            raise ValueError("Cosmos supports text generation only through the generation mode API.")
+        if self.bootstrap_generator is None or self.continuation_generator is None:
+            raise RuntimeError("Cosmos T2W + DFD generators are not initialized.")
+
+        conditioning_image, uses_continuation = self._select_conditioning_image(
+            segment_idx,
+            image_path,
+            reset_conditioning,
+        )
+        conditioned = conditioning_image is not None
+        generator = self.continuation_generator if conditioned else self.bootstrap_generator
+        sampling_param = self._sampling_param(conditioned=conditioned)
+        started = time.perf_counter()
+        try:
+            if conditioned:
+                sampling_param.pil_image = conditioning_image
+            result = generator.generate_video(prompt, sampling_param=sampling_param)
+        finally:
+            if conditioning_image is not None:
+                conditioning_image.close()
+        torch.cuda.synchronize()
+        generation_ms = (time.perf_counter() - started) * 1000.0
+
+        if not isinstance(result, dict):
+            raise RuntimeError("Cosmos generation did not return one result dictionary.")
+        frames = result.get("frames")
+        expected_frames = int(sampling_param.num_frames)
+        if not isinstance(frames, list) or len(frames) != expected_frames:
+            actual_frames = len(frames) if isinstance(frames, list) else None
+            raise RuntimeError(f"Cosmos generation returned {actual_frames} frames; expected {expected_frames}.")
+
+        save_started = time.perf_counter()
+        self._save_continuation_frame(frames[-1])
+        save_conditioning_ms = (time.perf_counter() - save_started) * 1000.0
+        fps = int(sampling_param.fps)
+        timings = {
+            "generation_ms": generation_ms,
+            "generation_time_ms": float(result.get("generation_time") or 0.0) * 1000.0,
+            "save_conditioning_ms": save_conditioning_ms,
+            "e2e_latency_ms": (time.perf_counter() - started) * 1000.0,
+        }
+        trim_frames = 1 if uses_continuation else 0
+        mode = "DFD continuation" if conditioned else "T2W bootstrap"
+        print(f"[GPU {self.gpu_id}] Cosmos {mode} segment {segment_idx}: "
+              f"{len(frames)} frames, gen={generation_ms:.0f}ms, "
+              f"save_conditioning={save_conditioning_ms:.0f}ms, "
+              f"e2e={timings['e2e_latency_ms']:.0f}ms")
+        return StepResult(
+            frames=frames,
+            audio=self._silent_audio(len(frames), fps),
+            audio_sample_rate=_SILENT_AUDIO_SAMPLE_RATE,
+            timings=timings,
+            head_trim_frames=trim_frames,
+            head_trim_audio_frames=trim_frames,
+        )
+
+    def warmup(self, prompt: str) -> dict[str, float]:
+        """Exercise both T2W bootstrap and retained-frame DFD request shapes."""
+        warmup_prompt = (prompt or "").strip()
+        if not warmup_prompt:
+            raise RuntimeError("Startup warmup prompt must be non-empty.")
+        print(f"[GPU {self.gpu_id}] Cosmos startup warmup starting "
+              "(synthetic segments: T2W bootstrap, DFD continuation)")
+        started = time.perf_counter()
+        bootstrap_result = self.generate_step(warmup_prompt, 1, None, True)
+        continuation_result = self.generate_step(warmup_prompt, 2, None, False)
+        total_ms = (time.perf_counter() - started) * 1000.0
+        self.clear_conditioning()
+        bootstrap_ms = float(bootstrap_result.timings.get("e2e_latency_ms", 0.0))
+        continuation_ms = float(continuation_result.timings.get("e2e_latency_ms", 0.0))
+        print(f"[GPU {self.gpu_id}] Cosmos startup warmup complete: "
+              f"bootstrap={bootstrap_ms:.0f}ms, continuation={continuation_ms:.0f}ms, total={total_ms:.0f}ms")
+        return {
+            "warmup_bootstrap_ms": bootstrap_ms,
+            "warmup_continuation_ms": continuation_ms,
+            "warmup_total_ms": total_ms,
+        }
+
+    def apply_lora_stack(self, stack: list[tuple[str, float]]) -> tuple[str | None, str | None]:
+        del stack
+        raise RuntimeError("Cosmos Predict2.5 DFD does not support DreamVerse runtime LoRA changes.")

apps/dreamverse/dreamverse/generation_contracts.py

@@ -0,0 +1,49 @@
+"""Shared contract between DreamVerse generation backends and GPU workers."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any, Protocol
+
+from dreamverse.generation_inputs import GenerationInputs
+
+
+@dataclass
+class StepResult:
+    """Decoded media and stream-trimming metadata for one DreamVerse segment."""
+
+    frames: list
+    audio: Any
+    audio_sample_rate: int | None
+    timings: dict[str, float]
+    head_trim_frames: int
+    head_trim_audio_frames: int
+
+
+class GenerationBackend(Protocol):
+    """Model-owned generation operations used by one GPU worker process."""
+
+    def initialize(self, model_config: dict | None = None) -> None:
+        ...
+
+    def shutdown(self) -> None:
+        ...
+
+    def clear_conditioning(self) -> None:
+        ...
+
+    def generate_step(
+        self,
+        prompt: str,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+        generation_inputs: GenerationInputs | None = None,
+    ) -> StepResult:
+        ...
+
+    def warmup(self, prompt: str) -> dict[str, float]:
+        ...
+
+    def apply_lora_stack(self, stack: list[tuple[str, float]]) -> tuple[str | None, str | None]:
+        ...

apps/dreamverse/dreamverse/generation_inputs.py

@@ -0,0 +1,102 @@
+"""GPU-independent validation for generation modes and ordered asset handles."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from PIL import Image, UnidentifiedImageError
+
+from dreamverse.assets import asset_store
+
+GENERATION_MODES = ("t2va", "fl2va", "ref2va")
+
+
+@dataclass(frozen=True)
+class GenerationAsset:
+    asset_id: str
+    kind: str
+    path: str
+    role: str
+
+
+@dataclass(frozen=True)
+class GenerationInputs:
+    mode: str | None = None
+    assets: tuple[GenerationAsset, ...] = ()
+
+    @property
+    def first_frame_path(self) -> str | None:
+        return next((asset.path for asset in self.assets if asset.role == "first_frame"), None)
+
+    @property
+    def last_frame_path(self) -> str | None:
+        return next((asset.path for asset in self.assets if asset.role == "last_frame"), None)
+
+    @property
+    def references(self) -> tuple[GenerationAsset, ...]:
+        return tuple(asset for asset in self.assets if asset.role == "reference")
+
+
+def supported_generation_modes(model_id: str) -> tuple[str, ...]:
+    return GENERATION_MODES if model_id in ("full-h3", "mock") else ("t2va", )
+
+
+def resolve_generation_inputs(payload: dict, model_id: str) -> GenerationInputs:
+    mode = payload.get("generation_mode")
+    raw_assets = payload.get("conditioning_assets", [])
+    if mode is None and "generation_mode" not in payload:
+        if raw_assets:
+            raise ValueError("Select a generation mode before attaching conditioning assets.")
+        return GenerationInputs()
+    if not isinstance(mode, str) or mode not in GENERATION_MODES:
+        raise ValueError("Unknown generation mode. Choose T2VA, FL2VA, or Ref2VA.")
+    if mode not in supported_generation_modes(model_id):
+        raise ValueError(f"{mode.upper()} requires the Full H3 runtime. This runtime is running {model_id}.")
+    if payload.get("initial_image") is not None:
+        raise ValueError("Use asset IDs for generation modes; do not combine them with the legacy initial_image field.")
+    if not isinstance(raw_assets, list) or len(raw_assets) > 12:
+        raise ValueError("conditioning_assets must be an ordered list with at most 12 assets.")
+    if mode == "t2va" and raw_assets:
+        raise ValueError("T2VA accepts text only. Remove conditioning assets or choose another mode.")
+    assets: list[GenerationAsset] = []
+    for item in raw_assets:
+        if not isinstance(item, dict) or set(item) != {"asset_id", "role"}:
+            raise ValueError("Each conditioning asset must contain only asset_id and role.")
+        role = item["role"]
+        if role not in ("first_frame", "last_frame", "reference"):
+            raise ValueError("Asset role must be first_frame, last_frame, or reference.")
+        stored = asset_store.get(item["asset_id"])
+        assets.append(GenerationAsset(stored.asset_id, stored.kind, stored.path, role))
+    if mode == "fl2va":
+        if any(asset.kind != "image" or asset.role == "reference" for asset in assets):
+            raise ValueError("FL2VA accepts only first-frame and last-frame images.")
+        if sum(asset.role == "first_frame" for asset in assets) != 1:
+            raise ValueError("FL2VA requires exactly one first-frame image.")
+        if sum(asset.role == "last_frame" for asset in assets) > 1:
+            raise ValueError("FL2VA accepts at most one last-frame image.")
+    elif mode == "ref2va":
+        if not assets or any(asset.role != "reference" for asset in assets):
+            raise ValueError("Ref2VA requires an ordered list of reference assets, without keyframe roles.")
+        if not any(asset.kind in ("image", "video") for asset in assets):
+            raise ValueError("Ref2VA requires at least one image or video; audio alone is not supported.")
+        for kind, limit in (("image", 9), ("video", 3), ("audio", 3)):
+            if sum(asset.kind == kind for asset in assets) > limit:
+                raise ValueError(f"Ref2VA accepts at most {limit} {kind} references.")
+        for asset in assets:
+            if asset.kind == "image":
+                try:
+                    with Image.open(asset.path) as image:
+                        if image.width > 4 * image.height or image.height > 4 * image.width:
+                            raise ValueError(
+                                "Ref2VA image aspect ratios must be between 1:4 and 4:1. Crop this image first.")
+                except (UnidentifiedImageError, OSError, Image.DecompressionBombError) as exc:
+                    raise ValueError("A selected reference image could not be decoded. Upload it again.") from exc
+    return GenerationInputs(mode, tuple(assets))
+
+
+def pin_generation_inputs(inputs: GenerationInputs) -> None:
+    asset_store.pin([asset.asset_id for asset in inputs.assets])
+
+
+def release_generation_inputs(inputs: GenerationInputs) -> None:
+    asset_store.release([asset.asset_id for asset in inputs.assets])

apps/dreamverse/dreamverse/generation_worker.py

@@ -0,0 +1,103 @@
+"""Select and own one model-specific generation backend per GPU process."""
+
+from __future__ import annotations
+
+from dreamverse.config import MODEL_CONFIG
+from dreamverse.generation_contracts import GenerationBackend, StepResult
+from dreamverse.generation_inputs import GenerationInputs
+
+
+def _create_generation_backend(backend_name: str, gpu_id: int) -> GenerationBackend:
+    """Construct the backend that owns the selected model family's behavior."""
+    if backend_name == "ltx2":
+        from dreamverse.ltx2_generation import LTX2GenerationBackend
+
+        return LTX2GenerationBackend(gpu_id)
+    if backend_name == "minimax_h3":
+        from dreamverse.minimax_h3_generation import MiniMaxH3GenerationBackend
+
+        return MiniMaxH3GenerationBackend(gpu_id)
+    if backend_name == "cosmos25_dfd":
+        from dreamverse.cosmos25_dfd_generation import Cosmos25DFDGenerationBackend
+
+        return Cosmos25DFDGenerationBackend(gpu_id)
+    raise ValueError(f"Unsupported DreamVerse generation backend: {backend_name!r}")
+
+
+class VideoGenerationWorker:
+    """Delegate GPU lifecycle and generation calls to the active model backend."""
+
+    def __init__(self, gpu_id: int):
+        self.gpu_id = gpu_id
+        self.model_config: dict = dict(MODEL_CONFIG)
+        self.backend_name: str | None = None
+        self.backend: GenerationBackend | None = None
+
+    def initialize(self, model_config: dict | None = None) -> None:
+        """Load the requested model through its generation backend.
+
+        Model selection belongs here so the GPU process and streaming layers
+        use one stable media contract without importing model-specific code.
+        """
+        requested_model_config = dict(model_config) if model_config is not None else dict(self.model_config)
+        backend_name = requested_model_config.get("generation_backend")
+        if not isinstance(backend_name, str) or not backend_name:
+            raise ValueError("DreamVerse model configuration requires `generation_backend`.")
+
+        candidate_backend = self.backend
+        if candidate_backend is None or self.backend_name != backend_name:
+            if candidate_backend is not None:
+                candidate_backend.shutdown()
+            candidate_backend = _create_generation_backend(backend_name, self.gpu_id)
+
+        try:
+            candidate_backend.initialize(requested_model_config)
+        except Exception:
+            try:
+                candidate_backend.shutdown()
+            except Exception as shutdown_error:
+                print(f"[GPU {self.gpu_id}] Backend cleanup after initialization failure: {shutdown_error}")
+            self.backend = None
+            self.backend_name = None
+            raise
+
+        self.model_config = requested_model_config
+        self.backend = candidate_backend
+        self.backend_name = backend_name
+
+    def _require_backend(self) -> GenerationBackend:
+        """Return the initialized backend or fail before processing a command."""
+        if self.backend is None:
+            raise RuntimeError("Generation backend is not initialized.")
+        return self.backend
+
+    def shutdown(self) -> None:
+        """Release model resources owned by the selected backend."""
+        if self.backend is not None:
+            self.backend.shutdown()
+
+    def clear_conditioning(self) -> None:
+        self._require_backend().clear_conditioning()
+
+    def generate_step(
+        self,
+        prompt: str,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+        generation_inputs: GenerationInputs | None = None,
+    ) -> StepResult:
+        """Generate one segment through the selected model backend."""
+        return self._require_backend().generate_step(
+            prompt,
+            segment_idx,
+            image_path,
+            reset_conditioning,
+            generation_inputs=generation_inputs,
+        )
+
+    def warmup(self, prompt: str) -> dict[str, float]:
+        return self._require_backend().warmup(prompt)
+
+    def apply_lora_stack(self, stack: list[tuple[str, float]]) -> tuple[str | None, str | None]:
+        return self._require_backend().apply_lora_stack(stack)

apps/dreamverse/dreamverse/gpu_pool.py

@@ -12,7 +12,7 @@ from enum import Enum
 from multiprocessing import Process, Queue
 
 from dreamverse.config import (
-    DEFAULT_MODEL_ID,
+    ACTIVE_MODEL_ID,
     DREAMVERSE_SP_SIZE,
     MODEL_REGISTRY,
     STARTUP_WARMUP_ENABLED,
@@ -29,6 +29,7 @@ from dreamverse.av_streaming import (
     generate_stream_id,
     stream_fmp4,
 )
+from dreamverse.generation_inputs import GenerationInputs, pin_generation_inputs, release_generation_inputs
 from dreamverse.worker_ipc import (
     CommandPayload,
     InitAck,
@@ -54,7 +55,7 @@ from dreamverse.worker_ipc import (
 def _parse_requested_gpu_limit() -> int | None:
     raw_value = os.getenv("FASTVIDEO_GPU_COUNT", "").strip().lower()
     if not raw_value:
-        return 1
+        return DREAMVERSE_SP_SIZE
     if raw_value == "all":
         return None
     try:
@@ -164,12 +165,12 @@ def gpu_worker_process(
     os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device
     os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "FLASH_ATTN"
 
-    from dreamverse.video_generation import VideoGenerationWorker
+    from dreamverse.generation_worker import VideoGenerationWorker
 
     worker = VideoGenerationWorker(gpu_id)
 
     def event_loop(first_cmd: Command = None):
-        """Blocking event loop for LTX2; dispatches user commands."""
+        """Block on generation commands after the model is initialized."""
         print(f"[GPU {gpu_id}] Entering event loop")
 
         def handle_command(cmd: Command):
@@ -189,6 +190,7 @@ def gpu_worker_process(
                         segment_idx,
                         image_path=payload.image_path,
                         reset_conditioning=payload.reset_conditioning,
+                        generation_inputs=payload.generation_inputs,
                     )
                     head_trim_frames = step_result.head_trim_frames
                     head_trim_audio_frames = step_result.head_trim_audio_frames
@@ -432,10 +434,11 @@ class GPUSlot:
         self.connected_users: set[str] = set()
         self._pending_futures: dict[str, asyncio.Future] = {}
         self._stream_queues: dict[str, asyncio.Queue] = {}
+        self._step_asset_inputs: dict[str, GenerationInputs] = {}
         self._response_reader_task: asyncio.Task | None = None
         self._active: bool = False
         self._reader_lock: asyncio.Lock | None = None
-        self.current_model_id: str = DEFAULT_MODEL_ID
+        self.current_model_id: str | None = ACTIVE_MODEL_ID
         self.shared_stream_buffer = None
         self.shared_stream_buffer_size = SHARED_STREAM_BUFFER_BYTES
 
@@ -663,6 +666,9 @@ class GPUSlot:
                 if isinstance(event, (StepComplete, WarmupComplete)):
                     event.timings["ipc_get_done_ns"] = time.time_ns()
 
+                if isinstance(event, (StepComplete, WorkerError)) and event.user_id is not None:
+                    self._release_step_assets(event.user_id)
+
                 user_id = event.user_id
                 if user_id and user_id in self._pending_futures:
                     future = self._pending_futures.pop(user_id)
@@ -690,7 +696,7 @@ class GPUSlot:
     async def join_user(self, user_id: str, model_id: str = None) -> JoinAck:
         """Add a user to this GPU."""
         if model_id is None:
-            model_id = DEFAULT_MODEL_ID
+            model_id = ACTIVE_MODEL_ID
 
         # Reload model if a different one is requested
         if model_id != self.current_model_id and model_id in MODEL_REGISTRY:
@@ -705,16 +711,23 @@ class GPUSlot:
             self.connected_users.clear()
 
             model_config = MODEL_REGISTRY[model_id]
-            reload_response = await self._send_command(Command(CommandType.RELOAD_MODEL,
-                                                               payload=ReloadModelPayload(model_config=model_config),
-                                                               user_id="__reload__"),
-                                                       timeout=600.0)
+            try:
+                reload_response = await self._send_command(Command(
+                    CommandType.RELOAD_MODEL,
+                    payload=ReloadModelPayload(model_config=model_config),
+                    user_id="__reload__"),
+                                                           timeout=600.0)
+            except Exception:
+                self.current_model_id = None
+                raise
             match reload_response:
                 case ReloadAck():
                     pass
                 case WorkerError(message=msg):
+                    self.current_model_id = None
                     raise RuntimeError(f"Model reload failed: {msg}")
                 case _:
+                    self.current_model_id = None
                     raise RuntimeError(f"Unexpected reload response: "
                                        f"{type(reload_response).__name__}")
 
@@ -746,6 +759,7 @@ class GPUSlot:
         segment_idx: int = 1,
         image_path: str | None = None,
         reset_conditioning: bool = False,
+        generation_inputs: GenerationInputs | None = None,
     ) -> dict[str, float]:
         """Execute a generation step for a specific user.
 
@@ -759,9 +773,19 @@ class GPUSlot:
             segment_idx=segment_idx,
             image_path=image_path,
             reset_conditioning=bool(reset_conditioning),
+            generation_inputs=generation_inputs,
         )
+        if generation_inputs is not None:
+            if user_id in self._step_asset_inputs:
+                raise RuntimeError("The previous generation is still using this project's assets.")
+            pin_generation_inputs(generation_inputs)
+            self._step_asset_inputs[user_id] = generation_inputs
+        # Pins intentionally survive a waiter timeout/cancellation: the GPU
+        # command keeps running. The response reader releases them when the
+        # worker actually completes (even if that response is now unmatched).
         response = await self._send_command_tagged(Command(CommandType.USER_STEP, payload=payload, user_id=user_id),
                                                    timeout=1800.0)
+        self._release_step_assets(user_id)
         match response:
             case StepComplete(timings=timings):
                 return timings
@@ -771,6 +795,11 @@ class GPUSlot:
                 raise RuntimeError(f"Unexpected step response for {user_id[:8]}: "
                                    f"{type(response).__name__}")
 
+    def _release_step_assets(self, user_id: str) -> None:
+        inputs = self._step_asset_inputs.pop(user_id, None)
+        if inputs is not None:
+            release_generation_inputs(inputs)
+
     async def apply_lora_stack(
         self,
         stack: list[tuple[str, float]],
@@ -793,7 +822,9 @@ class GPUSlot:
     async def leave_user(self, user_id: str) -> None:
         """Remove a user from this GPU."""
         try:
-            await self._send_command_tagged(Command(CommandType.USER_LEAVE, user_id=user_id), timeout=30.0)
+            response = await self._send_command_tagged(Command(CommandType.USER_LEAVE, user_id=user_id), timeout=30.0)
+            if isinstance(response, LeaveAck):
+                self._release_step_assets(user_id)
         except Exception as e:
             print(f"[GPU {self.gpu_id}] Leave user error: {e}")
         finally:
@@ -830,6 +861,10 @@ class GPUSlot:
                 except Exception:
                     pass
 
+        if self.process is None or not self.process.is_alive():
+            for user_id in list(self._step_asset_inputs):
+                self._release_step_assets(user_id)
+
         for q in (self.command_queue, self.response_queue):
             if q is not None:
                 try:

apps/dreamverse/dreamverse/ltx2_generation.py

@@ -1,9 +1,9 @@
-"""LTX2 model lifecycle and continuation conditioning.
+"""LTX-2 model lifecycle and continuation conditioning.
 
 Runs inside a GPU worker subprocess.  Owns the model, the audio
 encoder, and the per-session continuation state carried across
 segments.  Callers must set ``os.environ["CUDA_VISIBLE_DEVICES"]``
-before constructing ``VideoGenerationWorker`` — all ``fastvideo.*``
+before constructing ``LTX2GenerationBackend`` — all ``fastvideo.*``
 imports are deferred to method bodies so nothing touches CUDA at
 module import time.
 """
@@ -14,9 +14,6 @@ import gc
 import os
 import re
 import time
-from dataclasses import dataclass
-from typing import Any
-
 import numpy as np
 import torch
 
@@ -35,6 +32,8 @@ from dreamverse.config import (
     DREAMVERSE_LORA_STACK,
     _resolve_lora_spec,
 )
+from dreamverse.generation_contracts import StepResult
+from dreamverse.generation_inputs import GenerationInputs
 
 # Multi-frame decoded continuation defaults from
 # examples/inference/basic/basic_ltx2_distilled_video_continuation.py.
@@ -80,22 +79,6 @@ def _reset_lora_registry(worker) -> dict:
     return {"status": "lora_registry_reset"}
 
 
-@dataclass
-class StepResult:
-    """Output of one generation step.
-
-    ``head_trim_frames`` / ``head_trim_audio_frames`` are derived here
-    so downstream AV streaming never needs to import conditioning
-    constants.
-    """
-    frames: list
-    audio: Any
-    audio_sample_rate: int | None
-    timings: dict
-    head_trim_frames: int
-    head_trim_audio_frames: int
-
-
 class ContinuationState:
     """Per-session video + audio conditioning carried across segments."""
 
@@ -202,7 +185,7 @@ class ContinuationState:
         self.audio_latents = latents.detach().clone().cpu()
 
 
-class VideoGenerationWorker:
+class LTX2GenerationBackend:
     """Single-GPU LTX2 generator with continuation state.
 
     Caller must set ``os.environ["CUDA_VISIBLE_DEVICES"]`` before
@@ -308,7 +291,13 @@ class VideoGenerationWorker:
                     dynamic=False,
                 ),
                 use_fsdp_inference=False,
-                quantization=QuantizationConfig(transformer_quant="NVFP4"),
+                # The bundled LTX2 model enables a refinement LoRA during the
+                # first request. NVFP4 otherwise purges the dense weights that
+                # FastVideo's LoRA merge path requires.
+                quantization=QuantizationConfig(
+                    transformer_quant="NVFP4",
+                    transformer_retain_original_weights=True,
+                ),
             ),
             pipeline=PipelineSelection(
                 components=components,
@@ -472,8 +461,11 @@ class VideoGenerationWorker:
         segment_idx: int,
         image_path: str | None,
         reset_conditioning: bool,
+        generation_inputs: GenerationInputs | None = None,
     ) -> StepResult:
         """Execute one generation step; snapshot state for the next segment."""
+        if generation_inputs is not None and (generation_inputs.mode not in (None, "t2va") or generation_inputs.assets):
+            raise ValueError("LTX supports text generation only through the generation mode API.")
         timings: dict = {}
 
         prompt = self._inject_style_trigger(prompt)

apps/dreamverse/dreamverse/main.py

@@ -15,6 +15,7 @@ from dreamverse.gpu_pool import GPUPool, get_available_gpus
 from dreamverse.session_logger import SessionEventLogger
 
 from dreamverse.config import (
+    ACTIVE_MODEL_ID,
     AVAILABLE_LORAS,
     DEVTOOLS_ENABLED,
     FRONTEND_STATIC_DIR_CANDIDATES,
@@ -34,6 +35,8 @@ from dreamverse.routes.presets import (
     curated_presets_router,
 )
 from dreamverse.session.controller import SessionController
+from dreamverse.generation_inputs import supported_generation_modes
+from dreamverse.routes.assets import router as asset_router
 
 
 class _HeartbeatAccessLogFilter(logging.Filter):
@@ -92,10 +95,16 @@ app.add_middleware(
 app.include_router(build_health_router(lambda: runtime.gpu_pool))
 app.include_router(internal_monitor_router)
 app.include_router(prompt_config_router)
+app.include_router(asset_router)
 if DEVTOOLS_ENABLED:
     app.include_router(curated_presets_router)
 
 
+@app.get("/generation-capabilities")
+async def generation_capabilities() -> dict:
+    return {"model_id": ACTIVE_MODEL_ID, "modes": supported_generation_modes(ACTIVE_MODEL_ID), "mock": False}
+
+
 @app.websocket("/ws")
 async def websocket_endpoint(websocket: WebSocket):
     controller = SessionController(

apps/dreamverse/dreamverse/minimax_h3_generation.py

@@ -0,0 +1,363 @@
+"""Full/Preview H3 lifecycle, conditioning and per-project pipeline selection."""
+
+from __future__ import annotations
+
+import gc
+import os
+import time
+from typing import TYPE_CHECKING, Any
+
+import numpy as np
+import torch
+
+from dreamverse.config import DREAMVERSE_SP_SIZE
+from dreamverse.generation_contracts import StepResult
+from dreamverse.generation_inputs import GenerationInputs
+
+if TYPE_CHECKING:
+    from PIL.Image import Image
+
+
+def _required_config_str(model_config: dict, field_name: str) -> str:
+    """Read one required non-empty string from a DreamVerse model profile."""
+    value = model_config.get(field_name)
+    if not isinstance(value, str) or not value.strip():
+        raise ValueError(f"FastH3 model configuration requires `{field_name}`.")
+    return value.strip()
+
+
+class MiniMaxH3GenerationBackend:
+    """Own one H3 pipeline at a time and retain base-pipeline continuation."""
+
+    def __init__(self, gpu_id: int):
+        self.gpu_id = gpu_id
+        self.generator: Any | None = None
+        self.model_config: dict = {}
+        self.continuation_image: Image | None = None
+        self.pipeline_mode = "base"
+
+    def _gpu_mem(self) -> str:
+        allocated_gib = torch.cuda.memory_allocated() / 1024**3
+        reserved_gib = torch.cuda.memory_reserved() / 1024**3
+        return f"alloc={allocated_gib:.2f}GiB, reserved={reserved_gib:.2f}GiB"
+
+    @staticmethod
+    def _configure_environment(attention_backend: str) -> None:
+        """Apply the fixed boot-time switches from the FastH3 reference recipe."""
+        os.environ.update({
+            "FASTVIDEO_ATTENTION_BACKEND": attention_backend,
+            "FASTVIDEO_FA4": "1",
+            "FASTVIDEO_MINIMAX_H3_FUSIONS": "all",
+            "FASTVIDEO_VSA_SM100A": "0",
+        })
+        os.environ.pop("FASTVIDEO_INFERENCE_TORCH_COMPILE", None)
+
+    def initialize(self, model_config: dict | None = None) -> None:
+        """Load the profile's base pipeline; Ref2VA is loaded on first use."""
+        if model_config is not None:
+            self.model_config = dict(model_config)
+        if not self.model_config:
+            raise ValueError("FastH3 initialization requires a model configuration.")
+        self._load_pipeline("base")
+
+    def _load_pipeline(self, pipeline_mode: str) -> None:
+        """Unload the old executor before loading a base or reference transformer.
+
+        GPU worker commands are serialized, so a project boundary never swaps
+        weights while another request is using them. Keeping one executor also
+        avoids simultaneously retaining two large H3 transformers in VRAM. A
+        failed load leaves no executor behind so the next step retries it.
+        """
+        full_checkpoint = bool(self.model_config.get("full_checkpoint", False))
+        if pipeline_mode == "ref2va" and not full_checkpoint:
+            raise ValueError("Ref2VA requires the full-h3 model profile.")
+        if self.generator is not None:
+            previous_generator = self.generator
+            self.generator = None
+            previous_generator.shutdown()
+            del previous_generator
+            gc.collect()
+            torch.cuda.empty_cache()
+
+        self.clear_conditioning()
+        model_path = _required_config_str(self.model_config, "model_path")
+        attention_backend = _required_config_str(self.model_config, "attention_backend")
+        self._configure_environment(attention_backend)
+
+        from fastvideo import VideoGenerator
+        from fastvideo.api import (
+            CompileConfig,
+            ComponentConfig,
+            EngineConfig,
+            GeneratorConfig,
+            OffloadConfig,
+            ParallelismConfig,
+            PipelineSelection,
+        )
+
+        components = ComponentConfig()
+        if not full_checkpoint:
+            from huggingface_hub import hf_hub_download
+
+            adapter_repo = _required_config_str(self.model_config, "adapter_repo")
+            adapter_filename = _required_config_str(self.model_config, "adapter_filename")
+            components.lora_path = hf_hub_download(repo_id=adapter_repo, filename=adapter_filename)
+            components.lora_strength = 1.0
+            print(f"[GPU {self.gpu_id}] FastH3 adapter: {adapter_repo}/{adapter_filename}")
+        if pipeline_mode == "ref2va":
+            components.override_pipeline_cls_name = "MiniMaxH3Ref2VAModularPipeline"
+        experimental = {
+            "attention_backend": attention_backend,
+            "inference_torch_compile": not full_checkpoint and attention_backend == "FLASH_ATTN",
+            "vae_parallel_decode": True,
+            "vae_parallel_decode_strategy": "gather",
+        }
+        if attention_backend == "VIDEO_SPARSE_ATTN_H3":
+            experimental.update({
+                "VSA_sparsity": 0.9,
+                "VSA_tile_size": 64,
+            })
+        generator_config = GeneratorConfig(
+            model_path=model_path,
+            pipeline=PipelineSelection(
+                workload_type="i2v" if pipeline_mode == "ref2va" else None,
+                components=components,
+                experimental=experimental,
+            ),
+            engine=EngineConfig(
+                num_gpus=DREAMVERSE_SP_SIZE,
+                parallelism=ParallelismConfig(tp_size=1, sp_size=DREAMVERSE_SP_SIZE),
+                offload=OffloadConfig(
+                    dit=False,
+                    dit_layerwise=False,
+                    text_encoder=True,
+                    image_encoder=True,
+                    vae=True,
+                    pin_cpu_memory=not full_checkpoint,
+                ),
+                compile=CompileConfig(enabled=False, vae_enabled=True),
+                use_fsdp_inference=full_checkpoint and DREAMVERSE_SP_SIZE > 1,
+            ),
+        )
+
+        print(f"[GPU {self.gpu_id}] Loading H3 model: {model_path} ({pipeline_mode})")
+        print(f"[GPU {self.gpu_id}] Before model load: {self._gpu_mem()}")
+        try:
+            self.generator = VideoGenerator.from_config(generator_config)
+        except Exception:
+            # The old executor is already gone; leaving no executor behind lets
+            # the next step retry this load instead of stranding the GPU slot.
+            self.generator = None
+            raise
+        self.pipeline_mode = pipeline_mode
+        print(f"[GPU {self.gpu_id}] FastH3 loaded: {self._gpu_mem()} (warmup pending)")
+
+    def shutdown(self) -> None:
+        """Release the FastVideo generator and cached continuation image."""
+        self.clear_conditioning()
+        if self.generator is not None:
+            self.generator.shutdown()
+            self.generator = None
+
+    def clear_conditioning(self) -> None:
+        """Release the first-frame image retained for the next segment."""
+        if self.continuation_image is not None:
+            self.continuation_image.close()
+            self.continuation_image = None
+
+    @staticmethod
+    def _load_rgb_image(image_path: str) -> Image:
+        """Load an image into an independent RGB buffer with no open file handle."""
+        from PIL import Image
+
+        with Image.open(image_path) as image:
+            return image.convert("RGB").copy()
+
+    def _select_conditioning_image(
+        self,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+    ) -> tuple[Image | None, bool]:
+        """Select the initial upload or retained last frame for one segment."""
+        if reset_conditioning:
+            self.clear_conditioning()
+        if segment_idx > 1 and self.continuation_image is not None:
+            return self.continuation_image.copy(), True
+        if segment_idx > 1 and not reset_conditioning:
+            raise RuntimeError(f"FastH3 segment {segment_idx} requires a retained continuation frame.")
+        if segment_idx == 1 and image_path:
+            return self._load_rgb_image(image_path), False
+        return None, False
+
+    def _build_request(
+        self,
+        prompt: str,
+        conditioning_image: Image | None,
+        last_image: Image | None = None,
+        generation_inputs: GenerationInputs | None = None,
+    ):
+        """Build the typed FastVideo request owned by the FastH3 profile."""
+        from fastvideo.api import GenerationRequest, InputConfig, OutputConfig, SamplingConfig
+
+        references = None
+        if generation_inputs is not None and generation_inputs.mode == "ref2va":
+            from fastvideo.api import MiniMaxH3Reference
+
+            references = [
+                MiniMaxH3Reference(source=str(asset.path), media_type=asset.kind)
+                for asset in generation_inputs.references
+            ]
+        return GenerationRequest(
+            prompt=prompt,
+            negative_prompt="",
+            inputs=InputConfig(pil_image=conditioning_image, last_image=last_image, references=references),
+            sampling=SamplingConfig(
+                height=int(self.model_config["height"]),
+                width=int(self.model_config["width"]),
+                num_frames=int(self.model_config["num_frames"]),
+                fps=24,
+                num_inference_steps=int(self.model_config["num_inference_steps"]),
+                guidance_scale=1.0,
+                batch_cfg=False,
+                seed=int(self.model_config["seed"]),
+            ),
+            output=OutputConfig(save_video=False, return_frames=True),
+        )
+
+    def _save_continuation_frame(self, frames: list) -> None:
+        """Retain the last decoded frame as first-frame conditioning."""
+        from PIL import Image
+
+        self.clear_conditioning()
+        self.continuation_image = Image.fromarray(np.ascontiguousarray(frames[-1])).convert("RGB")
+
+    def generate_step(
+        self,
+        prompt: str,
+        segment_idx: int,
+        image_path: str | None,
+        reset_conditioning: bool,
+        generation_inputs: GenerationInputs | None = None,
+    ) -> StepResult:
+        """Generate one synchronized FastH3 segment and retain its last frame.
+
+        Later segments use MiniMax H3's first-frame-to-video path. The first
+        conditioned frame and its matching audio duration are trimmed before
+        streaming so adjacent segments do not duplicate media.
+        """
+        mode = generation_inputs.mode if generation_inputs is not None else None
+        if mode not in (None, "t2va", "fl2va", "ref2va"):
+            raise ValueError(f"Unsupported H3 generation mode: {mode!r}.")
+        if mode in ("fl2va", "ref2va") and not self.model_config.get("full_checkpoint", False):
+            raise ValueError(f"{mode.upper()} requires the full-h3 model profile.")
+        pipeline_mode = "ref2va" if mode == "ref2va" else "base"
+        if self.generator is None or self.pipeline_mode != pipeline_mode:
+            # A failed switch leaves no executor behind; reload here so the
+            # slot recovers on the next step instead of staying broken.
+            if segment_idx > 1 and not reset_conditioning:
+                raise ValueError("Generation mode cannot change in the middle of a project.")
+            self._load_pipeline(pipeline_mode)
+
+        conditioning_image = None
+        last_image = None
+        uses_continuation = False
+        if mode == "ref2va":
+            # The reference pipeline rejects first/last-frame inputs. Preserve
+            # all original references for every clip and do not trim overlap.
+            self.clear_conditioning()
+        else:
+            if mode == "fl2va" and generation_inputs is not None:
+                image_path = generation_inputs.first_frame_path
+            conditioning_image, uses_continuation = self._select_conditioning_image(
+                segment_idx,
+                image_path,
+                reset_conditioning,
+            )
+        started = time.perf_counter()
+        try:
+            if (mode == "fl2va" and segment_idx == 1 and generation_inputs is not None
+                    and generation_inputs.last_frame_path):
+                last_image = self._load_rgb_image(generation_inputs.last_frame_path)
+            request = self._build_request(prompt, conditioning_image, last_image, generation_inputs)
+            result = self.generator.generate(request)
+        finally:
+            if conditioning_image is not None:
+                conditioning_image.close()
+            if last_image is not None:
+                last_image.close()
+        torch.cuda.synchronize()
+        generation_ms = (time.perf_counter() - started) * 1000.0
+
+        if isinstance(result, list):
+            raise RuntimeError("FastH3 returned multiple results for one DreamVerse segment.")
+        frames = result.frames
+        if not isinstance(frames, list) or not frames:
+            raise RuntimeError("FastH3 generation did not return decoded frames.")
+        audio = result.audio
+        audio_sample_rate = result.audio_sample_rate
+        if audio is not None and audio_sample_rate is None:
+            raise RuntimeError("FastH3 returned audio without an audio sample rate.")
+
+        save_started = time.perf_counter()
+        if mode != "ref2va":
+            self._save_continuation_frame(frames)
+        save_conditioning_ms = (time.perf_counter() - save_started) * 1000.0
+        timings = {
+            "generation_ms": generation_ms,
+            "generation_time_ms": float(result.generation_time or 0.0) * 1000.0,
+            "save_conditioning_ms": save_conditioning_ms,
+            "e2e_latency_ms": (time.perf_counter() - started) * 1000.0,
+        }
+        trim_frames = 1 if uses_continuation else 0
+        print(f"[GPU {self.gpu_id}] FastH3 segment {segment_idx}: "
+              f"{len(frames)} frames, gen={generation_ms:.0f}ms, "
+              f"save_conditioning={save_conditioning_ms:.0f}ms, "
+              f"e2e={timings['e2e_latency_ms']:.0f}ms")
+        return StepResult(
+            frames=frames,
+            audio=audio,
+            audio_sample_rate=audio_sample_rate,
+            timings=timings,
+            head_trim_frames=trim_frames,
+            head_trim_audio_frames=trim_frames,
+        )
+
+    def warmup(self, prompt: str) -> dict[str, float]:
+        """Compile the FastH3 text and first-frame paths before readiness."""
+        warmup_prompt = (prompt or "").strip()
+        if not warmup_prompt:
+            raise RuntimeError("Startup warmup prompt must be non-empty.")
+        print(f"[GPU {self.gpu_id}] FastH3 startup warmup starting "
+              "(synthetic segments: text-to-video, first-frame-to-video)")
+        started = time.perf_counter()
+        text_result = self.generate_step(
+            warmup_prompt,
+            segment_idx=1,
+            image_path=None,
+            reset_conditioning=True,
+        )
+        first_frame_result = self.generate_step(
+            warmup_prompt,
+            segment_idx=2,
+            image_path=None,
+            reset_conditioning=False,
+        )
+        total_ms = (time.perf_counter() - started) * 1000.0
+        self.clear_conditioning()
+        text_ms = float(text_result.timings.get("e2e_latency_ms", 0.0))
+        first_frame_ms = float(first_frame_result.timings.get("e2e_latency_ms", 0.0))
+        print(f"[GPU {self.gpu_id}] FastH3 startup warmup complete: "
+              f"text_to_video={text_ms:.0f}ms, "
+              f"first_frame_to_video={first_frame_ms:.0f}ms, "
+              f"total={total_ms:.0f}ms")
+        return {
+            "warmup_text_to_video_ms": text_ms,
+            "warmup_first_frame_to_video_ms": first_frame_ms,
+            "warmup_total_ms": total_ms,
+        }
+
+    def apply_lora_stack(self, stack: list[tuple[str, float]]) -> tuple[str | None, str | None]:
+        """Reject runtime LoRA mutation because FastH3 uses one startup adapter."""
+        del stack
+        raise RuntimeError("FastH3 uses its fixed startup adapter and does not support runtime LoRA changes.")

apps/dreamverse/dreamverse/mock_server.py

@@ -32,6 +32,14 @@ from fastapi.staticfiles import StaticFiles
 from dreamverse._deps import require_dreamverse_runtime_deps
 from dreamverse.config import FRONTEND_STATIC_DIR_CANDIDATES, GENERATION_SEGMENT_CAP
 from dreamverse.session_init_image import cleanup_session_init_image, persist_session_init_image
+from dreamverse.generation_inputs import (
+    GenerationInputs,
+    pin_generation_inputs,
+    release_generation_inputs,
+    resolve_generation_inputs,
+    supported_generation_modes,
+)
+from dreamverse.routes.assets import router as asset_router
 
 LATENCY_MS = 200
 SESSION_TIMEOUT_SECONDS = 300
@@ -170,6 +178,12 @@ app.add_middleware(
     allow_methods=["*"],
     allow_headers=["*"],
 )
+app.include_router(asset_router)
+
+
+@app.get("/generation-capabilities")
+async def generation_capabilities():
+    return {"model_id": "mock", "modes": supported_generation_modes("mock"), "mock": True}
 
 
 @app.get("/healthz")
@@ -290,6 +304,7 @@ async def websocket_endpoint(websocket: WebSocket):
     send_lock = asyncio.Lock()
     stop_event = asyncio.Event()
     session_init_image = None
+    generation_inputs = GenerationInputs()
 
     async def ws_send_json(payload: dict) -> None:
         async with send_lock:
@@ -347,10 +362,13 @@ async def websocket_endpoint(websocket: WebSocket):
         generation_paused = bool(initial_rollout_prompt and not single_clip_mode and len(curated_prompts) == 0)
 
         try:
+            generation_inputs = resolve_generation_inputs(init_data, "mock")
+            pin_generation_inputs(generation_inputs)
             session_init_image = persist_session_init_image(init_data.get("initial_image"))
         except ValueError as exc:
             await ws_send_json({
                 "type": "error",
+                "error_code": "invalid_generation_input",
                 "message": str(exc),
             })
             await websocket.close(code=1003, reason="Invalid initial image")
@@ -362,6 +380,8 @@ async def websocket_endpoint(websocket: WebSocket):
             "type": "gpu_assigned",
             "gpu_id": 0,
             "session_timeout": SESSION_TIMEOUT_SECONDS,
+            "generation_mode": generation_inputs.mode,
+            "mock": True,
         })
 
         raw_prompt_queue: asyncio.Queue[PromptSubmission] = asyncio.Queue()
@@ -486,6 +506,7 @@ async def websocket_endpoint(websocket: WebSocket):
                 })
 
         async def apply_project_init_payload(payload: dict[str, object], ) -> bool:
+            nonlocal generation_inputs
             nonlocal preset_id
             nonlocal preset_label
             nonlocal initial_rollout_prompt
@@ -519,10 +540,19 @@ async def websocket_endpoint(websocket: WebSocket):
             ]
 
             try:
-                replace_session_image(payload.get("initial_image"))
+                next_inputs = resolve_generation_inputs(payload, "mock")
+                pin_generation_inputs(next_inputs)
+                try:
+                    replace_session_image(payload.get("initial_image"))
+                except ValueError:
+                    release_generation_inputs(next_inputs)
+                    raise
+                release_generation_inputs(generation_inputs)
+                generation_inputs = next_inputs
             except ValueError as exc:
                 await ws_send_json({
                     "type": "error",
+                    "error_code": "invalid_generation_input",
                     "message": str(exc),
                 })
                 return False
@@ -568,6 +598,7 @@ async def websocket_endpoint(websocket: WebSocket):
             return drained
 
         async def enter_project_idle() -> None:
+            nonlocal generation_inputs
             nonlocal seed_prompt_memory
             nonlocal curated_prompts
             nonlocal curated_idx
@@ -587,6 +618,8 @@ async def websocket_endpoint(websocket: WebSocket):
 
             dropped_raw = drain_queue_nowait(raw_prompt_queue)
             dropped_ready = drain_queue_nowait(ready_prompt_queue)
+            release_generation_inputs(generation_inputs)
+            generation_inputs = GenerationInputs()
             seed_prompt_memory = []
             curated_prompts = []
             curated_idx = 0
@@ -767,10 +800,16 @@ async def websocket_endpoint(websocket: WebSocket):
                         continue
 
                     try:
+                        if generation_inputs.mode is not None and data.get("initial_image") is not None:
+                            raise ValueError("Choose conditioning assets when starting a project; legacy initial_image "
+                                             "cannot replace generation mode inputs.")
+                        if "generation_mode" in data or "conditioning_assets" in data:
+                            raise ValueError("simple_generate cannot change the mode; use project_init_v1.")
                         replace_session_image(data.get("initial_image"))
                     except ValueError as exc:
                         await ws_send_json({
                             "type": "error",
+                            "error_code": "invalid_generation_input",
                             "message": str(exc),
                         })
                         continue
@@ -1182,6 +1221,7 @@ async def websocket_endpoint(websocket: WebSocket):
     finally:
         stop_event.set()
         cleanup_session_init_image(session_init_image)
+        release_generation_inputs(generation_inputs)
 
 
 for static_dir in FRONTEND_STATIC_DIR_CANDIDATES:

apps/dreamverse/dreamverse/routes/assets.py

@@ -0,0 +1,72 @@
+"""Raw, bounded media uploads keep large binary data out of websocket messages."""
+
+from __future__ import annotations
+
+import asyncio
+from urllib.parse import unquote
+
+from fastapi import APIRouter, HTTPException, Request, Response
+from fastapi.responses import FileResponse
+from starlette.concurrency import run_in_threadpool
+
+from dreamverse.assets import IMAGE_LIMIT, MEDIA_LIMIT, MIME_TYPES, asset_store
+
+router = APIRouter()
+_upload_lock = asyncio.Lock()
+
+
+@router.post("/assets", status_code=201)
+async def upload_asset(request: Request) -> dict:
+    mime_type = request.headers.get("content-type", "").split(";", 1)[0].lower()
+    if mime_type not in MIME_TYPES:
+        raise HTTPException(415, "Unsupported asset type. Select a supported image, video, or audio file.")
+    limit = IMAGE_LIMIT if MIME_TYPES[mime_type][0] == "image" else MEDIA_LIMIT
+    try:
+        if int(request.headers.get("content-length", "0")) > limit:
+            raise HTTPException(413, f"Asset exceeds the {limit // (1024 * 1024)} MB upload limit.")
+    except ValueError as exc:
+        raise HTTPException(400, "Invalid Content-Length.") from exc
+    async with _upload_lock:
+        try:
+            path = asset_store.staging_path(mime_type)
+        except ValueError as exc:
+            raise HTTPException(400, str(exc)) from exc
+        try:
+            size = 0
+            with path.open("xb") as handle:
+                async for chunk in request.stream():
+                    size += len(chunk)
+                    if size > limit:
+                        raise HTTPException(413, f"Asset exceeds the {limit // (1024 * 1024)} MB upload limit.")
+                    await run_in_threadpool(handle.write, chunk)
+            asset = await run_in_threadpool(asset_store.add, path,
+                                            unquote(request.headers.get("x-asset-name", "Untitled asset")), mime_type)
+            return asset.public()
+        except ValueError as exc:
+            path.unlink(missing_ok=True)
+            raise HTTPException(400, str(exc)) from exc
+        except BaseException:
+            path.unlink(missing_ok=True)
+            raise
+
+
+@router.api_route("/assets/{asset_id}", methods=["GET", "HEAD"])
+async def get_asset(asset_id: str) -> FileResponse:
+    try:
+        asset = asset_store.get(asset_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from exc
+    return FileResponse(asset.path, media_type=asset.mime_type, headers={"X-Content-Type-Options": "nosniff"})
+
+
+@router.delete("/assets/{asset_id}", status_code=204)
+async def delete_asset(asset_id: str) -> Response:
+    try:
+        asset_store.get(asset_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from exc
+    try:
+        asset_store.delete(asset_id)
+    except ValueError as exc:
+        raise HTTPException(409, str(exc)) from exc
+    return Response(status_code=204)

apps/dreamverse/dreamverse/session/controller.py

@@ -26,11 +26,17 @@ from typing import TYPE_CHECKING
 
 from fastapi import WebSocket, WebSocketDisconnect
 from dreamverse.gpu_pool import GPUSlot
+from dreamverse.generation_inputs import (
+    GenerationInputs,
+    pin_generation_inputs,
+    release_generation_inputs,
+    resolve_generation_inputs,
+)
 from dreamverse.session_init_image import cleanup_session_init_image, persist_session_init_image
 from dreamverse.worker_ipc import MediaChunk, MediaComplete, MediaInit
 
 from dreamverse.config import (
-    DEFAULT_MODEL_ID,
+    ACTIVE_MODEL_ID,
     GENERATION_SEGMENT_CAP,
     PROMPT_AUTO_SLEEP_MS,
     PROMPT_AUTO_TIMEOUT_MS,
@@ -156,6 +162,7 @@ class SessionController:
         prompt_worker_task: asyncio.Task | None = None
         rewrite_seed_prompts_task: asyncio.Task | None = None
         session_init_image = None
+        generation_inputs: GenerationInputs | None = None
 
         async def session_timeout():
             """Close the session after timeout."""
@@ -191,6 +198,18 @@ class SessionController:
                 init_data = {}
 
             init_type = init_data.get("type")
+            try:
+                next_generation_inputs = resolve_generation_inputs(init_data, ACTIVE_MODEL_ID)
+                pin_generation_inputs(next_generation_inputs)
+                generation_inputs = next_generation_inputs
+            except ValueError as exc:
+                await ws_send_json({
+                    "type": "error",
+                    "error_code": "invalid_generation_input",
+                    "message": str(exc),
+                })
+                await websocket.close(code=1008, reason="Invalid generation inputs")
+                return
             preset_id = init_data.get("preset_id")
             preset_label = str(init_data.get("preset_label") or "").strip()
             initial_rollout_prompt = str(init_data.get("initial_rollout_prompt") or "").strip()
@@ -264,13 +283,14 @@ class SessionController:
             timeout_task = asyncio.create_task(session_timeout())
 
             # Join the engine on this GPU.
-            await slot.join_user(client_id, model_id=DEFAULT_MODEL_ID)
+            await slot.join_user(client_id, model_id=ACTIVE_MODEL_ID)
 
             # Notify client they're connected to a GPU.
             await ws_send_json({
                 "type": "gpu_assigned",
                 "gpu_id": gpu_id,
                 "session_timeout": SESSION_TIMEOUT_SECONDS,
+                "generation_mode": generation_inputs.mode,
             })
             await log_event(
                 "gpu_assigned",
@@ -361,10 +381,16 @@ class SessionController:
                     return
 
                 try:
+                    if generation_inputs.mode is not None and payload.get("initial_image") is not None:
+                        raise ValueError("Choose conditioning assets when starting a project; legacy initial_image "
+                                         "cannot replace generation mode inputs.")
+                    if "generation_mode" in payload or "conditioning_assets" in payload:
+                        raise ValueError("simple_generate cannot change the mode; use project_init_v1.")
                     replace_session_init_image(payload.get("initial_image"))
                 except ValueError as exc:
                     await ws_send_json({
                         "type": "error",
+                        "error_code": "invalid_generation_input",
                         "message": str(exc),
                     })
                     return
@@ -417,6 +443,7 @@ class SessionController:
                     })
 
             async def apply_project_init_payload(payload: dict[str, object]) -> bool:
+                nonlocal generation_inputs
                 nonlocal preset_id
                 nonlocal preset_label
                 nonlocal initial_rollout_prompt
@@ -496,15 +523,26 @@ class SessionController:
                     })
                     return False
 
+                next_generation_inputs = None
+                next_inputs_pinned = False
                 try:
+                    next_generation_inputs = resolve_generation_inputs(payload, ACTIVE_MODEL_ID)
+                    pin_generation_inputs(next_generation_inputs)
+                    next_inputs_pinned = True
                     replace_session_init_image(payload.get("initial_image"))
                 except ValueError as exc:
+                    if next_inputs_pinned:
+                        release_generation_inputs(next_generation_inputs)
                     await ws_send_json({
                         "type": "error",
+                        "error_code": "invalid_generation_input",
                         "message": str(exc),
                     })
                     return False
 
+                release_generation_inputs(generation_inputs)
+                generation_inputs = next_generation_inputs
+
                 initial_rollout_prompt = next_initial_rollout_prompt
                 enhancement_enabled = next_enhancement_enabled
                 auto_extension_enabled = next_auto_extension_enabled
@@ -1171,6 +1209,7 @@ class SessionController:
                 return drained
 
             async def enter_project_idle() -> None:
+                nonlocal generation_inputs
                 nonlocal curated_prompts
                 nonlocal seed_prompt_memory
                 nonlocal curated_idx
@@ -1221,6 +1260,9 @@ class SessionController:
                 project_active = False
                 pending_project_end = False
 
+                release_generation_inputs(generation_inputs)
+                generation_inputs = GenerationInputs()
+
                 if project_stream_started:
                     project_stream_started = False
                     await ws_send_json({"type": "ltx2_stream_complete"})
@@ -1627,6 +1669,7 @@ class SessionController:
                         segment_idx=segment_idx,
                         image_path=step_image_path,
                         reset_conditioning=step_reset_conditioning,
+                        generation_inputs=generation_inputs,
                     ))
                 segment_generation_active = True
                 try:
@@ -1681,10 +1724,10 @@ class SessionController:
                                 print(f"[GPU {gpu_id}] Unknown AV event: "
                                       f"{type(event).__name__}")
 
-                    if not step_task.done():
-                        step_task.cancel()
-                    else:
-                        timings = await step_task
+                    # A GPU command cannot be cancelled by cancelling its
+                    # asyncio waiter. Await completion before releasing pinned
+                    # asset files or making this GPU available to a new user.
+                    timings = await step_task
                 finally:
                     segment_generation_active = False
                     if not step_task.done():
@@ -1808,3 +1851,5 @@ class SessionController:
                     await self.gpu_pool.release(client_id)
             finally:
                 cleanup_session_init_image(session_init_image)
+                if generation_inputs is not None:
+                    release_generation_inputs(generation_inputs)

apps/dreamverse/dreamverse/tests/conftest.py

@@ -3,7 +3,6 @@ from __future__ import annotations
 import sys
 from pathlib import Path
 
-
 TESTS_DIR = Path(__file__).resolve().parent
 DREAMVERSE_PACKAGE_DIR = TESTS_DIR.parent
 DREAMVERSE_APP_DIR = DREAMVERSE_PACKAGE_DIR.parent

apps/dreamverse/dreamverse/tests/test_config.py

@@ -2,14 +2,14 @@ from __future__ import annotations
 
 import importlib.util
 from pathlib import Path
+from types import ModuleType
 
 import pytest
 
-
 SERVER_DIR = Path(__file__).resolve().parents[1]
 
 
-def _load_config_module():
+def _load_config_module() -> ModuleType:
     spec = importlib.util.spec_from_file_location(
         "server_config_test_module",
         SERVER_DIR / "config.py",
@@ -21,7 +21,7 @@ def _load_config_module():
     return module
 
 
-def _set_required_prompt_keys(monkeypatch):
+def _set_required_prompt_keys(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setenv("CEREBRAS_API_KEY", "cerebras-key")
     monkeypatch.setenv("GROQ_API_KEY", "groq-key")
 
@@ -53,9 +53,7 @@ def test_config_defaults_to_cerebras_with_parallel_groq_fallback_stage(monkeypat
     module = _load_config_module()
 
     assert module.PROMPT_PROVIDER == "cerebras"
-    assert module.PROMPT_PROVIDER_RUNTIME_STAGES == (
-        ("cerebras", "groq"),
-    )
+    assert module.PROMPT_PROVIDER_RUNTIME_STAGES == (("cerebras", "groq"), )
     assert module.PROMPT_PROVIDER_PRIORITY == (
         "cerebras",
         "groq",
@@ -86,9 +84,7 @@ def test_config_ignores_legacy_groq_primary_override(monkeypatch):
     module = _load_config_module()
 
     assert module.PROMPT_PROVIDER == "cerebras"
-    assert module.PROMPT_PROVIDER_RUNTIME_STAGES == (
-        ("cerebras", "groq"),
-    )
+    assert module.PROMPT_PROVIDER_RUNTIME_STAGES == (("cerebras", "groq"), )
     assert module.PROMPT_PROVIDER_PRIORITY == (
         "cerebras",
         "groq",
@@ -106,24 +102,17 @@ def test_config_uses_local_overlay_paths_when_devtools_enabled(monkeypatch, tmp_
 
     assert module.DEVTOOLS_ENABLED is True
     assert module.FRONTEND_ROOT.as_posix().endswith("apps/dreamverse/web")
-    assert module.PROMPT_ENHANCE_SYSTEM_PROMPT_PATH.endswith(
-        "dreamverse/prompts.local/next_segment_system_prompt.md"
-    )
+    assert module.PROMPT_ENHANCE_SYSTEM_PROMPT_PATH.endswith("dreamverse/prompts.local/next_segment_system_prompt.md")
     assert module.PROMPT_ENHANCE_SYSTEM_PROMPT_FALLBACK_PATH.endswith(
-        "dreamverse/prompts/next_segment_system_prompt.md"
-    )
+        "dreamverse/prompts/next_segment_system_prompt.md")
     assert module.PROMPT_REWRITE_USER_SYSTEM_PROMPT_PATH.endswith(
-        "dreamverse/prompts.local/rewrite_user_system_prompt.md"
-    )
+        "dreamverse/prompts.local/rewrite_user_system_prompt.md")
     assert module.PROMPT_REWRITE_USER_SYSTEM_PROMPT_FALLBACK_PATH.endswith(
-        "dreamverse/prompts/rewrite_user_system_prompt.md"
-    )
+        "dreamverse/prompts/rewrite_user_system_prompt.md")
     assert module.CURATED_PRESETS_FILE_PATH.endswith(
-        "apps/dreamverse/web/prompts.local/selected_ltx2_continuation_story_presets.json"
-    )
+        "apps/dreamverse/web/prompts.local/selected_ltx2_continuation_story_presets.json")
     assert module.CURATED_PRESETS_FALLBACK_FILE_PATH.endswith(
-        "apps/dreamverse/web/prompts/selected_ltx2_continuation_story_presets.json"
-    )
+        "apps/dreamverse/web/prompts/selected_ltx2_continuation_story_presets.json")
     assert module.FRONTEND_STATIC_DIR_CANDIDATES[:2] == (
         str(module.FRONTEND_ROOT / "out"),
         str(module.FRONTEND_ROOT / "dist"),
@@ -150,15 +139,140 @@ def test_config_enables_prompt_safety_when_requested(monkeypatch):
 
 def test_config_uses_five_minute_session_timeout(monkeypatch):
     _set_required_prompt_keys(monkeypatch)
+    monkeypatch.delenv("DREAMVERSE_MODEL_ID", raising=False)
+    monkeypatch.delenv("DREAMVERSE_SESSION_TIMEOUT_SECONDS", raising=False)
+    monkeypatch.delenv("FASTVIDEO_SESSION_TIMEOUT_SECONDS", raising=False)
 
     module = _load_config_module()
 
     assert module.SESSION_TIMEOUT_SECONDS == 300
 
 
+def test_config_uses_thirty_minute_cosmos25_session_timeout(monkeypatch):
+    _set_required_prompt_keys(monkeypatch)
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "cosmos25-dfd")
+    monkeypatch.delenv("DREAMVERSE_SESSION_TIMEOUT_SECONDS", raising=False)
+    monkeypatch.delenv("FASTVIDEO_SESSION_TIMEOUT_SECONDS", raising=False)
+
+    module = _load_config_module()
+
+    assert module.SESSION_TIMEOUT_SECONDS == 1800
+
+
+def test_config_allows_session_timeout_override(monkeypatch):
+    _set_required_prompt_keys(monkeypatch)
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "cosmos25-dfd")
+    monkeypatch.delenv("DREAMVERSE_SESSION_TIMEOUT_SECONDS", raising=False)
+    monkeypatch.setenv("FASTVIDEO_SESSION_TIMEOUT_SECONDS", "900")
+
+    module = _load_config_module()
+
+    assert module.SESSION_TIMEOUT_SECONDS == 900
+
+
+def test_config_prefers_dreamverse_session_timeout_over_alias(monkeypatch):
+    _set_required_prompt_keys(monkeypatch)
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "cosmos25-dfd")
+    monkeypatch.setenv("DREAMVERSE_SESSION_TIMEOUT_SECONDS", "1200")
+    monkeypatch.setenv("FASTVIDEO_SESSION_TIMEOUT_SECONDS", "900")
+
+    module = _load_config_module()
+
+    assert module.SESSION_TIMEOUT_SECONDS == 1200
+
+
 def test_config_rejects_invalid_prompt_provider(monkeypatch):
     monkeypatch.setenv("FASTVIDEO_PROMPT_PROVIDER", "unsupported")
     _set_required_prompt_keys(monkeypatch)
 
     with pytest.raises(RuntimeError, match="Invalid FASTVIDEO_PROMPT_PROVIDER"):
         _load_config_module()
+
+
+def test_config_registers_vsa_datafree_fasth3_profile(monkeypatch):
+    """The FastH3 registry entry owns the complete fixed Preview recipe."""
+    _set_required_prompt_keys(monkeypatch)
+
+    module = _load_config_module()
+
+    assert module.MODEL_REGISTRY["fast-h3"] == {
+        "name": "FastH3",
+        "generation_backend": "minimax_h3",
+        "default_sp_size": 4,
+        "model_path": "MiniMaxAI/MiniMax-H3",
+        "adapter_repo": "FastVideo/FastVideo-FastH3-4-step-Preview-v1-LoRA",
+        "adapter_filename": "vsa-datafree/adapter_model.safetensors",
+        "attention_backend": "VIDEO_SPARSE_ATTN_H3",
+        "height": 768,
+        "width": 1344,
+        "num_frames": 124,
+        "num_inference_steps": 5,
+        "seed": 1000,
+    }
+
+
+def test_config_uses_fasth3_sequence_parallel_default(monkeypatch):
+    """Selecting FastH3 defaults DreamVerse to its four-GPU topology."""
+    _set_required_prompt_keys(monkeypatch)
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "fast-h3")
+    monkeypatch.delenv("DREAMVERSE_SP_SIZE", raising=False)
+
+    module = _load_config_module()
+
+    assert module.ACTIVE_MODEL_ID == "fast-h3"
+    assert module.MODEL_CONFIG["generation_backend"] == "minimax_h3"
+    assert module.DREAMVERSE_SP_SIZE == 4
+
+
+def test_full_h3_profile_has_no_preview_adapter_and_longer_session(monkeypatch):
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "full-h3")
+    monkeypatch.delenv("DREAMVERSE_SP_SIZE", raising=False)
+    monkeypatch.delenv("DREAMVERSE_SESSION_TIMEOUT_SECONDS", raising=False)
+    monkeypatch.delenv("FASTVIDEO_SESSION_TIMEOUT_SECONDS", raising=False)
+    module = _load_config_module()
+    assert module.MODEL_CONFIG["full_checkpoint"] is True
+    assert "adapter_repo" not in module.MODEL_CONFIG
+    assert module.MODEL_CONFIG["num_inference_steps"] == 50
+    assert module.DREAMVERSE_SP_SIZE == 4
+    assert module.SESSION_TIMEOUT_SECONDS == 7200
+
+
+def test_config_registers_cosmos25_dfd_profile(monkeypatch):
+    _set_required_prompt_keys(monkeypatch)
+
+    module = _load_config_module()
+
+    assert module.MODEL_REGISTRY["cosmos25-dfd"] == {
+        "name": "Cosmos Predict2.5 DFD",
+        "generation_backend": "cosmos25_dfd",
+        "default_sp_size": 1,
+        "model_path": "FastVideo/Cosmos-Predict2.5-2B-Distilled-TrigFlow",
+        "continuation_model_path": "FastVideo/Cosmos-Predict2.5-2B-DFD",
+        "attention_backend": "TORCH_SDPA",
+        "height": 704,
+        "width": 1280,
+        "bootstrap_num_frames": 77,
+        "continuation_num_frames": 81,
+        "fps": 24,
+        "num_inference_steps": 4,
+        "seed": 42,
+        "session_timeout_seconds": 1800,
+    }
+
+
+def test_config_selects_cosmos25_package_roles(monkeypatch, tmp_path):
+    _set_required_prompt_keys(monkeypatch)
+    bootstrap_path = tmp_path / "cosmos25-t2w"
+    continuation_path = tmp_path / "cosmos25-dfd"
+    monkeypatch.setenv("DREAMVERSE_MODEL_ID", "cosmos25-dfd")
+    monkeypatch.setenv("DREAMVERSE_MODEL_PATH", str(bootstrap_path))
+    monkeypatch.setenv("DREAMVERSE_COSMOS25_DFD_MODEL_PATH", str(continuation_path))
+    monkeypatch.delenv("DREAMVERSE_SP_SIZE", raising=False)
+
+    module = _load_config_module()
+
+    assert module.ACTIVE_MODEL_ID == "cosmos25-dfd"
+    assert module.MODEL_CONFIG["generation_backend"] == "cosmos25_dfd"
+    assert module.MODEL_CONFIG["model_path"] == str(bootstrap_path)
+    assert module.MODEL_CONFIG["continuation_model_path"] == str(continuation_path)
+    assert module.DREAMVERSE_SP_SIZE == 1

apps/dreamverse/dreamverse/tests/test_cosmos25_dfd_generation.py

@@ -0,0 +1,219 @@
+from __future__ import annotations
+
+import os
+from pathlib import Path
+from types import SimpleNamespace
+
+import numpy as np
+import pytest
+
+from dreamverse.cosmos25_dfd_generation import Cosmos25DFDGenerationBackend
+from dreamverse.generation_inputs import GenerationInputs
+
+COSMOS_CONFIG = {
+    "name": "Cosmos Predict2.5 DFD",
+    "generation_backend": "cosmos25_dfd",
+    "default_sp_size": 1,
+    "model_path": "/models/cosmos25-t2w",
+    "continuation_model_path": "/models/cosmos25-dfd",
+    "attention_backend": "TORCH_SDPA",
+    "height": 704,
+    "width": 1280,
+    "bootstrap_num_frames": 77,
+    "continuation_num_frames": 81,
+    "fps": 24,
+    "num_inference_steps": 4,
+    "seed": 42,
+}
+
+
+class _RecordingGenerator:
+    def __init__(self, pixel_value: int = 20) -> None:
+        self.pixel_value = pixel_value
+        self.calls: list[dict] = []
+        self.shutdown_calls = 0
+
+    def generate_video(self, prompt, sampling_param):
+        condition = sampling_param.pil_image
+        self.calls.append({
+            "prompt": prompt,
+            "sampling": sampling_param,
+            "conditioning_pixels": None if condition is None else np.asarray(condition).copy(),
+        })
+        frames = [
+            np.full((2, 3, 3), self.pixel_value, dtype=np.uint8)
+            for _ in range(sampling_param.num_frames)
+        ]
+        frames[-1] = np.full((2, 3, 3), self.pixel_value + 1, dtype=np.uint8)
+        return {
+            "frames": frames,
+            "generation_time": 0.25,
+        }
+
+    def shutdown(self):
+        self.shutdown_calls += 1
+
+
+@pytest.fixture
+def backend(monkeypatch) -> Cosmos25DFDGenerationBackend:
+    instance = Cosmos25DFDGenerationBackend(gpu_id=0)
+    instance.model_config = dict(COSMOS_CONFIG)
+    instance.bootstrap_generator = _RecordingGenerator(pixel_value=20)
+    instance.continuation_generator = _RecordingGenerator(pixel_value=40)
+    monkeypatch.setattr("dreamverse.cosmos25_dfd_generation.torch.cuda.synchronize", lambda: None)
+
+    def fake_sampling_param(*, conditioned):
+        return SimpleNamespace(
+            negative_prompt="",
+            save_video=False,
+            return_frames=True,
+            height=704,
+            width=1280,
+            num_frames=81 if conditioned else 77,
+            fps=24,
+            num_inference_steps=4,
+            guidance_scale=1.0,
+            seed=42,
+            num_cond_frames=1 if conditioned else 0,
+            pil_image=None,
+        )
+
+    monkeypatch.setattr(instance, "_sampling_param", fake_sampling_param)
+    return instance
+
+
+def test_initialize_loads_both_package_roles(monkeypatch):
+    loaded_paths = []
+    generators = [_RecordingGenerator(), _RecordingGenerator()]
+    backend = Cosmos25DFDGenerationBackend(gpu_id=0)
+
+    def fake_load(model_path):
+        loaded_paths.append(model_path)
+        return generators[len(loaded_paths) - 1]
+
+    monkeypatch.setattr(backend, "_load_generator", fake_load)
+    monkeypatch.setattr(backend, "_gpu_mem", lambda: "alloc=0.00GiB, reserved=0.00GiB")
+    monkeypatch.setattr("dreamverse.cosmos25_dfd_generation.gc.collect", lambda: 0)
+    monkeypatch.setattr("dreamverse.cosmos25_dfd_generation.torch.cuda.is_available", lambda: False)
+    monkeypatch.setenv("FASTVIDEO_ATTENTION_BACKEND", "test-attention")
+    monkeypatch.setenv("FASTVIDEO_INFERENCE_TORCH_COMPILE", "1")
+
+    backend.initialize(COSMOS_CONFIG)
+
+    assert loaded_paths == [
+        "/models/cosmos25-t2w",
+        "/models/cosmos25-dfd",
+    ]
+    assert backend.bootstrap_generator is generators[0]
+    assert backend.continuation_generator is generators[1]
+    assert backend.model_config == COSMOS_CONFIG
+    assert os.environ["FASTVIDEO_ATTENTION_BACKEND"] == "TORCH_SDPA"
+    assert "FASTVIDEO_INFERENCE_TORCH_COMPILE" not in os.environ
+
+
+def test_unconditioned_start_uses_t2w_and_retains_terminal_frame(backend):
+    result = backend.generate_step("first prompt", 1, None, True)
+
+    assert len(backend.bootstrap_generator.calls) == 1
+    assert backend.continuation_generator.calls == []
+    sampling = backend.bootstrap_generator.calls[0]["sampling"]
+    assert sampling.height == 704
+    assert sampling.width == 1280
+    assert sampling.num_frames == 77
+    assert sampling.fps == 24
+    assert sampling.num_inference_steps == 4
+    assert sampling.guidance_scale == 1.0
+    assert sampling.seed == 42
+    assert sampling.num_cond_frames == 0
+    assert sampling.pil_image is None
+    assert result.head_trim_frames == 0
+    assert result.head_trim_audio_frames == 0
+    assert result.audio_sample_rate == 24_000
+    assert result.audio.shape == (77_000, )
+    assert result.audio.count_nonzero() == 0
+    assert np.asarray(backend.continuation_image).tolist() == np.full((2, 3, 3), 21).tolist()
+
+
+def test_retained_frame_uses_dfd_and_trims_repeated_boundary(backend):
+    backend.generate_step("first prompt", 1, None, True)
+    result = backend.generate_step("pivot right", 2, None, False)
+
+    assert len(backend.continuation_generator.calls) == 1
+    call = backend.continuation_generator.calls[0]
+    sampling = call["sampling"]
+    assert sampling.num_frames == 81
+    assert sampling.num_cond_frames == 1
+    assert call["conditioning_pixels"].tolist() == np.full((2, 3, 3), 21).tolist()
+    assert result.head_trim_frames == 1
+    assert result.head_trim_audio_frames == 1
+    assert result.audio.shape == (81_000, )
+    assert np.asarray(backend.continuation_image).tolist() == np.full((2, 3, 3), 41).tolist()
+
+
+def test_initial_image_uses_dfd_without_stream_trim(backend, tmp_path: Path):
+    from PIL import Image
+
+    image_path = tmp_path / "initial.png"
+    Image.fromarray(np.full((2, 3, 3), 7, dtype=np.uint8)).save(image_path)
+
+    result = backend.generate_step("animate", 1, str(image_path), True)
+
+    assert backend.bootstrap_generator.calls == []
+    call = backend.continuation_generator.calls[0]
+    assert call["conditioning_pixels"].tolist() == np.full((2, 3, 3), 7).tolist()
+    assert result.head_trim_frames == 0
+    assert result.head_trim_audio_frames == 0
+
+
+def test_generation_mode_api_accepts_text_only_and_rejects_conditioning_modes(backend):
+    result = backend.generate_step("first prompt", 1, None, True, generation_inputs=GenerationInputs(mode="t2va"))
+
+    assert len(backend.bootstrap_generator.calls) == 1
+    assert result.head_trim_frames == 0
+
+    with pytest.raises(ValueError, match="text generation only"):
+        backend.generate_step("pivot right", 2, None, False, generation_inputs=GenerationInputs(mode="fl2va"))
+    assert backend.continuation_generator.calls == []
+
+
+def test_missing_later_continuation_fails_before_generation(backend):
+    with pytest.raises(RuntimeError, match="requires a retained continuation frame"):
+        backend.generate_step("later prompt", 2, None, False)
+
+    assert backend.bootstrap_generator.calls == []
+    assert backend.continuation_generator.calls == []
+
+
+def test_reset_later_segment_uses_fresh_t2w_bootstrap(backend):
+    backend.generate_step("first prompt", 1, None, True)
+
+    result = backend.generate_step("new scene", 2, None, True)
+
+    assert len(backend.bootstrap_generator.calls) == 2
+    assert backend.continuation_generator.calls == []
+    assert result.head_trim_frames == 0
+
+
+def test_warmup_exercises_bootstrap_and_dfd_paths(backend):
+    timings = backend.warmup("warmup prompt")
+
+    assert len(backend.bootstrap_generator.calls) == 1
+    assert len(backend.continuation_generator.calls) == 1
+    assert backend.continuation_image is None
+    assert "warmup_bootstrap_ms" in timings
+    assert "warmup_continuation_ms" in timings
+    assert "warmup_total_ms" in timings
+
+
+def test_shutdown_releases_both_generators_and_conditioning(backend):
+    bootstrap = backend.bootstrap_generator
+    continuation = backend.continuation_generator
+    backend.generate_step("first prompt", 1, None, True)
+
+    backend.shutdown()
+
+    assert bootstrap.shutdown_calls == 1
+    assert continuation.shutdown_calls == 1
+    assert backend.bootstrap_generator is None
+    assert backend.continuation_generator is None
+    assert backend.continuation_image is None

apps/dreamverse/dreamverse/tests/test_entrypoints.py

@@ -9,6 +9,7 @@ from fastapi.testclient import TestClient
 import fastvideo.entrypoints.streaming as streaming_entrypoints
 import pytest
 
+
 def _install_stack03_import_stubs(monkeypatch):
     """Keep entrypoint tests focused while later-stack runtime modules are absent."""
     if not hasattr(streaming_entrypoints, "build_health_router"):
@@ -17,6 +18,7 @@ def _install_stack03_import_stubs(monkeypatch):
     gpu_pool_stub = types.ModuleType("dreamverse.gpu_pool")
 
     class GPUPool:
+
         def __init__(self, _gpu_ids):
             pass
 
@@ -49,6 +51,7 @@ def _install_stack03_import_stubs(monkeypatch):
     controller_stub = types.ModuleType("dreamverse.session.controller")
 
     class SessionController:
+
         def __init__(self, **_kwargs):
             pass
 
@@ -76,13 +79,11 @@ def _run_cli(module, monkeypatch, argv: list[str]) -> list[dict[str, object]]:
     uvicorn_stub = types.ModuleType("uvicorn")
 
     def run(app, host: str, port: int) -> None:
-        calls.append(
-            {
-                "app": app,
-                "host": host,
-                "port": port,
-            }
-        )
+        calls.append({
+            "app": app,
+            "host": host,
+            "port": port,
+        })
 
     uvicorn_stub.run = run
     monkeypatch.setitem(sys.modules, "uvicorn", uvicorn_stub)
@@ -99,13 +100,11 @@ def test_server_cli_defaults_to_local_web_port(monkeypatch):
     server_main = _import_server_main(monkeypatch)
     calls = _run_cli(server_main, monkeypatch, ["dreamverse-server"])
 
-    assert calls == [
-        {
-            "app": server_main.app,
-            "host": "0.0.0.0",
-            "port": 8009,
-        }
-    ]
+    assert calls == [{
+        "app": server_main.app,
+        "host": "0.0.0.0",
+        "port": 8009,
+    }]
 
 
 def test_server_cli_allows_explicit_host_and_port(monkeypatch):
@@ -116,13 +115,11 @@ def test_server_cli_allows_explicit_host_and_port(monkeypatch):
         ["dreamverse-server", "--host", "127.0.0.1", "--port", "8123"],
     )
 
-    assert calls == [
-        {
-            "app": server_main.app,
-            "host": "127.0.0.1",
-            "port": 8123,
-        }
-    ]
+    assert calls == [{
+        "app": server_main.app,
+        "host": "127.0.0.1",
+        "port": 8123,
+    }]
 
 
 def test_server_does_not_expose_backend_source_as_static_assets(monkeypatch):
@@ -142,13 +139,11 @@ def test_mock_server_cli_defaults_to_local_web_port(monkeypatch):
         ["dreamverse-mock-server"],
     )
 
-    assert calls == [
-        {
-            "app": mock_server.app,
-            "host": "0.0.0.0",
-            "port": 8009,
-        }
-    ]
+    assert calls == [{
+        "app": mock_server.app,
+        "host": "0.0.0.0",
+        "port": 8009,
+    }]
 
 
 def test_mock_server_cli_updates_latency(monkeypatch):
@@ -161,13 +156,11 @@ def test_mock_server_cli_updates_latency(monkeypatch):
             ["dreamverse-mock-server", "--latency", "321", "--port", "8111"],
         )
 
-        assert calls == [
-            {
-                "app": mock_server.app,
-                "host": "0.0.0.0",
-                "port": 8111,
-            }
-        ]
+        assert calls == [{
+            "app": mock_server.app,
+            "host": "0.0.0.0",
+            "port": 8111,
+        }]
         assert mock_server.LATENCY_MS == 321
     finally:
         mock_server.LATENCY_MS = old_latency_ms

apps/dreamverse/dreamverse/tests/test_generation_inputs.py

@@ -0,0 +1,248 @@
+"""Contract regressions independent of CUDA and model weights."""
+
+import io
+import asyncio
+import shutil
+import subprocess
+
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+from PIL import Image
+
+from dreamverse import assets, generation_inputs
+from dreamverse.generation_inputs import resolve_generation_inputs
+from dreamverse.routes import assets as asset_routes
+from dreamverse.tests.test_mock_server import _FakeWebSocket
+
+
+@pytest.fixture
+def library(monkeypatch):
+    store = assets.AssetStore()
+    monkeypatch.setattr(asset_routes, "asset_store", store)
+    monkeypatch.setattr(generation_inputs, "asset_store", store)
+    app = FastAPI()
+    app.include_router(asset_routes.router)
+    with TestClient(app) as client:
+        yield store, client
+
+
+def upload_image(client, color="red"):
+    image_bytes = io.BytesIO()
+    Image.new("RGB", (32, 32), color).save(image_bytes, format="PNG")
+    response = client.post("/assets", content=image_bytes.getvalue(),
+                           headers={"Content-Type": "image/png", "X-Asset-Name": "frame.png"})
+    assert response.status_code == 201, response.text
+    return response.json()
+
+
+def conditioning(asset, role):
+    return {"asset_id": asset["asset_id"], "role": role}
+
+
+def test_assets_validate_content_and_support_head_range_and_delete(library):
+    store, client = library
+    asset = upload_image(client)
+    assert set(asset) == {"asset_id", "kind", "name", "mime_type", "size", "url"}
+    assert client.head(asset["url"]).status_code == 200
+    response = client.get(asset["url"], headers={"Range": "bytes=0-7"})
+    assert response.status_code == 206
+    assert response.content == b"\x89PNG\r\n\x1a\n"
+    assert client.post("/assets", content=b"not an image", headers={"Content-Type": "image/png"}).status_code == 400
+    assert client.post("/assets", content=b"<svg/>", headers={"Content-Type": "image/svg+xml"}).status_code == 415
+    assert client.post("/assets", content=b"", headers={"Content-Type": "image/png",
+                                                      "Content-Length": str(assets.IMAGE_LIMIT + 1)}).status_code == 413
+    with pytest.raises(ValueError, match="Invalid asset ID"):
+        store.get("../../etc/passwd")
+    assert client.delete(asset["url"]).status_code == 204
+    assert client.head(asset["url"]).status_code == 404
+
+
+def test_generation_pin_prevents_deletion_until_session_releases(library):
+    _, client = library
+    asset = upload_image(client)
+    inputs = resolve_generation_inputs({"generation_mode": "fl2va", "conditioning_assets": [
+        conditioning(asset, "first_frame")
+    ]}, "full-h3")
+    generation_inputs.pin_generation_inputs(inputs)
+    generation_inputs.pin_generation_inputs(inputs)
+    assert client.delete(asset["url"]).status_code == 409
+    generation_inputs.release_generation_inputs(inputs)
+    assert client.delete(asset["url"]).status_code == 409
+    generation_inputs.release_generation_inputs(inputs)
+    assert client.delete(asset["url"]).status_code == 204
+
+
+def test_legacy_init_remains_compatible_but_explicit_t2va_is_text_only(library):
+    _, client = library
+    assert resolve_generation_inputs({"initial_image": {"old": "payload"}}, "fast-ltx2").mode is None
+    assert resolve_generation_inputs({"generation_mode": "t2va"}, "fast-ltx2").mode == "t2va"
+    with pytest.raises(ValueError, match="legacy initial_image"):
+        resolve_generation_inputs({"generation_mode": "t2va", "initial_image": {}}, "full-h3")
+    asset = upload_image(client)
+    with pytest.raises(ValueError, match="text only"):
+        resolve_generation_inputs({"generation_mode": "t2va", "conditioning_assets": [
+            conditioning(asset, "reference")
+        ]}, "full-h3")
+
+
+@pytest.mark.parametrize("mode", ["unknown", None, 3, [], {}])
+def test_unknown_mode_fails_before_assets_are_resolved(mode):
+    with pytest.raises(ValueError, match="Unknown generation mode"):
+        resolve_generation_inputs({"generation_mode": mode}, "full-h3")
+
+
+@pytest.mark.parametrize("model_id", ["fast-h3", "fast-ltx2", "fast-ltx23"])
+def test_preview_and_ltx_cannot_advertise_full_h3_modes(model_id):
+    with pytest.raises(ValueError, match="Full H3"):
+        resolve_generation_inputs({"generation_mode": "ref2va"}, model_id)
+
+
+def test_fl2va_first_required_last_optional_and_roles_unique(library):
+    _, client = library
+    first = upload_image(client)
+    last = upload_image(client, "blue")
+    payload = {"generation_mode": "fl2va", "conditioning_assets": [conditioning(first, "first_frame")]}
+    inputs = resolve_generation_inputs(payload, "full-h3")
+    assert inputs.first_frame_path.endswith(".png")
+    assert inputs.last_frame_path is None
+    payload["conditioning_assets"].append(conditioning(last, "last_frame"))
+    assert resolve_generation_inputs(payload, "full-h3").last_frame_path is not None
+    payload["conditioning_assets"].append(conditioning(first, "first_frame"))
+    with pytest.raises(ValueError, match="exactly one first-frame"):
+        resolve_generation_inputs(payload, "full-h3")
+    with pytest.raises(ValueError, match="exactly one first-frame"):
+        resolve_generation_inputs({"generation_mode": "fl2va", "conditioning_assets": [
+            conditioning(last, "last_frame")
+        ]}, "full-h3")
+
+
+def test_ref_order_is_preserved_and_limits_are_enforced(library):
+    _, client = library
+    first, second = upload_image(client), upload_image(client, "blue")
+    refs = [conditioning(second, "reference"), conditioning(first, "reference")]
+    payload = {"generation_mode": "ref2va", "conditioning_assets": refs}
+    inputs = resolve_generation_inputs(payload, "full-h3")
+    assert [asset.asset_id for asset in inputs.references] == [second["asset_id"], first["asset_id"]]
+    with pytest.raises(ValueError, match="at most 9 image"):
+        resolve_generation_inputs({**payload, "conditioning_assets": refs * 5}, "full-h3")
+    with pytest.raises(ValueError, match="without keyframe roles"):
+        resolve_generation_inputs({**payload, "conditioning_assets": [conditioning(first, "first_frame")]}, "full-h3")
+    with pytest.raises(ValueError, match="at most 12"):
+        resolve_generation_inputs({**payload, "conditioning_assets": refs * 7}, "full-h3")
+
+
+def test_ref_audio_requires_visual_reference(library, monkeypatch):
+    store, _ = library
+    monkeypatch.setattr(store, "get", lambda asset_id: assets.StoredAsset(asset_id, "audio", "/audio.wav", "audio",
+                                                                         "audio/wav", 100))
+    with pytest.raises(ValueError, match="audio alone"):
+        resolve_generation_inputs({"generation_mode": "ref2va", "conditioning_assets": [
+            {"asset_id": "a" * 32, "role": "reference"}
+        ]}, "full-h3")
+
+
+def test_ref_rejects_extreme_image_aspect_before_gpu(library):
+    _, client = library
+    content = io.BytesIO()
+    Image.new("RGB", (500, 50), "blue").save(content, format="PNG")
+    response = client.post("/assets", content=content.getvalue(), headers={"Content-Type": "image/png"})
+    assert response.status_code == 201
+    with pytest.raises(ValueError, match="aspect ratios"):
+        resolve_generation_inputs({"generation_mode": "ref2va", "conditioning_assets": [
+            conditioning(response.json(), "reference")
+        ]}, "full-h3")
+
+
+def test_ref_reports_undecodable_image_as_invalid_input(library, monkeypatch, tmp_path):
+    store, _ = library
+    broken = tmp_path / "broken.png"
+    broken.write_bytes(b"not an image")
+    monkeypatch.setattr(store, "get", lambda asset_id: assets.StoredAsset(asset_id, "image", str(broken), "broken.png",
+                                                                         "image/png", 11))
+    with pytest.raises(ValueError, match="could not be decoded"):
+        resolve_generation_inputs({"generation_mode": "ref2va", "conditioning_assets": [
+            {"asset_id": "a" * 32, "role": "reference"}
+        ]}, "full-h3")
+
+
+def test_audio_upload_rejects_surround_sound(library, monkeypatch):
+    import json
+    _, client = library
+    monkeypatch.setattr(assets.shutil, "which", lambda name: "/usr/bin/ffprobe")
+    info = {"format": {"format_name": "wav", "duration": "1"},
+            "streams": [{"codec_type": "audio", "channels": 6}]}
+    monkeypatch.setattr(assets.subprocess, "run", lambda *args, **kwargs: subprocess.CompletedProcess(
+        [], 0, stdout=json.dumps(info).encode(), stderr=b""))
+    response = client.post("/assets", content=b"surround wav", headers={"Content-Type": "audio/wav"})
+    assert response.status_code == 400
+    assert "mono or stereo" in response.json()["detail"]
+
+
+@pytest.mark.parametrize("mime,format_name", [("audio/x-m4a", "mov,mp4,m4a,3gp,3g2,mj2"), ("audio/x-flac", "flac")])
+def test_legacy_audio_mime_aliases_are_accepted(library, monkeypatch, mime, format_name):
+    """Browsers report x- variants for the M4A and FLAC formats the docs promise."""
+    import json
+    _, client = library
+    monkeypatch.setattr(assets.shutil, "which", lambda name: "/usr/bin/ffprobe")
+    info = {"format": {"format_name": format_name, "duration": "1"},
+            "streams": [{"codec_type": "audio", "channels": 2}]}
+    monkeypatch.setattr(assets.subprocess, "run", lambda *args, **kwargs: subprocess.CompletedProcess(
+        [], 0, stdout=json.dumps(info).encode(), stderr=b""))
+    response = client.post("/assets", content=b"audio bytes", headers={"Content-Type": mime})
+    assert response.status_code == 201, response.text
+    assert response.json()["kind"] == "audio"
+    assert response.json()["mime_type"] == mime
+
+
+@pytest.mark.parametrize("entries", [None, {}, "x", [{"path": "/etc/passwd", "role": "reference"}],
+                                    [{"asset_id": "x", "role": "unknown"}]])
+def test_malformed_conditioning_is_rejected(entries):
+    with pytest.raises(ValueError):
+        resolve_generation_inputs({"generation_mode": "ref2va", "conditioning_assets": entries}, "full-h3")
+
+
+@pytest.mark.parametrize("mode", ["t2va", "fl2va", "ref2va"])
+def test_mock_streams_all_valid_modes_and_releases_assets(library, monkeypatch, mode):
+    from dreamverse import mock_server
+    _, client = library
+    monkeypatch.setattr(mock_server, "MOCK_SEGMENT_BYTES", b"mock-fmp4")
+    monkeypatch.setattr(mock_server, "LATENCY_MS", 1)
+    image = upload_image(client)
+    refs = [] if mode == "t2va" else [conditioning(image, "first_frame" if mode == "fl2va" else "reference")]
+    ws = _FakeWebSocket([
+        (0, {"type": "session_init_v2", "generation_mode": mode, "conditioning_assets": refs,
+             "curated_prompts": ["A bird flies over a lake."], "single_clip_mode": True,
+             "enhancement_enabled": False}),
+        (0.15, {"type": "leave"}),
+    ])
+    asyncio.run(mock_server.websocket_endpoint(ws))
+    assert not [event for event in ws.sent_json if event["type"] == "error"]
+    assert any(event["type"] == "media_segment_complete" for event in ws.sent_json)
+    assert ws.sent_bytes
+    assert client.delete(image["url"]).status_code == 204
+
+
+def test_mock_rejects_invalid_mode_before_gpu_assignment(library):
+    from dreamverse import mock_server
+    ws = _FakeWebSocket([(0, {"type": "session_init_v2", "generation_mode": "fl2va"})])
+    asyncio.run(mock_server.websocket_endpoint(ws))
+    assert not any(event["type"] == "gpu_assigned" for event in ws.sent_json)
+    errors = [event for event in ws.sent_json if event["type"] == "error"]
+    assert errors[0]["error_code"] == "invalid_generation_input"
+    assert "first-frame" in errors[0]["message"]
+
+
+@pytest.mark.skipif(not shutil.which("ffmpeg") or not shutil.which("ffprobe"), reason="ffmpeg + ffprobe required")
+@pytest.mark.parametrize("kind,mime,suffix", [("video", "video/mp4", ".mp4"), ("audio", "audio/wav", ".wav")])
+def test_actual_video_and_audio_upload_validation(library, tmp_path, kind, mime, suffix):
+    _, client = library
+    media_path = tmp_path / f"sample{suffix}"
+    source = "testsrc2=size=64x64:rate=24" if kind == "video" else "sine=frequency=440:sample_rate=24000"
+    command = [shutil.which("ffmpeg"), "-v", "error", "-f", "lavfi", "-i", source, "-t", "0.5", str(media_path)]
+    subprocess.run(command, check=True, capture_output=True, timeout=30)
+    response = client.post("/assets", content=media_path.read_bytes(), headers={"Content-Type": mime})
+    assert response.status_code == 201, response.text
+    assert response.json()["kind"] == kind
+    response = client.post("/assets", content=b"#EXTM3U\nhttp://example.com/stream", headers={"Content-Type": mime})
+    assert response.status_code == 400

apps/dreamverse/dreamverse/tests/test_generation_mode_routing.py

@@ -0,0 +1,214 @@
+"""CPU contract tests; fake executors do not validate generated-media quality."""
+
+from __future__ import annotations
+
+import importlib.util
+import pickle
+import sys
+from pathlib import Path
+from types import ModuleType, SimpleNamespace
+from unittest.mock import Mock
+
+import numpy as np
+import pytest
+from PIL import Image
+
+from dreamverse.config import MODEL_REGISTRY
+from dreamverse.generation_inputs import GenerationAsset, GenerationInputs
+from dreamverse.generation_worker import VideoGenerationWorker
+from dreamverse.minimax_h3_generation import MiniMaxH3GenerationBackend
+from dreamverse.worker_ipc import UserStepPayload
+
+
+@pytest.fixture
+def fastvideo_api(monkeypatch):
+    """Use the actual lightweight API schema with only GPU execution replaced."""
+    schema_path = Path(__file__).resolve().parents[4] / "fastvideo/api/schema.py"
+    spec = importlib.util.spec_from_file_location("dreamverse_test_api_schema", schema_path)
+    assert spec is not None and spec.loader is not None
+    schema = importlib.util.module_from_spec(spec)
+    monkeypatch.setitem(sys.modules, spec.name, schema)
+    spec.loader.exec_module(schema)
+    package = ModuleType("fastvideo")
+    package.__path__ = []
+    package.VideoGenerator = SimpleNamespace(from_config=Mock())
+    monkeypatch.setitem(sys.modules, "fastvideo", package)
+    monkeypatch.setitem(sys.modules, "fastvideo.api", schema)
+    schema.MiniMaxH3Reference = lambda **kwargs: SimpleNamespace(**kwargs)
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.torch.cuda.synchronize", lambda: None)
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.torch.cuda.empty_cache", lambda: None)
+    return package.VideoGenerator.from_config
+
+
+class RecordingGenerator:
+    def __init__(self):
+        self.requests = []
+        self.images = []
+        self.closed = False
+
+    def shutdown(self):
+        self.closed = True
+
+    def generate(self, request):
+        self.requests.append(request)
+        self.images.append(tuple(None if image is None else np.asarray(image).copy()
+                                 for image in (request.inputs.pil_image, request.inputs.last_image)))
+        return SimpleNamespace(
+            frames=[np.full((2, 3, 3), 7, dtype=np.uint8), np.full((2, 3, 3), 29, dtype=np.uint8)],
+            audio=np.zeros((2, 16), dtype=np.float32),
+            audio_sample_rate=44100,
+            generation_time=0.1,
+        )
+
+
+def prepared_backend(monkeypatch):
+    backend = MiniMaxH3GenerationBackend(0)
+    backend.model_config = dict(MODEL_REGISTRY["full-h3"])
+    backend.generator = RecordingGenerator()
+    monkeypatch.setattr(backend, "_gpu_mem", lambda: "fake executor")
+    return backend
+
+
+def test_ipc_preserves_immutable_ordered_references():
+    inputs = GenerationInputs("ref2va", (
+        GenerationAsset("second", "video", "/assets/second.mp4", "reference"),
+        GenerationAsset("first", "image", "/assets/first.png", "reference"),
+    ))
+    payload = UserStepPayload("follow the references", 1, None, True, inputs)
+    restored = pickle.loads(pickle.dumps(payload))
+    assert restored == payload
+    assert [asset.asset_id for asset in restored.generation_inputs.references] == ["second", "first"]
+
+
+def test_worker_passes_conditioning_to_selected_backend():
+    inputs = GenerationInputs("t2va")
+    worker = VideoGenerationWorker(0)
+    worker.backend = Mock()
+    worker.generate_step("prompt", 1, None, True, inputs)
+    worker.backend.generate_step.assert_called_once_with("prompt", 1, None, True, generation_inputs=inputs)
+
+
+def test_full_h3_uses_full_weights_without_preview_lora(monkeypatch, fastvideo_api):
+    backend = prepared_backend(monkeypatch)
+    old_generator = backend.generator
+    fastvideo_api.return_value = RecordingGenerator()
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.DREAMVERSE_SP_SIZE", 4)
+    backend.initialize(MODEL_REGISTRY["full-h3"])
+    config = fastvideo_api.call_args.args[0]
+    assert old_generator.closed
+    assert config.pipeline.components.lora_path is None
+    assert config.pipeline.components.override_pipeline_cls_name is None
+    assert config.engine.use_fsdp_inference
+    assert config.engine.num_gpus == 4
+    assert not config.pipeline.experimental["inference_torch_compile"]
+
+
+def test_fl2va_maps_endpoints_only_on_initial_segment(monkeypatch, fastvideo_api, tmp_path):
+    first = tmp_path / "first.png"
+    last = tmp_path / "last.png"
+    Image.new("RGB", (3, 2), (10, 20, 30)).save(first)
+    Image.new("RGB", (3, 2), (40, 50, 60)).save(last)
+    inputs = GenerationInputs("fl2va", (
+        GenerationAsset("first", "image", str(first), "first_frame"),
+        GenerationAsset("last", "image", str(last), "last_frame"),
+    ))
+    backend = prepared_backend(monkeypatch)
+    first_result = backend.generate_step("first", 1, None, True, inputs)
+    later_result = backend.generate_step("later", 2, None, False, inputs)
+    assert backend.generator.images[0][0][0, 0].tolist() == [10, 20, 30]
+    assert backend.generator.images[0][1][0, 0].tolist() == [40, 50, 60]
+    assert backend.generator.images[1][0][0, 0].tolist() == [29, 29, 29]
+    assert backend.generator.images[1][1] is None
+    assert first_result.head_trim_frames == 0
+    assert later_result.head_trim_frames == 1
+    assert backend.generator.requests[0].sampling.num_inference_steps == 50
+
+
+def test_ref2va_switches_pipeline_and_preserves_reference_order(monkeypatch, fastvideo_api):
+    inputs = GenerationInputs("ref2va", (
+        GenerationAsset("video", "video", "/assets/reference.mp4", "reference"),
+        GenerationAsset("audio", "audio", "/assets/reference.wav", "reference"),
+        GenerationAsset("image", "image", "/assets/reference.png", "reference"),
+    ))
+    backend = prepared_backend(monkeypatch)
+    base_generator = backend.generator
+    reference_generator = RecordingGenerator()
+
+    def load(config):
+        assert base_generator.closed, "Old executor must release memory before loading reference weights"
+        assert config.pipeline.components.override_pipeline_cls_name == "MiniMaxH3Ref2VAModularPipeline"
+        assert config.pipeline.workload_type == "i2v"
+        assert config.pipeline.components.lora_path is None
+        return reference_generator
+
+    fastvideo_api.side_effect = load
+    backend.generate_step("first", 1, None, True, inputs)
+    result = backend.generate_step("second", 2, None, False, inputs)
+    assert fastvideo_api.call_count == 1
+    for request in reference_generator.requests:
+        assert [(reference.media_type, reference.source) for reference in request.inputs.references] == [
+            ("video", "/assets/reference.mp4"), ("audio", "/assets/reference.wav"), ("image", "/assets/reference.png")
+        ]
+        assert request.inputs.pil_image is None
+        assert request.inputs.last_image is None
+    assert result.head_trim_frames == result.head_trim_audio_frames == 0
+    assert backend.continuation_image is None
+
+    fastvideo_api.side_effect = None
+    fastvideo_api.return_value = RecordingGenerator()
+    backend.generate_step("new project", 1, None, True, GenerationInputs("t2va"))
+    assert reference_generator.closed
+    config = fastvideo_api.call_args.args[0]
+    assert config.pipeline.components.override_pipeline_cls_name is None
+    assert backend.pipeline_mode == "base"
+
+
+def test_ref2va_pipeline_switch_failure_drops_unloaded_executor(monkeypatch, fastvideo_api):
+    backend = prepared_backend(monkeypatch)
+    old_generator = backend.generator
+    fastvideo_api.side_effect = RuntimeError("checkpoint unavailable")
+    with pytest.raises(RuntimeError, match="checkpoint unavailable"):
+        backend.generate_step("prompt", 1, None, True, GenerationInputs("ref2va"))
+    assert old_generator.closed
+    assert backend.generator is None
+
+
+def test_failed_pipeline_switch_reloads_on_the_next_step(monkeypatch, fastvideo_api):
+    """A failed base<->ref2va switch must not strand the slot for later steps."""
+    backend = prepared_backend(monkeypatch)
+    fastvideo_api.side_effect = RuntimeError("checkpoint unavailable")
+    with pytest.raises(RuntimeError, match="checkpoint unavailable"):
+        backend.generate_step("prompt", 1, None, True, GenerationInputs("ref2va"))
+
+    fastvideo_api.side_effect = None
+    fastvideo_api.return_value = RecordingGenerator()
+    backend.generate_step("retry", 1, None, True, GenerationInputs("ref2va"))
+
+    assert fastvideo_api.call_count == 2
+    assert backend.pipeline_mode == "ref2va"
+    assert backend.generator is not None
+
+
+def test_mode_cannot_switch_mid_project(monkeypatch, fastvideo_api):
+    backend = prepared_backend(monkeypatch)
+    with pytest.raises(ValueError, match="middle of a project"):
+        backend.generate_step("prompt", 2, None, False, GenerationInputs("ref2va"))
+    fastvideo_api.assert_not_called()
+
+
+@pytest.mark.parametrize("mode", ["fl2va", "ref2va"])
+def test_preview_rejects_unsupported_generation_modes(monkeypatch, fastvideo_api, mode):
+    backend = prepared_backend(monkeypatch)
+    backend.model_config = dict(MODEL_REGISTRY["fast-h3"])
+    with pytest.raises(ValueError, match="full-h3"):
+        backend.generate_step("prompt", 1, None, True, GenerationInputs(mode))
+    assert backend.generator.requests == []
+
+
+def test_legacy_h3_continuation_is_preserved(monkeypatch, fastvideo_api):
+    backend = prepared_backend(monkeypatch)
+    backend.generate_step("first", 1, None, False)
+    result = backend.generate_step("second", 2, None, False)
+    assert backend.generator.requests[0].inputs.pil_image is None
+    assert backend.generator.images[1][0][0, 0].tolist() == [29, 29, 29]
+    assert result.head_trim_frames == 1

apps/dreamverse/dreamverse/tests/test_generation_session_inputs.py

@@ -0,0 +1,278 @@
+"""Session-mode validation and IPC handoff without a GPU worker process."""
+
+from __future__ import annotations
+
+import asyncio
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType, SimpleNamespace
+from unittest.mock import Mock
+
+import pytest
+
+from dreamverse.generation_inputs import GenerationAsset, GenerationInputs
+from dreamverse.worker_ipc import MediaChunk, MediaComplete, MediaInit
+
+
+@pytest.fixture
+def controller_module(monkeypatch):
+    gpu_pool = ModuleType("dreamverse.gpu_pool")
+    gpu_pool.GPUSlot = object
+    monkeypatch.setitem(sys.modules, "dreamverse.gpu_pool", gpu_pool)
+    path = Path(__file__).resolve().parents[1] / "session/controller.py"
+    spec = importlib.util.spec_from_file_location("dreamverse_test_session_controller", path)
+    assert spec is not None and spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    monkeypatch.setattr(module, "ACTIVE_MODEL_ID", "full-h3")
+    monkeypatch.setattr(module, "pin_generation_inputs", Mock())
+    monkeypatch.setattr(module, "release_generation_inputs", Mock())
+    return module
+
+
+class Socket:
+    def __init__(self):
+        self.incoming = asyncio.Queue()
+        self.outgoing = asyncio.Queue()
+        self.messages = []
+        self.closed = False
+
+    async def accept(self):
+        pass
+
+    async def receive_json(self):
+        return await self.incoming.get()
+
+    async def send_json(self, payload):
+        self.messages.append(payload)
+        await self.outgoing.put(payload)
+
+    async def send_bytes(self, payload):
+        pass
+
+    async def close(self, **kwargs):
+        self.closed = True
+
+    async def wait_for(self, kind):
+        while True:
+            payload = await asyncio.wait_for(self.outgoing.get(), 3)
+            if payload["type"] == kind:
+                return payload
+
+
+class Slot:
+    def __init__(self):
+        self.shared_stream_buffer = None
+        self.queue = asyncio.Queue()
+        self.calls = []
+
+    async def join_user(self, *args, **kwargs):
+        pass
+
+    def register_stream_queue(self, client_id):
+        return self.queue
+
+    def unregister_stream_queue(self, client_id):
+        pass
+
+    async def user_step(self, client_id, **kwargs):
+        self.calls.append(kwargs)
+        segment_idx = kwargs["segment_idx"]
+        await self.queue.put(MediaInit(client_id, segment_idx, "test", "video/mp4", False))
+        await self.queue.put(MediaChunk(client_id, segment_idx, "test", chunk=b"test"))
+        await self.queue.put(MediaComplete(client_id, segment_idx, "test", 1))
+        return {"e2e_latency_ms": 1.0}
+
+
+class Pool:
+    def __init__(self):
+        self.slot = Slot()
+        self.acquire_count = 0
+
+    def get_status(self):
+        return {"queue_size": 0, "available_gpus": 1, "total_gpus": 1}
+
+    async def acquire(self, *args):
+        self.acquire_count += 1
+        return 0, self.slot
+
+    async def release(self, *args):
+        pass
+
+
+def start_controller(module, socket, pool):
+    enhancer = SimpleNamespace(
+        resolve_rewrite_model=lambda value: "test-model",
+        resolve_rewrite_system_prompt=lambda value: "test-system",
+        resolve_rewrite_temperature=lambda value: 1.0,
+    )
+    controller = module.SessionController(socket, pool, enhancer, None, None)
+    return asyncio.create_task(controller.run())
+
+
+def test_invalid_initial_mode_does_not_acquire_gpu(controller_module):
+    async def scenario():
+        socket, pool = Socket(), Pool()
+        await socket.incoming.put({"type": "session_init_v2", "generation_mode": "unknown"})
+        await asyncio.wait_for(start_controller(controller_module, socket, pool), 3)
+        error = next(message for message in socket.messages if message["type"] == "error")
+        assert error["error_code"] == "invalid_generation_input"
+        assert pool.acquire_count == 0
+        assert socket.closed
+        controller_module.pin_generation_inputs.assert_not_called()
+
+    asyncio.run(scenario())
+
+
+def test_new_project_replaces_conditioning_and_passes_it_to_gpu(controller_module, monkeypatch):
+    first = GenerationInputs("t2va")
+    second = GenerationInputs("fl2va", (GenerationAsset("first", "image", "/assets/first.png", "first_frame"),))
+    monkeypatch.setattr(controller_module, "resolve_generation_inputs", Mock(side_effect=[first, second]))
+
+    async def scenario():
+        socket, pool = Socket(), Pool()
+        await socket.incoming.put({
+            "type": "session_init_v2", "generation_mode": "t2va", "curated_prompts": ["first prompt"],
+            "enhancement_enabled": False,
+        })
+        task = start_controller(controller_module, socket, pool)
+        try:
+            await socket.wait_for("media_segment_complete")
+            await socket.incoming.put({"type": "end_project_keep_session"})
+            await socket.wait_for("project_idle")
+            assert first in [call.args[0] for call in controller_module.release_generation_inputs.call_args_list]
+            await socket.incoming.put({
+                "type": "project_init_v1", "generation_mode": "fl2va", "curated_prompts": ["second prompt"],
+                "enhancement_enabled": False,
+            })
+            await socket.wait_for("media_segment_complete")
+            assert [call["generation_inputs"] for call in pool.slot.calls] == [first, second]
+            assert pool.slot.calls[1]["segment_idx"] == 1
+            assert pool.slot.calls[1]["reset_conditioning"]
+            await socket.incoming.put({"type": "leave"})
+            await asyncio.wait_for(task, 3)
+        finally:
+            if not task.done():
+                task.cancel()
+                await asyncio.gather(task, return_exceptions=True)
+        assert [call.args[0] for call in controller_module.pin_generation_inputs.call_args_list] == [first, second]
+        assert second in [call.args[0] for call in controller_module.release_generation_inputs.call_args_list]
+
+    asyncio.run(scenario())
+
+
+@pytest.mark.parametrize("injection", [
+    {"initial_image": {"data_url": "not allowed"}},
+    {"generation_mode": "ref2va"},
+    {"conditioning_assets": []},
+])
+def test_simple_generate_cannot_replace_locked_inputs(controller_module, injection):
+    async def scenario():
+        socket, pool = Socket(), Pool()
+        await socket.incoming.put({
+            "type": "session_init_v2", "generation_mode": "t2va", "single_clip_mode": True,
+            "enhancement_enabled": False,
+        })
+        task = start_controller(controller_module, socket, pool)
+        try:
+            await socket.wait_for("gpu_assigned")
+            await socket.incoming.put({"type": "simple_generate", "prompt": "prompt", **injection})
+            error = await socket.wait_for("error")
+            assert error["error_code"] == "invalid_generation_input"
+            assert "project" in error["message"]
+            assert pool.slot.calls == []
+            await socket.incoming.put({"type": "leave"})
+            await asyncio.wait_for(task, 3)
+        finally:
+            if not task.done():
+                task.cancel()
+                await asyncio.gather(task, return_exceptions=True)
+
+    asyncio.run(scenario())
+
+
+def test_disconnect_waits_for_worker_before_releasing_assets(controller_module, monkeypatch):
+    async def scenario():
+        socket, pool = Socket(), Pool()
+        worker_started = asyncio.Event()
+        worker_finished = asyncio.Event()
+        proceed = asyncio.Event()
+
+        async def slow_step(client_id, **kwargs):
+            worker_started.set()
+            await proceed.wait()
+            worker_finished.set()
+            return {"e2e_latency_ms": 1.0}
+
+        pool.slot.user_step = slow_step
+        await socket.incoming.put({
+            "type": "session_init_v2", "generation_mode": "t2va", "curated_prompts": ["prompt"],
+            "enhancement_enabled": False,
+        })
+        task = start_controller(controller_module, socket, pool)
+        try:
+            await asyncio.wait_for(worker_started.wait(), 3)
+            await socket.incoming.put({"type": "leave"})
+            await asyncio.sleep(0.07)
+            assert not task.done()
+            controller_module.release_generation_inputs.assert_not_called()
+            proceed.set()
+            await asyncio.wait_for(task, 3)
+            assert worker_finished.is_set()
+            controller_module.release_generation_inputs.assert_called_once()
+        finally:
+            if not task.done():
+                task.cancel()
+                await asyncio.gather(task, return_exceptions=True)
+
+    asyncio.run(scenario())
+
+
+@pytest.fixture
+def gpu_pool_module(monkeypatch):
+    streaming = ModuleType("dreamverse.av_streaming")
+    for name in ("StreamChunk", "StreamComplete", "StreamEvent", "StreamInit", "generate_stream_id", "stream_fmp4"):
+        setattr(streaming, name, object)
+    streaming.SHARED_STREAM_BUFFER_BYTES = 1024
+    streaming.USE_SHARED_STREAM_BUFFER = False
+    monkeypatch.setitem(sys.modules, "dreamverse.av_streaming", streaming)
+    path = Path(__file__).resolve().parents[1] / "gpu_pool.py"
+    spec = importlib.util.spec_from_file_location("dreamverse_test_gpu_pool", path)
+    assert spec is not None and spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    monkeypatch.setitem(sys.modules, spec.name, module)
+    spec.loader.exec_module(module)
+    monkeypatch.setattr(module, "pin_generation_inputs", Mock())
+    monkeypatch.setattr(module, "release_generation_inputs", Mock())
+    return module
+
+
+def test_gpu_step_timeout_keeps_assets_pinned_until_late_worker_completion(gpu_pool_module):
+    from dreamverse.worker_ipc import StepComplete
+
+    async def scenario():
+        slot = gpu_pool_module.GPUSlot(0, "0")
+        inputs = GenerationInputs("ref2va", (GenerationAsset("ref", "image", "/assets/ref.png", "reference"),))
+
+        async def timeout(command, timeout):
+            assert command.payload.generation_inputs == inputs
+            raise asyncio.TimeoutError
+
+        slot._send_command_tagged = timeout
+        with pytest.raises(asyncio.TimeoutError):
+            await slot.user_step("user", "prompt", generation_inputs=inputs)
+        gpu_pool_module.pin_generation_inputs.assert_called_once_with(inputs)
+        gpu_pool_module.release_generation_inputs.assert_not_called()
+
+        def late_response(timeout):
+            slot._active = False
+            return StepComplete("user", 1, {})
+
+        slot.response_queue = SimpleNamespace(get=late_response)
+        slot._active = True
+        await slot._response_reader()
+        gpu_pool_module.release_generation_inputs.assert_called_once_with(inputs)
+        assert slot._step_asset_inputs == {}
+
+    asyncio.run(scenario())

apps/dreamverse/dreamverse/tests/test_generation_worker.py

@@ -0,0 +1,18 @@
+from dreamverse.generation_worker import _create_generation_backend
+from dreamverse.ltx2_generation import LTX2GenerationBackend
+
+
+def test_create_generation_backend_ltx2_module_import():
+    backend = _create_generation_backend("ltx2", gpu_id=3)
+
+    assert isinstance(backend, LTX2GenerationBackend)
+    assert backend.gpu_id == 3
+
+
+def test_create_generation_backend_cosmos25_dfd_module_import():
+    from dreamverse.cosmos25_dfd_generation import Cosmos25DFDGenerationBackend
+
+    backend = _create_generation_backend("cosmos25_dfd", gpu_id=2)
+
+    assert isinstance(backend, Cosmos25DFDGenerationBackend)
+    assert backend.gpu_id == 2

apps/dreamverse/dreamverse/tests/test_gpu_pool.py

@@ -7,7 +7,6 @@ from types import SimpleNamespace
 
 import pytest
 
-
 import dreamverse.gpu_pool as gpu_pool
 
 
@@ -64,6 +63,14 @@ def test_get_available_gpus_defaults_to_first_visible_device(monkeypatch):
     assert gpu_pool.get_available_gpus() == [3]
 
 
+def test_get_available_gpus_defaults_to_active_model_sequence_parallel_size(monkeypatch):
+    monkeypatch.setenv("CUDA_VISIBLE_DEVICES", "0,1,2,3,4")
+    monkeypatch.delenv("FASTVIDEO_GPU_COUNT", raising=False)
+    monkeypatch.setattr(gpu_pool, "DREAMVERSE_SP_SIZE", 4)
+
+    assert gpu_pool.get_available_gpus() == [0, 1, 2, 3]
+
+
 def test_get_available_gpus_rejects_invalid_gpu_count(monkeypatch):
     monkeypatch.delenv("CUDA_VISIBLE_DEVICES", raising=False)
     monkeypatch.setenv("FASTVIDEO_GPU_COUNT", "zero")
@@ -72,6 +79,23 @@ def test_get_available_gpus_rejects_invalid_gpu_count(monkeypatch):
         gpu_pool.get_available_gpus()
 
 
+def test_join_user_failed_reload_marks_model_uninitialized(monkeypatch):
+    """A failed model reload forces the next join to reload a model."""
+    slot = gpu_pool.GPUSlot(gpu_id=0, cuda_device="0")
+    slot.current_model_id = "fast-ltx2"
+
+    async def fake_send_command(command, timeout):
+        del command, timeout
+        return gpu_pool.WorkerError(user_id="__reload__", message="load failed")
+
+    monkeypatch.setattr(slot, "_send_command", fake_send_command)
+
+    with pytest.raises(RuntimeError, match="Model reload failed"):
+        asyncio.run(slot.join_user("client-id", model_id="fast-h3"))
+
+    assert slot.current_model_id is None
+
+
 def test_send_command_raises_on_worker_death():
     """A worker that consumes a command and exits without replying must
     surface as RuntimeError via sentinel detection, not after the long
@@ -85,9 +109,7 @@ def test_send_command_raises_on_worker_death():
     cmd_q = ctx.Queue()
     resp_q = ctx.Queue()
 
-    proc = ctx.Process(
-        target=_child_consume_and_exit, args=(cmd_q, resp_q)
-    )
+    proc = ctx.Process(target=_child_consume_and_exit, args=(cmd_q, resp_q))
     proc.start()
 
     # Wait for the spawn child to fully boot.  Allow generous time —
@@ -95,9 +117,9 @@ def test_send_command_raises_on_worker_death():
     ready = resp_q.get(timeout=30.0)
     assert ready == "READY"
 
-    async def runner():
+    async def runner() -> None:
         slot = gpu_pool.GPUSlot(gpu_id=0, cuda_device="0")
-        slot.process = proc
+        slot.process = proc  # type: ignore[assignment]
         slot.command_queue = cmd_q
         slot.response_queue = resp_q
 

apps/dreamverse/dreamverse/tests/test_import_contract.py

@@ -1,4 +1,6 @@
 import ast
+import subprocess
+import sys
 from pathlib import Path
 
 ALLOWED_PREFIXES = (
@@ -7,7 +9,7 @@ ALLOWED_PREFIXES = (
     "fastvideo.entrypoints.video_generator",
     "fastvideo.configs",
 )
-ALLOWED_EXACT = ("fastvideo",)
+ALLOWED_EXACT = ("fastvideo", )
 FORBIDDEN_PREFIXES = (
     "fastvideo.pipelines",
     "fastvideo.models",
@@ -17,11 +19,11 @@ FORBIDDEN_PREFIXES = (
 )
 ALLOWED_INTERNAL_IMPORTS = {
     (
-        "video_generation.py",
+        "ltx2_generation.py",
         "fastvideo.models.audio.ltx2_audio_processing",
     ),
     (
-        "video_generation.py",
+        "ltx2_generation.py",
         "fastvideo.models.loader.component_loader",
     ),
 }
@@ -38,19 +40,49 @@ def test_dreamverse_server_imports_only_public_fastvideo_surfaces() -> None:
         except SyntaxError as task_exc:
             raise AssertionError(f"Failed to parse {path}") from task_exc
         for node in ast.walk(tree):
-            names = (
-                [a.name for a in node.names] if isinstance(node, ast.Import)
-                else [node.module] if isinstance(node, ast.ImportFrom) and node.module
-                else []
-            )
+            names = ([a.name for a in node.names] if isinstance(node, ast.Import) else
+                     [node.module] if isinstance(node, ast.ImportFrom) and node.module else [])
             for name in names:
                 if not name:
                     continue
                 rel_path = str(path.relative_to(root))
-                if (
-                    name.startswith(FORBIDDEN_PREFIXES)
-                    and (rel_path, name) not in ALLOWED_INTERNAL_IMPORTS
-                ):
+                if (name.startswith(FORBIDDEN_PREFIXES) and (rel_path, name) not in ALLOWED_INTERNAL_IMPORTS):
                     bad.append((str(path.relative_to(root)), getattr(node, "lineno", 0), name))
 
     assert bad == [], f"Forbidden internal imports: {bad}"
+
+
+def test_h3_reference_public_export_is_lazy_and_preserves_type_identity() -> None:
+    """Only explicit reference usage should load H3's optional GPU dependencies."""
+    repo_root = Path(__file__).resolve().parents[4]
+    # Isolate the import graph: keep the real public API implementation/schema,
+    # substituting only the unrelated legacy sampling module and heavy H3 leaf.
+    script = r'''
+import importlib
+import sys
+from pathlib import Path
+from types import ModuleType
+
+root = Path(sys.argv[1])
+fastvideo = ModuleType("fastvideo")
+fastvideo.__path__ = [str(root / "fastvideo")]
+sys.modules["fastvideo"] = fastvideo
+sampling = ModuleType("fastvideo.api.sampling_param")
+sampling.SamplingParam = type("SamplingParam", (), {})
+sys.modules[sampling.__name__] = sampling
+
+api = importlib.import_module("fastvideo.api")
+assert "MiniMaxH3Reference" in api.__all__
+assert "MiniMaxH3Reference" not in vars(api)
+assert not any(name.startswith("fastvideo.pipelines") for name in sys.modules)
+
+internal = ModuleType("fastvideo.pipelines.basic.minimax_h3.reference")
+internal.MiniMaxH3Reference = type("MiniMaxH3Reference", (), {})
+sys.modules[internal.__name__] = internal
+from fastvideo.api import MiniMaxH3Reference
+assert MiniMaxH3Reference is internal.MiniMaxH3Reference
+assert api.MiniMaxH3Reference is internal.MiniMaxH3Reference
+assert not hasattr(api, "UnknownReference")
+'''
+    result = subprocess.run([sys.executable, "-c", script, str(repo_root)], capture_output=True, text=True, timeout=30)
+    assert result.returncode == 0, result.stderr

apps/dreamverse/dreamverse/tests/test_minimax_h3_generation.py

@@ -0,0 +1,247 @@
+from __future__ import annotations
+
+import os
+from types import SimpleNamespace
+from typing import Any
+
+import numpy as np
+import pytest
+
+import dreamverse.generation_worker as generation_worker
+from dreamverse.minimax_h3_generation import MiniMaxH3GenerationBackend
+
+
+FASTH3_MODEL_CONFIG = {
+    "name": "FastH3",
+    "generation_backend": "minimax_h3",
+    "default_sp_size": 4,
+    "model_path": "MiniMaxAI/MiniMax-H3",
+    "adapter_repo": "FastVideo/FastVideo-FastH3-4-step-Preview-v1-LoRA",
+    "adapter_filename": "vsa-datafree/adapter_model.safetensors",
+    "attention_backend": "VIDEO_SPARSE_ATTN_H3",
+    "height": 768,
+    "width": 1344,
+    "num_frames": 124,
+    "num_inference_steps": 5,
+    "seed": 1000,
+}
+
+
+class _RecordingGenerator:
+    """Record typed requests and return small synchronized media fixtures."""
+
+    def __init__(self) -> None:
+        self.requests: list[Any] = []
+        self.conditioning_pixels: list[np.ndarray | None] = []
+
+    def generate(self, request):
+        """Capture the request and return two tiny video frames with audio."""
+        self.requests.append(request)
+        conditioning_image = request.inputs.pil_image
+        self.conditioning_pixels.append(
+            None if conditioning_image is None else np.asarray(conditioning_image).copy())
+        frames = [
+            np.full((2, 3, 3), 10, dtype=np.uint8),
+            np.full((2, 3, 3), 20, dtype=np.uint8),
+        ]
+        return SimpleNamespace(
+            frames=frames,
+            audio=np.zeros((2, 16), dtype=np.float32),
+            audio_sample_rate=44100,
+            generation_time=0.25,
+        )
+
+
+def test_initialize_builds_vsa_datafree_fasth3_generator(monkeypatch):
+    """Initialization translates the DreamVerse profile into typed FastVideo config."""
+    from fastvideo import VideoGenerator
+
+    captured = {}
+    fake_generator = SimpleNamespace(shutdown=lambda: None)
+
+    def fake_from_config(config):
+        captured["config"] = config
+        return fake_generator
+
+    def fake_download(**kwargs):
+        captured["download"] = kwargs
+        return f"/models/{kwargs['filename']}"
+
+    monkeypatch.setattr("huggingface_hub.hf_hub_download", fake_download)
+    monkeypatch.setattr(VideoGenerator, "from_config", fake_from_config)
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.DREAMVERSE_SP_SIZE", 4)
+    monkeypatch.setenv("FASTVIDEO_ATTENTION_BACKEND", "test-attention")
+    monkeypatch.setenv("FASTVIDEO_FA4", "0")
+    monkeypatch.setenv("FASTVIDEO_MINIMAX_H3_FUSIONS", "0")
+    monkeypatch.setenv("FASTVIDEO_VSA_SM100A", "1")
+    monkeypatch.setenv("FASTVIDEO_INFERENCE_TORCH_COMPILE", "1")
+
+    backend = MiniMaxH3GenerationBackend(gpu_id=0)
+    monkeypatch.setattr(backend, "_gpu_mem", lambda: "alloc=0.00GiB, reserved=0.00GiB")
+    backend.initialize(FASTH3_MODEL_CONFIG)
+
+    config = captured["config"]
+    assert captured["download"] == {
+        "repo_id": "FastVideo/FastVideo-FastH3-4-step-Preview-v1-LoRA",
+        "filename": "vsa-datafree/adapter_model.safetensors",
+    }
+    assert config.model_path == "MiniMaxAI/MiniMax-H3"
+    assert config.pipeline.components.lora_path.endswith("vsa-datafree/adapter_model.safetensors")
+    assert config.pipeline.components.lora_strength == 1.0
+    assert config.pipeline.experimental == {
+        "attention_backend": "VIDEO_SPARSE_ATTN_H3",
+        "inference_torch_compile": False,
+        "vae_parallel_decode": True,
+        "vae_parallel_decode_strategy": "gather",
+        "VSA_sparsity": 0.9,
+        "VSA_tile_size": 64,
+    }
+    assert config.engine.num_gpus == 4
+    assert config.engine.parallelism.tp_size == 1
+    assert config.engine.parallelism.sp_size == 4
+    assert config.engine.offload.dit is False
+    assert config.engine.offload.dit_layerwise is False
+    assert config.engine.offload.text_encoder is True
+    assert config.engine.offload.vae is True
+    assert config.engine.compile.vae_enabled is True
+    assert config.engine.use_fsdp_inference is False
+    assert os.environ["FASTVIDEO_ATTENTION_BACKEND"] == "VIDEO_SPARSE_ATTN_H3"
+    assert os.environ["FASTVIDEO_FA4"] == "1"
+    assert os.environ["FASTVIDEO_MINIMAX_H3_FUSIONS"] == "all"
+    assert os.environ["FASTVIDEO_VSA_SM100A"] == "0"
+    assert "FASTVIDEO_INFERENCE_TORCH_COMPILE" not in os.environ
+
+
+def test_initialize_selects_declared_generation_backend(monkeypatch):
+    """The GPU worker constructs the backend that the active model profile declares."""
+    from unittest.mock import Mock
+
+    selected_backend = Mock()
+    monkeypatch.setattr(
+        generation_worker,
+        "_create_generation_backend",
+        lambda backend_name, gpu_id: selected_backend,
+    )
+    worker = generation_worker.VideoGenerationWorker(gpu_id=3)
+
+    worker.initialize(FASTH3_MODEL_CONFIG)
+
+    assert worker.backend_name == "minimax_h3"
+    assert worker.backend is selected_backend
+    selected_backend.initialize.assert_called_once_with(FASTH3_MODEL_CONFIG)
+
+
+def test_initialize_failure_clears_backend_ownership(monkeypatch):
+    """A failed family change leaves the GPU worker explicitly uninitialized."""
+    ltx_backend = SimpleNamespace(initialize=lambda config: None, shutdown=lambda: None)
+
+    def fail_initialize(config):
+        del config
+        raise RuntimeError("load failed")
+
+    fasth3_backend = SimpleNamespace(
+        initialize=fail_initialize,
+        shutdown=lambda: None,
+    )
+    backends = {
+        "ltx2": ltx_backend,
+        "minimax_h3": fasth3_backend,
+    }
+    monkeypatch.setattr(
+        generation_worker,
+        "_create_generation_backend",
+        lambda backend_name, gpu_id: backends[backend_name],
+    )
+    worker = generation_worker.VideoGenerationWorker(gpu_id=3)
+    worker.initialize({"generation_backend": "ltx2"})
+
+    with pytest.raises(RuntimeError, match="load failed"):
+        worker.initialize(FASTH3_MODEL_CONFIG)
+
+    assert worker.backend is None
+    assert worker.backend_name is None
+    assert worker.model_config == {"generation_backend": "ltx2"}
+
+
+def test_generate_step_uses_last_frame_for_continuation(monkeypatch):
+    """A later segment receives the prior segment's last decoded frame."""
+    backend = MiniMaxH3GenerationBackend(gpu_id=0)
+    backend.model_config = dict(FASTH3_MODEL_CONFIG)
+    backend.generator = _RecordingGenerator()
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.torch.cuda.synchronize", lambda: None)
+
+    first_result = backend.generate_step(
+        "first prompt",
+        segment_idx=1,
+        image_path=None,
+        reset_conditioning=True,
+    )
+    second_result = backend.generate_step(
+        "second prompt",
+        segment_idx=2,
+        image_path=None,
+        reset_conditioning=False,
+    )
+
+    first_request = backend.generator.requests[0]
+    assert first_request.inputs.pil_image is None
+    assert first_request.negative_prompt == ""
+    assert first_request.sampling.height == 768
+    assert first_request.sampling.width == 1344
+    assert first_request.sampling.num_frames == 124
+    assert first_request.sampling.num_inference_steps == 5
+    assert first_request.sampling.fps == 24
+    assert first_request.sampling.guidance_scale == 1.0
+    assert first_request.sampling.batch_cfg is False
+    assert first_request.sampling.seed == 1000
+    assert first_request.output.save_video is False
+    assert first_request.output.return_frames is True
+    assert backend.generator.conditioning_pixels[1].tolist() == np.full((2, 3, 3), 20).tolist()
+    assert first_result.head_trim_frames == 0
+    assert first_result.head_trim_audio_frames == 0
+    assert second_result.head_trim_frames == 1
+    assert second_result.head_trim_audio_frames == 1
+    assert second_result.audio_sample_rate == 44100
+
+
+def test_generate_step_reset_uses_text_to_video_path(monkeypatch):
+    """Resetting continuation produces an unconditioned text-to-video request."""
+    backend = MiniMaxH3GenerationBackend(gpu_id=0)
+    backend.model_config = dict(FASTH3_MODEL_CONFIG)
+    backend.generator = _RecordingGenerator()
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.torch.cuda.synchronize", lambda: None)
+
+    backend.generate_step("first prompt", 1, None, True)
+    reset_result = backend.generate_step("reset prompt", 2, None, True)
+
+    assert backend.generator.requests[-1].inputs.pil_image is None
+    assert reset_result.head_trim_frames == 0
+    assert reset_result.head_trim_audio_frames == 0
+
+
+def test_generate_step_missing_continuation_frame(monkeypatch):
+    """A later segment fails when no reset or retained frame defines its input."""
+    backend = MiniMaxH3GenerationBackend(gpu_id=0)
+    backend.model_config = dict(FASTH3_MODEL_CONFIG)
+    backend.generator = _RecordingGenerator()
+
+    with pytest.raises(RuntimeError, match="requires a retained continuation frame"):
+        backend.generate_step("later prompt", 2, None, False)
+
+    assert backend.generator.requests == []
+
+
+def test_warmup_exercises_text_and_first_frame_paths(monkeypatch):
+    """Warmup covers both request shapes used by a DreamVerse session."""
+    backend = MiniMaxH3GenerationBackend(gpu_id=0)
+    backend.model_config = dict(FASTH3_MODEL_CONFIG)
+    backend.generator = _RecordingGenerator()
+    monkeypatch.setattr("dreamverse.minimax_h3_generation.torch.cuda.synchronize", lambda: None)
+
+    timings = backend.warmup("warmup prompt")
+
+    assert backend.generator.conditioning_pixels[0] is None
+    assert backend.generator.conditioning_pixels[1] is not None
+    assert backend.continuation_image is None
+    assert "warmup_text_to_video_ms" in timings
+    assert "warmup_first_frame_to_video_ms" in timings

apps/dreamverse/dreamverse/tests/test_mock_server.py

@@ -6,7 +6,6 @@ import os
 
 from fastapi import WebSocketDisconnect
 
-
 os.environ.setdefault("CEREBRAS_API_KEY", "dummy")
 os.environ.setdefault("GROQ_API_KEY", "dummy")
 
@@ -14,6 +13,7 @@ import dreamverse.mock_server as mock_server
 
 
 class _FakeWebSocket:
+
     def __init__(self, messages: list[tuple[float, dict[str, object]]]):
         self._messages = messages
         self._index = 0
@@ -49,34 +49,34 @@ def test_mock_server_matches_current_single5s_protocol():
         mock_server.MOCK_SEGMENT_BYTES = b"mock-fmp4-bytes"
         mock_server.LATENCY_MS = 1
 
-        ws = _FakeWebSocket(
-            [
-                (
-                    0.0,
-                    {
-                        "type": "session_init_v2",
-                        "preset_id": "simple_prompt_1",
-                        "curated_prompts": ["selected prompt"],
-                        "single_clip_mode": True,
-                        "enhancement_enabled": False,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (
-                    0.01,
-                    {
-                        "type": "simple_generate",
-                        "preset_id": "simple_custom_prompt",
-                        "prompt_id": "simple_custom_prompt",
-                        "prompt": "custom prompt",
-                        "enhancement_enabled": True,
-                        "initial_image": None,
-                    },
-                ),
-                (0.20, {"type": "leave"}),
-            ]
-        )
+        ws = _FakeWebSocket([
+            (
+                0.0,
+                {
+                    "type": "session_init_v2",
+                    "preset_id": "simple_prompt_1",
+                    "curated_prompts": ["selected prompt"],
+                    "single_clip_mode": True,
+                    "enhancement_enabled": False,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (
+                0.01,
+                {
+                    "type": "simple_generate",
+                    "preset_id": "simple_custom_prompt",
+                    "prompt_id": "simple_custom_prompt",
+                    "prompt": "custom prompt",
+                    "enhancement_enabled": True,
+                    "initial_image": None,
+                },
+            ),
+            (0.20, {
+                "type": "leave"
+            }),
+        ])
 
         asyncio.run(mock_server.websocket_endpoint(ws))
 
@@ -92,24 +92,14 @@ def test_mock_server_matches_current_single5s_protocol():
         assert message_types.count("ltx2_stream_complete") == 2
         assert "prompt_sources_blocked" not in message_types
 
-        segment_start_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "ltx2_segment_start"
-        ]
-        gpu_assigned_event = next(
-            payload for payload in ws.sent_json if payload["type"] == "gpu_assigned"
-        )
+        segment_start_events = [payload for payload in ws.sent_json if payload["type"] == "ltx2_segment_start"]
+        gpu_assigned_event = next(payload for payload in ws.sent_json if payload["type"] == "gpu_assigned")
         assert gpu_assigned_event["session_timeout"] == mock_server.SESSION_TIMEOUT_SECONDS
         assert [payload["segment_idx"] for payload in segment_start_events] == [1, 1]
         assert segment_start_events[0]["prompt"] == "selected prompt"
         assert segment_start_events[1]["prompt"] == "custom prompt"
 
-        step_complete_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "step_complete"
-        ]
+        step_complete_events = [payload for payload in ws.sent_json if payload["type"] == "step_complete"]
         assert len(step_complete_events) == 2
         assert step_complete_events[0]["latency_ms"] == {
             "total": 121.0,
@@ -134,29 +124,29 @@ def test_mock_server_regular_cap_waits_for_rewrite_rollout():
         mock_server.LATENCY_MS = 1
         mock_server.GENERATION_SEGMENT_CAP = 1
 
-        ws = _FakeWebSocket(
-            [
-                (
-                    0.0,
-                    {
-                        "type": "session_init_v2",
-                        "preset_id": "test_preset",
-                        "curated_prompts": ["segment one"],
-                        "enhancement_enabled": True,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (
-                    0.02,
-                    {
-                        "type": "rewrite_seed_prompts",
-                        "rewrite_instruction": "start a new rollout",
-                    },
-                ),
-                (0.20, {"type": "leave"}),
-            ]
-        )
+        ws = _FakeWebSocket([
+            (
+                0.0,
+                {
+                    "type": "session_init_v2",
+                    "preset_id": "test_preset",
+                    "curated_prompts": ["segment one"],
+                    "enhancement_enabled": True,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (
+                0.02,
+                {
+                    "type": "rewrite_seed_prompts",
+                    "rewrite_instruction": "start a new rollout",
+                },
+            ),
+            (0.20, {
+                "type": "leave"
+            }),
+        ])
 
         asyncio.run(mock_server.websocket_endpoint(ws))
 
@@ -166,11 +156,7 @@ def test_mock_server_regular_cap_waits_for_rewrite_rollout():
         assert "generation_cap_reached" not in message_types
         assert "prompt_sources_blocked" not in message_types
 
-        segment_start_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "ltx2_segment_start"
-        ]
+        segment_start_events = [payload for payload in ws.sent_json if payload["type"] == "ltx2_segment_start"]
         assert [payload["segment_idx"] for payload in segment_start_events] == [1, 1]
         assert segment_start_events[0]["prompt"] == "segment one"
         assert segment_start_events[1]["prompt"] == "segment one [start a new rollout]"
@@ -187,54 +173,40 @@ def test_mock_server_rewrite_during_active_segment_restarts_from_first_rewritten
         mock_server.MOCK_SEGMENT_BYTES = b"mock-fmp4-bytes"
         mock_server.LATENCY_MS = 100
 
-        ws = _FakeWebSocket(
-            [
-                (
-                    0.0,
-                    {
-                        "type": "session_init_v2",
-                        "preset_id": "test_preset",
-                        "curated_prompts": ["segment one", "segment two"],
-                        "enhancement_enabled": True,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (
-                    0.02,
-                    {
-                        "type": "rewrite_seed_prompts",
-                        "rewrite_instruction": "restart from rewrite",
-                    },
-                ),
-                (0.40, {"type": "leave"}),
-            ]
-        )
+        ws = _FakeWebSocket([
+            (
+                0.0,
+                {
+                    "type": "session_init_v2",
+                    "preset_id": "test_preset",
+                    "curated_prompts": ["segment one", "segment two"],
+                    "enhancement_enabled": True,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (
+                0.02,
+                {
+                    "type": "rewrite_seed_prompts",
+                    "rewrite_instruction": "restart from rewrite",
+                },
+            ),
+            (0.40, {
+                "type": "leave"
+            }),
+        ])
 
         asyncio.run(mock_server.websocket_endpoint(ws))
 
-        segment_start_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "ltx2_segment_start"
-        ]
+        segment_start_events = [payload for payload in ws.sent_json if payload["type"] == "ltx2_segment_start"]
         assert [payload["prompt"] for payload in segment_start_events[:2]] == [
             "segment one",
             "segment one [restart from rewrite]",
         ]
-        assert all(
-            payload["prompt"] != "segment two"
-            for payload in segment_start_events[1:]
-        )
-        reset_events = [
-            payload
-            for payload in ws.sent_json
-            if payload.get("type") == "seed_prompts_reset_applied"
-        ]
-        assert any(
-            payload.get("reason") == "rewrite_during_generation"
-            for payload in reset_events
-        )
+        assert all(payload["prompt"] != "segment two" for payload in segment_start_events[1:])
+        reset_events = [payload for payload in ws.sent_json if payload.get("type") == "seed_prompts_reset_applied"]
+        assert any(payload.get("reason") == "rewrite_during_generation" for payload in reset_events)
     finally:
         mock_server.MOCK_SEGMENT_BYTES = old_segment_bytes
         mock_server.LATENCY_MS = old_latency_ms
@@ -247,24 +219,24 @@ def test_mock_server_supports_initial_custom_rollout_prompt():
         mock_server.MOCK_SEGMENT_BYTES = b"mock-fmp4-bytes"
         mock_server.LATENCY_MS = 1
 
-        ws = _FakeWebSocket(
-            [
-                (
-                    0.0,
-                    {
-                        "type": "session_init_v2",
-                        "preset_id": "custom_editable",
-                        "preset_label": "Custom rollout",
-                        "curated_prompts": [],
-                        "initial_rollout_prompt": "A moonbase corridor thriller with flooding",
-                        "enhancement_enabled": True,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (0.20, {"type": "leave"}),
-            ]
-        )
+        ws = _FakeWebSocket([
+            (
+                0.0,
+                {
+                    "type": "session_init_v2",
+                    "preset_id": "custom_editable",
+                    "preset_label": "Custom rollout",
+                    "curated_prompts": [],
+                    "initial_rollout_prompt": "A moonbase corridor thriller with flooding",
+                    "enhancement_enabled": True,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (0.20, {
+                "type": "leave"
+            }),
+        ])
 
         asyncio.run(mock_server.websocket_endpoint(ws))
 
@@ -276,15 +248,9 @@ def test_mock_server_supports_initial_custom_rollout_prompt():
         assert "ltx2_stream_start" in message_types
         assert "prompt_sources_blocked" not in message_types
 
-        segment_start_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "ltx2_segment_start"
-        ]
+        segment_start_events = [payload for payload in ws.sent_json if payload["type"] == "ltx2_segment_start"]
         assert segment_start_events
-        assert segment_start_events[0]["prompt"] == (
-            "A moonbase corridor thriller with flooding [segment 1]"
-        )
+        assert segment_start_events[0]["prompt"] == ("A moonbase corridor thriller with flooding [segment 1]")
     finally:
         mock_server.MOCK_SEGMENT_BYTES = old_segment_bytes
         mock_server.LATENCY_MS = old_latency_ms
@@ -297,35 +263,37 @@ def test_mock_server_can_start_new_project_without_reconnecting():
         mock_server.MOCK_SEGMENT_BYTES = b"mock-fmp4-bytes"
         mock_server.LATENCY_MS = 40
 
-        ws = _FakeWebSocket(
-            [
-                (
-                    0.0,
-                    {
-                        "type": "session_init_v2",
-                        "preset_id": "test_preset",
-                        "curated_prompts": ["segment one"],
-                        "enhancement_enabled": True,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (0.02, {"type": "end_project_keep_session"}),
-                (
-                    0.20,
-                    {
-                        "type": "project_init_v1",
-                        "preset_id": "test_preset_2",
-                        "preset_label": "Test Preset 2",
-                        "curated_prompts": ["segment two"],
-                        "enhancement_enabled": True,
-                        "auto_extension_enabled": False,
-                        "loop_generation_enabled": False,
-                    },
-                ),
-                (0.40, {"type": "leave"}),
-            ]
-        )
+        ws = _FakeWebSocket([
+            (
+                0.0,
+                {
+                    "type": "session_init_v2",
+                    "preset_id": "test_preset",
+                    "curated_prompts": ["segment one"],
+                    "enhancement_enabled": True,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (0.02, {
+                "type": "end_project_keep_session"
+            }),
+            (
+                0.20,
+                {
+                    "type": "project_init_v1",
+                    "preset_id": "test_preset_2",
+                    "preset_label": "Test Preset 2",
+                    "curated_prompts": ["segment two"],
+                    "enhancement_enabled": True,
+                    "auto_extension_enabled": False,
+                    "loop_generation_enabled": False,
+                },
+            ),
+            (0.40, {
+                "type": "leave"
+            }),
+        ])
 
         asyncio.run(mock_server.websocket_endpoint(ws))
 
@@ -336,16 +304,11 @@ def test_mock_server_can_start_new_project_without_reconnecting():
 
         project_idle_index = message_types.index("project_idle")
         stream_start_indexes = [
-            index for index, message_type in enumerate(message_types)
-            if message_type == "ltx2_stream_start"
+            index for index, message_type in enumerate(message_types) if message_type == "ltx2_stream_start"
         ]
         assert stream_start_indexes[0] < project_idle_index < stream_start_indexes[1]
 
-        segment_start_events = [
-            payload
-            for payload in ws.sent_json
-            if payload["type"] == "ltx2_segment_start"
-        ]
+        segment_start_events = [payload for payload in ws.sent_json if payload["type"] == "ltx2_segment_start"]
         assert [payload["prompt"] for payload in segment_start_events[:2]] == [
             "segment one",
             "segment two",

apps/dreamverse/dreamverse/tests/test_prompt_enhancer.py

@@ -6,7 +6,6 @@ import os
 import re
 import time
 
-
 os.environ.setdefault("CEREBRAS_API_KEY", "dummy")
 os.environ.setdefault("GROQ_API_KEY", "dummy")
 
@@ -22,6 +21,7 @@ from dreamverse.prompt_enhancer import (
 
 
 class _FakeResponse:
+
     def __init__(self, payload: dict):
         self._payload = payload
 
@@ -30,6 +30,7 @@ class _FakeResponse:
 
 
 class _FakeSyncCompletions:
+
     def __init__(self, payload: dict):
         self._payload = payload
 
@@ -38,6 +39,7 @@ class _FakeSyncCompletions:
 
 
 class _FakeSyncClient:
+
     def __init__(self, payload: dict):
         self.chat = type(
             "_FakeChat",
@@ -47,6 +49,7 @@ class _FakeSyncClient:
 
 
 class _DelayedSyncCompletions:
+
     def __init__(self, payload: dict, delay_s: float = 0.0, exc: Exception | None = None):
         self._payload = payload
         self._delay_s = delay_s
@@ -61,29 +64,26 @@ class _DelayedSyncCompletions:
 
 
 class _DelayedSyncClient:
+
     def __init__(self, payload: dict, delay_s: float = 0.0, exc: Exception | None = None):
         self.chat = type(
             "_FakeChat",
             (),
-            {
-                "completions": _DelayedSyncCompletions(
-                    payload,
-                    delay_s=delay_s,
-                    exc=exc,
-                )
-            },
+            {"completions": _DelayedSyncCompletions(
+                payload,
+                delay_s=delay_s,
+                exc=exc,
+            )},
         )()
 
 
 def _chat_payload_with_content(content: str) -> dict:
     return {
-        "choices": [
-            {
-                "message": {
-                    "content": content,
-                }
+        "choices": [{
+            "message": {
+                "content": content,
             }
-        ]
+        }]
     }
 
 
@@ -172,6 +172,7 @@ def _build_staged_enhancer(
 
 
 class _FakeOpenAIClient:
+
     def __init__(self, **kwargs):
         self.kwargs = kwargs
         self.chat = type(
@@ -182,6 +183,7 @@ class _FakeOpenAIClient:
 
 
 class _FakeCerebrasClient:
+
     def __init__(self, **kwargs):
         self.kwargs = kwargs
         self.chat = type(
@@ -192,16 +194,12 @@ class _FakeCerebrasClient:
 
 
 def test_parse_json_response_accepts_fenced_json_with_prose():
-    parsed = _parse_json_response(
-        "Here is the rewrite:\n```json\n{\"segment_prompts\":[\"A\",\"B\"]}\n```\nThanks."
-    )
+    parsed = _parse_json_response("Here is the rewrite:\n```json\n{\"segment_prompts\":[\"A\",\"B\"]}\n```\nThanks.")
     assert parsed == {"segment_prompts": ["A", "B"]}
 
 
 def test_parse_json_response_extracts_first_embedded_object():
-    parsed = _parse_json_response(
-        "Model output:\n{\"segment_prompts\":[\"A\",\"B\"]}\n(complete)"
-    )
+    parsed = _parse_json_response("Model output:\n{\"segment_prompts\":[\"A\",\"B\"]}\n(complete)")
     assert parsed == {"segment_prompts": ["A", "B"]}
 
 
@@ -268,16 +266,12 @@ def test_build_client_supports_groq_provider(monkeypatch):
 
 def test_rewrite_prompt_sequence_accepts_segment_prompts_output():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.rollout_id == "preset_a"
@@ -286,15 +280,12 @@ def test_rewrite_prompt_sequence_accepts_segment_prompts_output():
 
 
 def test_rewrite_prompt_sequence_accepts_legacy_rewritten_prompts_output():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content('{"rewritten_prompts":["A","B"]}')
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"rewritten_prompts":["A","B"]}'))
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.rollout_id == "current_rollout"
@@ -303,19 +294,14 @@ def test_rewrite_prompt_sequence_accepts_legacy_rewritten_prompts_output():
 
 
 def test_rewrite_prompt_sequence_accepts_segment_dicts_without_top_level_rollout_metadata():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"segments":[{"prompt":"A"},{"text":"B"}]}'
-        )
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"segments":[{"prompt":"A"},{"text":"B"}]}'))
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             preset_id="preset_a",
             preset_label="Preset A",
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.rollout_id == "preset_a"
@@ -329,14 +315,12 @@ def test_rewrite_prompt_sequence_accepts_numbered_prose_output():
             "The user is asking for a cinematic rewrite.\n\n"
             "1. A dog bounds across the moon's dusty surface, kicking up silver regolith as it chases a rabbit beneath the black sky.\n"
             "2. The rabbit darts around a crater rim while the dog lunges after it, Earth glowing blue in the distance.\n"
-        )
-    )
+        ))
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.rollout_id == "current_rollout"
@@ -347,29 +331,27 @@ def test_rewrite_prompt_sequence_accepts_numbered_prose_output():
     ]
 
 
-def test_enhance_prompt_prefers_cerebras_before_groq_fallback():
+def test_enhance_prompt_uses_groq_when_it_returns_first():
     enhancer = _build_staged_enhancer(
         cerebras_payload=_chat_payload_with_content('{"prompt":"Cerebras prompt"}'),
         groq_payload=_chat_payload_with_content('{"prompt":"Groq prompt"}'),
-        cerebras_delay_s=0.01,
+        cerebras_delay_s=0.08,
         groq_delay_s=0.01,
     )
 
-    result = asyncio.run(
-        enhancer.enhance_prompt(
-            "A rainy alley at night",
-            mode="single_clip",
-        )
-    )
+    result = asyncio.run(enhancer.enhance_prompt(
+        "A rainy alley at night",
+        mode="single_clip",
+    ))
 
     assert result.fallback_used is False
     assert result.error is None
-    assert result.provider == "cerebras"
+    assert result.provider == "groq"
     assert result.model == "gpt-test"
-    assert result.prompt == "Cerebras prompt"
+    assert result.prompt == "Groq prompt"
     assert enhancer.get_provider_success_counts() == {
-        "cerebras": 1,
-        "groq": 0,
+        "cerebras": 0,
+        "groq": 1,
     }
 
 
@@ -381,12 +363,10 @@ def test_enhance_prompt_uses_groq_when_cerebras_fails():
         groq_delay_s=0.01,
     )
 
-    result = asyncio.run(
-        enhancer.enhance_prompt(
-            "A rainy alley at night",
-            mode="single_clip",
-        )
-    )
+    result = asyncio.run(enhancer.enhance_prompt(
+        "A rainy alley at night",
+        mode="single_clip",
+    ))
 
     assert result.fallback_used is False
     assert result.error is None
@@ -408,13 +388,11 @@ def test_enhance_prompt_can_use_groq_when_cerebras_times_out():
     enhancer.http_timeout_ms = 50
     enhancer.default_timeout_ms = 50
 
-    result = asyncio.run(
-        enhancer.enhance_prompt(
-            "A rainy alley at night",
-            mode="single_clip",
-            timeout_ms=50,
-        )
-    )
+    result = asyncio.run(enhancer.enhance_prompt(
+        "A rainy alley at night",
+        mode="single_clip",
+        timeout_ms=50,
+    ))
 
     assert result.fallback_used is False
     assert result.error is None
@@ -434,12 +412,10 @@ def test_enhance_prompt_can_use_cerebras_when_it_returns_first():
         groq_delay_s=0.08,
     )
 
-    result = asyncio.run(
-        enhancer.enhance_prompt(
-            "A rainy alley at night",
-            mode="single_clip",
-        )
-    )
+    result = asyncio.run(enhancer.enhance_prompt(
+        "A rainy alley at night",
+        mode="single_clip",
+    ))
 
     assert result.fallback_used is False
     assert result.error is None
@@ -453,15 +429,12 @@ def test_enhance_prompt_can_use_cerebras_when_it_returns_first():
 
 
 def test_rewrite_prompt_sequence_keeps_raw_output_on_parse_error():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content("I cannot comply with JSON right now.")
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content("I cannot comply with JSON right now."))
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert "No JSON object found in assistant response." in (result.error or "")
     assert result.raw_response_text == "I cannot comply with JSON right now."
@@ -473,9 +446,7 @@ def test_rewrite_prompt_sequence_keeps_raw_output_on_parse_error():
 def test_rewrite_prompt_sequence_uses_current_rollout_payload_shape():
     enhancer = _build_test_enhancer(
         _chat_payload_with_content(
-            '{"id":"rewritten_rollout","label":"Rewritten Rollout","segment_prompts":["A","B"]}'
-        )
-    )
+            '{"id":"rewritten_rollout","label":"Rewritten Rollout","segment_prompts":["A","B"]}'))
     captured = {
         "body": None,
         "timeout_seconds": None,
@@ -486,8 +457,7 @@ def test_rewrite_prompt_sequence_uses_current_rollout_payload_shape():
         captured["timeout_seconds"] = timeout_seconds
         return (
             _chat_payload_with_content(
-                '{"id":"rewritten_rollout","label":"Rewritten Rollout","segment_prompts":["A","B"]}'
-            ),
+                '{"id":"rewritten_rollout","label":"Rewritten Rollout","segment_prompts":["A","B"]}'),
             '{"id":"rewritten_rollout","label":"Rewritten Rollout","segment_prompts":["A","B"]}',
         )
 
@@ -502,8 +472,7 @@ def test_rewrite_prompt_sequence_uses_current_rollout_payload_shape():
             rewrite_model="gpt-test",
             rewrite_temperature=0.2,
             timeout_ms=800,
-        )
-    )
+        ))
 
     assert result.fallback_used is False
     assert captured["body"]["messages"][0] == {
@@ -512,12 +481,12 @@ def test_rewrite_prompt_sequence_uses_current_rollout_payload_shape():
     }
     assert captured["body"]["messages"][1]["role"] == "user"
     assert prompt_enhancer_module.json.loads(captured["body"]["messages"][1]["content"]) == {
-        "mode": "edit_existing_rollout",
-        "request": (
-            "Rewrite all segment prompts with improved continuity and cinematic detail. "
-            "Keep count and ordering identical."
-        ),
-        "user_instruction": "make it cinematic",
+        "mode":
+        "edit_existing_rollout",
+        "request": ("Rewrite all segment prompts with improved continuity and cinematic detail. "
+                    "Keep count and ordering identical."),
+        "user_instruction":
+        "make it cinematic",
         "current_rollout": {
             "id": "preset_a",
             "label": "Preset A",
@@ -528,11 +497,8 @@ def test_rewrite_prompt_sequence_uses_current_rollout_payload_shape():
 
 def test_rewrite_prompt_sequence_supports_new_rollout_mode():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"custom_editable","label":"Custom rollout","segment_prompts":['
-            '"A","B","C","D","E","F"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"custom_editable","label":"Custom rollout","segment_prompts":['
+                                   '"A","B","C","D","E","F"]}'))
     captured = {
         "body": None,
     }
@@ -541,10 +507,8 @@ def test_rewrite_prompt_sequence_supports_new_rollout_mode():
         del timeout_seconds
         captured["body"] = body
         return (
-            _chat_payload_with_content(
-                '{"id":"custom_editable","label":"Custom rollout","segment_prompts":['
-                '"A","B","C","D","E","F"]}'
-            ),
+            _chat_payload_with_content('{"id":"custom_editable","label":"Custom rollout","segment_prompts":['
+                                       '"A","B","C","D","E","F"]}'),
             '{"id":"custom_editable","label":"Custom rollout","segment_prompts":['
             '"A","B","C","D","E","F"]}',
         )
@@ -560,30 +524,29 @@ def test_rewrite_prompt_sequence_supports_new_rollout_mode():
             rewrite_model="gpt-test",
             rewrite_temperature=0.2,
             timeout_ms=800,
-        )
-    )
+        ))
 
     assert result.fallback_used is False
     assert result.prompts == ["A", "B", "C", "D", "E", "F"]
     assert prompt_enhancer_module.json.loads(captured["body"]["messages"][1]["content"]) == {
-        "mode": "new_rollout",
-        "request": (
-            "Rewrite all segment prompts with improved continuity and cinematic detail. "
-            "Keep count and ordering identical."
-        ),
-        "user_instruction": "A moonbase corridor thriller with flooding and red alarms",
-        "desired_segment_count": 6,
-        "rollout_id_hint": "custom_editable",
-        "rollout_label_hint": "Custom rollout",
+        "mode":
+        "new_rollout",
+        "request": ("Rewrite all segment prompts with improved continuity and cinematic detail. "
+                    "Keep count and ordering identical."),
+        "user_instruction":
+        "A moonbase corridor thriller with flooding and red alarms",
+        "desired_segment_count":
+        6,
+        "rollout_id_hint":
+        "custom_editable",
+        "rollout_label_hint":
+        "Custom rollout",
     }
 
 
 def test_rewrite_prompt_sequence_uses_session_override_system_prompt():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     enhancer.rewrite_all_system_prompt = "shared system prompt"
     captured = {
         "body": None,
@@ -593,9 +556,7 @@ def test_rewrite_prompt_sequence_uses_session_override_system_prompt():
         del timeout_seconds
         captured["body"] = body
         return (
-            _chat_payload_with_content(
-                '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-            ),
+            _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'),
             '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}',
         )
 
@@ -609,8 +570,7 @@ def test_rewrite_prompt_sequence_uses_session_override_system_prompt():
             rewrite_instruction="make it cinematic",
             rewrite_model="gpt-test",
             system_prompt_override="session specific system prompt",
-        )
-    )
+        ))
 
     assert result.fallback_used is False
     assert captured["body"]["messages"][0] == {
@@ -621,10 +581,7 @@ def test_rewrite_prompt_sequence_uses_session_override_system_prompt():
 
 def test_resolve_rewrite_new_rollout_system_prompt_uses_dedicated_prompt():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     enhancer.rewrite_all_system_prompt = "shared rewrite system prompt"
     enhancer.rewrite_user_system_prompt = "new rollout rewrite system prompt"
 
@@ -635,24 +592,17 @@ def test_resolve_rewrite_new_rollout_system_prompt_uses_dedicated_prompt():
 
 def test_resolve_rewrite_new_rollout_system_prompt_prefers_override():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     enhancer.rewrite_all_system_prompt = "shared rewrite system prompt"
     enhancer.rewrite_user_system_prompt = "new rollout rewrite system prompt"
 
-    resolved = enhancer.resolve_rewrite_new_rollout_system_prompt(
-        "session specific system prompt"
-    )
+    resolved = enhancer.resolve_rewrite_new_rollout_system_prompt("session specific system prompt")
 
     assert resolved == "session specific system prompt"
 
 
 def test_generate_auto_prompt_uses_selected_model():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content('{"next_prompt":"Auto next"}')
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"next_prompt":"Auto next"}'))
     enhancer.auto_system_prompt = "auto system prompt"
     enhancer.rewrite_model_options = ["gpt-test", "gpt-alt"]
     enhancer.rewrite_default_model = "gpt-test"
@@ -680,8 +630,7 @@ def test_generate_auto_prompt_uses_selected_model():
             next_segment_idx=2,
             model="gpt-alt",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.prompt == "Auto next"
@@ -690,9 +639,7 @@ def test_generate_auto_prompt_uses_selected_model():
 
 
 def test_enhance_prompt_uses_selected_model():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content('{"next_prompt":"Enhanced next"}')
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"next_prompt":"Enhanced next"}'))
     enhancer.enhance_system_prompt = "enhance system prompt"
     enhancer.auto_system_prompt = "auto system prompt"
     enhancer.rewrite_model_options = ["gpt-test", "gpt-alt"]
@@ -722,8 +669,7 @@ def test_enhance_prompt_uses_selected_model():
             next_segment_idx=2,
             model="gpt-alt",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.prompt == "Enhanced next"
@@ -732,9 +678,7 @@ def test_enhance_prompt_uses_selected_model():
 
 
 def test_enhance_prompt_single_clip_uses_auto_extension_prompt_and_prompt_field():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content('{"prompt":"Extended single clip"}')
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"prompt":"Extended single clip"}'))
     enhancer.enhance_system_prompt = "enhance system prompt"
     enhancer.auto_system_prompt = "auto system prompt"
     enhancer.rewrite_model_options = ["gpt-test", "gpt-alt"]
@@ -764,14 +708,12 @@ def test_enhance_prompt_single_clip_uses_auto_extension_prompt_and_prompt_field(
 
     enhancer._request_content = _fake_request_content  # type: ignore[attr-defined]
 
-    result = asyncio.run(
-        enhancer.enhance_prompt(
-            "short 5s idea",
-            mode="single_clip",
-            model="gpt-alt",
-            timeout_ms=800,
-        )
-    )
+    result = asyncio.run(enhancer.enhance_prompt(
+        "short 5s idea",
+        mode="single_clip",
+        model="gpt-alt",
+        timeout_ms=800,
+    ))
     assert result.fallback_used is False
     assert result.error is None
     assert result.prompt == "Extended single clip"
@@ -784,17 +726,15 @@ def test_enhance_prompt_single_clip_uses_auto_extension_prompt_and_prompt_field(
             "single 5-second LTX-2.3 video clip. Respond with "
             'valid JSON only as {"prompt": "..."}.'  # noqa: E501
         ),
-        "user_prompt": "short 5s idea",
+        "user_prompt":
+        "short 5s idea",
     }
 
 
 def test_enhance_prompt_single_clip_rejects_plain_text_response():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            "Medium shot of a woman by a rainy cafe window as she lifts her "
-            "phone, exhales softly, and the camera makes a slow push in."
-        )
-    )
+        _chat_payload_with_content("Medium shot of a woman by a rainy cafe window as she lifts her "
+                                   "phone, exhales softly, and the camera makes a slow push in."))
     enhancer.auto_system_prompt = "auto system prompt"
 
     result = asyncio.run(
@@ -803,17 +743,14 @@ def test_enhance_prompt_single_clip_rejects_plain_text_response():
             mode="single_clip",
             model="gpt-test",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert "No JSON object found in assistant response." in result.error
     assert result.prompt == ""
 
 
 def test_enhance_prompt_single_clip_rejects_segment_prompts_json():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content('{"segment_prompts":["A","B"]}')
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content('{"segment_prompts":["A","B"]}'))
     enhancer.auto_system_prompt = "auto system prompt"
     enhancer.rewrite_model_options = ["gpt-test"]
     enhancer.rewrite_default_model = "gpt-test"
@@ -824,17 +761,14 @@ def test_enhance_prompt_single_clip_rejects_segment_prompts_json():
             mode="single_clip",
             model="gpt-test",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert result.prompt == ""
     assert "Missing prompt string." in (result.error or "")
 
 
 def test_enhance_prompt_requires_json_and_does_not_fallback_to_raw_text():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content("A cinematic continuation with slow dolly movement.")
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content("A cinematic continuation with slow dolly movement."))
     enhancer.enhance_system_prompt = "enhance system prompt"
     enhancer.rewrite_model_options = ["gpt-test"]
     enhancer.rewrite_default_model = "gpt-test"
@@ -846,17 +780,14 @@ def test_enhance_prompt_requires_json_and_does_not_fallback_to_raw_text():
             next_segment_idx=2,
             model="gpt-test",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert result.prompt == ""
     assert "No JSON object found in assistant response." in (result.error or "")
 
 
 def test_generate_auto_prompt_requires_json_and_does_not_fallback_to_raw_text():
-    enhancer = _build_test_enhancer(
-        _chat_payload_with_content("A calm, grounded continuation with subtle motion.")
-    )
+    enhancer = _build_test_enhancer(_chat_payload_with_content("A calm, grounded continuation with subtle motion."))
     enhancer.auto_system_prompt = "auto system prompt"
     enhancer.rewrite_model_options = ["gpt-test"]
     enhancer.rewrite_default_model = "gpt-test"
@@ -867,34 +798,30 @@ def test_generate_auto_prompt_requires_json_and_does_not_fallback_to_raw_text():
             next_segment_idx=2,
             model="gpt-test",
             timeout_ms=800,
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert result.prompt == ""
     assert "No JSON object found in assistant response." in (result.error or "")
 
 
 def test_rewrite_prompt_sequence_includes_raw_json_when_content_empty():
-    enhancer = _build_test_enhancer(
-        {
-            "choices": [
-                {
-                    "finish_reason": "length",
-                    "message": {
-                        "content": [],
-                        "refusal": None,
-                    },
-                }
-            ],
-            "usage": {"completion_tokens": 0},
-        }
-    )
+    enhancer = _build_test_enhancer({
+        "choices": [{
+            "finish_reason": "length",
+            "message": {
+                "content": [],
+                "refusal": None,
+            },
+        }],
+        "usage": {
+            "completion_tokens": 0
+        },
+    })
     result = asyncio.run(
         enhancer.rewrite_prompt_sequence(
             ["prompt one", "prompt two"],
             rewrite_instruction="make it cinematic",
-        )
-    )
+        ))
     assert result.fallback_used is True
     assert "No rewrite segment prompts found in assistant response." in (result.error or "")
     assert isinstance(result.raw_response_text, str)
@@ -903,10 +830,7 @@ def test_rewrite_prompt_sequence_includes_raw_json_when_content_empty():
 
 def test_get_rewrite_model_config_returns_fixed_defaults():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     enhancer.rewrite_default_model = "gpt-oss-120b"
     enhancer.rewrite_model_options = ["gpt-oss-120b"]
 
@@ -918,10 +842,7 @@ def test_get_rewrite_model_config_returns_fixed_defaults():
 
 def test_get_prompt_config_includes_auto_extension_prompt():
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     enhancer.enhance_system_prompt_path = "/tmp/next.md"
     enhancer.auto_system_prompt_path = "/tmp/auto.md"
     enhancer.rewrite_all_system_prompt_path = "/tmp/rewrite.md"
@@ -948,19 +869,14 @@ def test_get_prompt_config_includes_auto_extension_prompt():
 
 def test_get_prompt_config_reports_loaded_fallback_prompt_path(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     rewrite_fallback_path = tmp_path / "rewrite_window_system_prompt.md"
     rewrite_fallback_path.write_text("rewrite prompt\n", encoding="utf-8")
     next_path = tmp_path / "next.md"
     next_path.write_text("next prompt\n", encoding="utf-8")
     auto_path = tmp_path / "auto.md"
     auto_path.write_text("auto prompt\n", encoding="utf-8")
-    enhancer.rewrite_all_system_prompt_path = str(
-        tmp_path / "prompts.local" / "rewrite_window_system_prompt.md"
-    )
+    enhancer.rewrite_all_system_prompt_path = str(tmp_path / "prompts.local" / "rewrite_window_system_prompt.md")
     enhancer.rewrite_all_system_prompt_fallback_path = str(rewrite_fallback_path)
     enhancer.enhance_system_prompt_path = str(next_path)
     enhancer.auto_system_prompt_path = str(auto_path)
@@ -973,14 +889,9 @@ def test_get_prompt_config_reports_loaded_fallback_prompt_path(tmp_path):
     assert config["rewrite_window_system_prompt_path"] == str(rewrite_fallback_path)
 
 
-def test_reload_system_prompts_falls_back_to_rewrite_window_when_user_prompt_empty(
-    tmp_path,
-):
+def test_reload_system_prompts_falls_back_to_rewrite_window_when_user_prompt_empty(tmp_path, ):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite_window_system_prompt.md"
@@ -1007,10 +918,7 @@ def test_reload_system_prompts_falls_back_to_rewrite_window_when_user_prompt_emp
 
 def test_save_prompt_config_updates_auto_extension_prompt(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite.md"
@@ -1021,9 +929,7 @@ def test_save_prompt_config_updates_auto_extension_prompt(tmp_path):
     enhancer.auto_system_prompt_path = str(auto_path)
     enhancer.rewrite_all_system_prompt_path = str(rewrite_path)
 
-    config = enhancer.save_prompt_config(
-        auto_extension_system_prompt="auto updated",
-    )
+    config = enhancer.save_prompt_config(auto_extension_system_prompt="auto updated", )
 
     assert auto_path.read_text(encoding="utf-8").strip() == "auto updated"
     assert config["auto_extension_system_prompt"] == "auto updated"
@@ -1031,10 +937,7 @@ def test_save_prompt_config_updates_auto_extension_prompt(tmp_path):
 
 def test_save_prompt_config_updates_rewrite_user_prompt(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite.md"
@@ -1052,9 +955,7 @@ def test_save_prompt_config_updates_rewrite_user_prompt(tmp_path):
     enhancer.rewrite_all_system_prompt_fallback_path = None
     enhancer.rewrite_user_system_prompt_fallback_path = None
 
-    config = enhancer.save_prompt_config(
-        rewrite_user_system_prompt="rewrite user updated",
-    )
+    config = enhancer.save_prompt_config(rewrite_user_system_prompt="rewrite user updated", )
 
     assert rewrite_user_path.read_text(encoding="utf-8").strip() == "rewrite user updated"
     assert config["rewrite_user_system_prompt"] == "rewrite user updated"
@@ -1062,10 +963,7 @@ def test_save_prompt_config_updates_rewrite_user_prompt(tmp_path):
 
 def test_save_prompt_config_updates_rewrite_model(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite.md"
@@ -1081,9 +979,7 @@ def test_save_prompt_config_updates_rewrite_model(tmp_path):
     enhancer.rewrite_default_model = "gpt-test"
     enhancer.rewrite_model_options = ["gpt-test", "gpt-alt"]
 
-    config = enhancer.save_prompt_config(
-        rewrite_model="gpt-alt",
-    )
+    config = enhancer.save_prompt_config(rewrite_model="gpt-alt", )
 
     assert enhancer.rewrite_default_model == "gpt-alt"
     assert config["rewrite_model"] == "gpt-alt"
@@ -1092,10 +988,7 @@ def test_save_prompt_config_updates_rewrite_model(tmp_path):
 
 def test_save_prompt_config_updates_rewrite_temperature(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite.md"
@@ -1109,9 +1002,7 @@ def test_save_prompt_config_updates_rewrite_temperature(tmp_path):
     enhancer.auto_system_prompt_fallback_path = None
     enhancer.rewrite_all_system_prompt_fallback_path = None
 
-    config = enhancer.save_prompt_config(
-        rewrite_temperature=1.3,
-    )
+    config = enhancer.save_prompt_config(rewrite_temperature=1.3, )
 
     assert enhancer.rewrite_default_temperature == 1.3
     assert config["rewrite_temperature"] == 1.3
@@ -1119,10 +1010,7 @@ def test_save_prompt_config_updates_rewrite_temperature(tmp_path):
 
 def test_save_prompt_config_creates_versioned_backup_for_existing_prompt(tmp_path):
     enhancer = _build_test_enhancer(
-        _chat_payload_with_content(
-            '{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'
-        )
-    )
+        _chat_payload_with_content('{"id":"preset_a","label":"Preset A","segment_prompts":["A","B"]}'))
     next_path = tmp_path / "next.md"
     auto_path = tmp_path / "auto.md"
     rewrite_path = tmp_path / "rewrite_window_system_prompt.md"
@@ -1136,13 +1024,9 @@ def test_save_prompt_config_creates_versioned_backup_for_existing_prompt(tmp_pat
     enhancer.auto_system_prompt_fallback_path = None
     enhancer.rewrite_all_system_prompt_fallback_path = None
 
-    enhancer.save_prompt_config(
-        rewrite_window_system_prompt="rewrite updated",
-    )
+    enhancer.save_prompt_config(rewrite_window_system_prompt="rewrite updated", )
 
-    backup_paths = sorted(
-        tmp_path.glob("rewrite_window_system_prompt.*.bak.md")
-    )
+    backup_paths = sorted(tmp_path.glob("rewrite_window_system_prompt.*.bak.md"))
 
     assert rewrite_path.read_text(encoding="utf-8").strip() == "rewrite updated"
     assert len(backup_paths) == 1

apps/dreamverse/dreamverse/tests/test_realtime_stress.py

@@ -27,13 +27,8 @@ try:
 except ModuleNotFoundError:
     websockets = None  # type: ignore[assignment]
 
-
-DEFAULT_PRESET_FILE = (
-    Path(__file__).resolve().parents[2]
-    / "web"
-    / "prompts"
-    / "selected_ltx2_continuation_story_presets.json"
-)
+DEFAULT_PRESET_FILE = (Path(__file__).resolve().parents[2] / "web" / "prompts" /
+                       "selected_ltx2_continuation_story_presets.json")
 
 
 def utc_now_iso() -> str:
@@ -65,10 +60,7 @@ def safe_percentile(values: list[float], percentile: float) -> float | None:
     if lower == upper:
         return sorted_values[lower]
     fraction = rank - lower
-    return (
-        sorted_values[lower]
-        + (sorted_values[upper] - sorted_values[lower]) * fraction
-    )
+    return (sorted_values[lower] + (sorted_values[upper] - sorted_values[lower]) * fraction)
 
 
 def summarize_series(values: list[float]) -> dict[str, float | int | None]:
@@ -145,24 +137,16 @@ def load_curated_prompts(
     selected_id = str(selected.get("id", "")).strip() or "unknown_preset"
     raw_prompts = selected.get("segment_prompts", [])
     if not isinstance(raw_prompts, list):
-        raise ValueError(
-            f"Preset {selected_id} has invalid segment_prompts (must be list)."
-        )
+        raise ValueError(f"Preset {selected_id} has invalid segment_prompts (must be list).")
 
-    prompts = [
-        str(prompt).strip()
-        for prompt in raw_prompts
-        if isinstance(prompt, str) and str(prompt).strip()
-    ]
+    prompts = [str(prompt).strip() for prompt in raw_prompts if isinstance(prompt, str) and str(prompt).strip()]
     if not prompts:
         raise ValueError(f"Preset {selected_id} has no non-empty prompts.")
 
     limited = prompts[:curated_limit]
     if not limited:
-        raise ValueError(
-            f"curated_limit={curated_limit} produced no prompts for preset "
-            f"{selected_id}."
-        )
+        raise ValueError(f"curated_limit={curated_limit} produced no prompts for preset "
+                         f"{selected_id}.")
     return selected_id, limited, len(prompts)
 
 
@@ -224,11 +208,11 @@ async def run_single_session(
 
     try:
         async with websockets.connect(
-            url,
-            max_size=None,
-            ping_interval=None,
-            open_timeout=connect_timeout_s,
-            close_timeout=2.0,
+                url,
+                max_size=None,
+                ping_interval=None,
+                open_timeout=connect_timeout_s,
+                close_timeout=2.0,
         ) as ws:
             connect_finish_monotonic = time.monotonic()
             session_data["connect_finish_ts_utc"] = utc_now_iso()
@@ -249,9 +233,7 @@ async def run_single_session(
                 timeout_remaining = session_timeout_s - elapsed_s
                 if timeout_remaining <= 0:
                     session_data["status"] = "timeout"
-                    session_data["error"] = (
-                        f"Session timed out after {session_timeout_s:.1f}s."
-                    )
+                    session_data["error"] = (f"Session timed out after {session_timeout_s:.1f}s.")
                     break
 
                 recv_start_epoch = time.time()
@@ -265,9 +247,7 @@ async def run_single_session(
                     )
                 except asyncio.TimeoutError:
                     session_data["status"] = "timeout"
-                    session_data["error"] = (
-                        "Timed out waiting for websocket message."
-                    )
+                    session_data["error"] = ("Timed out waiting for websocket message.")
                     break
                 except Exception as exc:
                     session_data["status"] = "failed"
@@ -288,20 +268,16 @@ async def run_single_session(
 
                     chunk_gap_ms: float | None = None
                     if last_chunk_finish_monotonic is not None:
-                        chunk_gap_ms = (
-                            recv_finish_monotonic - last_chunk_finish_monotonic
-                        ) * 1000.0
+                        chunk_gap_ms = (recv_finish_monotonic - last_chunk_finish_monotonic) * 1000.0
 
-                    session_data["chunks"].append(
-                        {
-                            "segment_idx": current_segment_idx,
-                            "chunk_idx": session_data["total_chunks"],
-                            "size_bytes": len(message),
-                            "chunk_start_ts_utc": recv_start_iso,
-                            "chunk_finish_ts_utc": recv_finish_iso,
-                            "chunk_gap_ms": chunk_gap_ms,
-                        }
-                    )
+                    session_data["chunks"].append({
+                        "segment_idx": current_segment_idx,
+                        "chunk_idx": session_data["total_chunks"],
+                        "size_bytes": len(message),
+                        "chunk_start_ts_utc": recv_start_iso,
+                        "chunk_finish_ts_utc": recv_finish_iso,
+                        "chunk_gap_ms": chunk_gap_ms,
+                    })
                     last_chunk_finish_monotonic = recv_finish_monotonic
                     last_chunk_finish_epoch = recv_finish_epoch
                     session_data["last_chunk_finish_ts_utc"] = recv_finish_iso
@@ -321,9 +297,7 @@ async def run_single_session(
                 if msg_type == "gpu_assigned":
                     session_data["gpu_assigned_ts_utc"] = recv_finish_iso
                     if connect_finish_monotonic is not None:
-                        session_data["queue_wait_ms"] = (
-                            recv_finish_monotonic - connect_finish_monotonic
-                        ) * 1000.0
+                        session_data["queue_wait_ms"] = (recv_finish_monotonic - connect_finish_monotonic) * 1000.0
                 elif msg_type == "ltx2_stream_start":
                     if initial_total_segments is None:
                         parsed_total = parse_int(data.get("total_segments"))
@@ -338,20 +312,13 @@ async def run_single_session(
                     session_data["media_segments_completed"] += 1
                     if first_media_segment_complete_epoch is None:
                         first_media_segment_complete_epoch = recv_finish_epoch
-                        session_data[
-                            "first_media_segment_complete_ts_utc"
-                        ] = recv_finish_iso
+                        session_data["first_media_segment_complete_ts_utc"] = recv_finish_iso
                 elif msg_type == "ltx2_segment_complete":
                     session_data["segments_completed"] += 1
                     seg_idx = parse_int(data.get("segment_idx"))
-                    if (
-                        initial_total_segments is not None
-                        and seg_idx is not None
-                        and seg_idx >= initial_total_segments
-                    ):
-                        session_data[
-                            "target_segment_complete_ts_utc"
-                        ] = recv_finish_iso
+                    if (initial_total_segments is not None and seg_idx is not None
+                            and seg_idx >= initial_total_segments):
+                        session_data["target_segment_complete_ts_utc"] = recv_finish_iso
                         await asyncio.sleep(post_complete_wait_s)
                         session_data["leave_sent_ts_utc"] = utc_now_iso()
                         try:
@@ -362,15 +329,11 @@ async def run_single_session(
                         break
                 elif msg_type == "session_timeout":
                     session_data["status"] = "timeout"
-                    session_data["error"] = str(
-                        data.get("message") or "Backend session timeout"
-                    )
+                    session_data["error"] = str(data.get("message") or "Backend session timeout")
                     break
                 elif msg_type == "error":
                     session_data["status"] = "failed"
-                    session_data["error"] = str(
-                        data.get("message") or "Backend error message"
-                    )
+                    session_data["error"] = str(data.get("message") or "Backend error message")
                     break
 
             if session_data["status"] == "failed" and session_data["error"] is None:
@@ -379,29 +342,18 @@ async def run_single_session(
         session_data["status"] = "failed"
         session_data["error"] = f"WebSocket connect/run failed: {exc}"
 
-    if (
-        first_chunk_finish_epoch is not None
-        and last_chunk_finish_epoch is not None
-        and session_data["total_chunk_bytes"] > 0
-    ):
+    if (first_chunk_finish_epoch is not None and last_chunk_finish_epoch is not None
+            and session_data["total_chunk_bytes"] > 0):
         duration_s = last_chunk_finish_epoch - first_chunk_finish_epoch
         if duration_s > 0:
-            session_data["session_goodput_mbps"] = (
-                session_data["total_chunk_bytes"] * 8.0 / duration_s / 1_000_000.0
-            )
+            session_data["session_goodput_mbps"] = (session_data["total_chunk_bytes"] * 8.0 / duration_s / 1_000_000.0)
 
-    if (
-        first_chunk_finish_epoch is not None
-        and first_media_segment_complete_epoch is not None
-    ):
-        session_data["first_chunk_before_first_media_complete"] = (
-            first_chunk_finish_epoch < first_media_segment_complete_epoch
-        )
+    if (first_chunk_finish_epoch is not None and first_media_segment_complete_epoch is not None):
+        session_data["first_chunk_before_first_media_complete"] = (first_chunk_finish_epoch
+                                                                   < first_media_segment_complete_epoch)
 
     session_data["close_ts_utc"] = utc_now_iso()
-    session_data["duration_ms"] = (
-        time.monotonic() - session_start_monotonic
-    ) * 1000.0
+    session_data["duration_ms"] = (time.monotonic() - session_start_monotonic) * 1000.0
     return session_data
 
 
@@ -412,14 +364,11 @@ async def run_worker_sessions(
     config: dict[str, Any],
 ) -> list[dict[str, Any]]:
     tasks = [
-        asyncio.create_task(
-            run_single_session(
-                worker_id=worker_id,
-                worker_session_idx=idx,
-                config=config,
-            )
-        )
-        for idx in range(session_count)
+        asyncio.create_task(run_single_session(
+            worker_id=worker_id,
+            worker_session_idx=idx,
+            config=config,
+        )) for idx in range(session_count)
     ]
     if not tasks:
         return []
@@ -437,29 +386,23 @@ def worker_entry(
     try:
         ready_queue.put({"worker_id": worker_id, "status": "ready"})
         start_event.wait()
-        sessions = asyncio.run(
-            run_worker_sessions(
-                worker_id=worker_id,
-                session_count=session_count,
-                config=config,
-            )
-        )
-        result_queue.put(
-            {
-                "worker_id": worker_id,
-                "status": "ok",
-                "sessions": sessions,
-            }
-        )
+        sessions = asyncio.run(run_worker_sessions(
+            worker_id=worker_id,
+            session_count=session_count,
+            config=config,
+        ))
+        result_queue.put({
+            "worker_id": worker_id,
+            "status": "ok",
+            "sessions": sessions,
+        })
     except Exception as exc:
-        result_queue.put(
-            {
-                "worker_id": worker_id,
-                "status": "error",
-                "error": str(exc),
-                "traceback": traceback.format_exc(),
-            }
-        )
+        result_queue.put({
+            "worker_id": worker_id,
+            "status": "error",
+            "error": str(exc),
+            "traceback": traceback.format_exc(),
+        })
 
 
 def build_summary(
@@ -517,33 +460,22 @@ def build_summary(
     if len(all_chunk_finish_epochs) >= 2 and total_chunk_bytes > 0:
         duration_s = max(all_chunk_finish_epochs) - min(all_chunk_finish_epochs)
         if duration_s > 0:
-            global_goodput_mbps = (
-                total_chunk_bytes * 8.0 / duration_s / 1_000_000.0
-            )
+            global_goodput_mbps = (total_chunk_bytes * 8.0 / duration_s / 1_000_000.0)
 
-    bucket_throughputs_mbps = [
-        (bytes_count * 8.0) / 1_000_000.0
-        for _, bytes_count in sorted(bucket_bytes.items())
-    ]
+    bucket_throughputs_mbps = [(bytes_count * 8.0) / 1_000_000.0 for _, bytes_count in sorted(bucket_bytes.items())]
     bucket_stats = summarize_series(bucket_throughputs_mbps)
 
-    chunk_gap_threshold_breaches = [
-        value for value in chunk_gaps if value >= chunk_gap_threshold_ms
-    ]
+    chunk_gap_threshold_breaches = [value for value in chunk_gaps if value >= chunk_gap_threshold_ms]
     non_success = len(sessions) - status_counts.get("success", 0)
 
     fail_reasons: list[str] = []
     if non_success > 0:
-        fail_reasons.append(
-            f"{non_success} session(s) did not complete successfully."
-        )
+        fail_reasons.append(f"{non_success} session(s) did not complete successfully.")
     if not chunk_gaps:
         fail_reasons.append("No chunk gap data collected.")
     if chunk_gap_threshold_breaches:
-        fail_reasons.append(
-            f"{len(chunk_gap_threshold_breaches)} chunk gap(s) were >= "
-            f"{chunk_gap_threshold_ms:.0f}ms."
-        )
+        fail_reasons.append(f"{len(chunk_gap_threshold_breaches)} chunk gap(s) were >= "
+                            f"{chunk_gap_threshold_ms:.0f}ms.")
 
     passed = len(fail_reasons) == 0
     progressive_ratio = None
@@ -554,20 +486,18 @@ def build_summary(
         "passed": passed,
         "fail_reasons": fail_reasons,
         "sessions": {
-            "total": len(sessions),
-            "success": status_counts.get("success", 0),
-            "failed": status_counts.get("failed", 0),
-            "timeout": status_counts.get("timeout", 0),
-            "protocol_error": status_counts.get("protocol_error", 0),
-            "other": (
-                len(sessions)
-                - (
-                    status_counts.get("success", 0)
-                    + status_counts.get("failed", 0)
-                    + status_counts.get("timeout", 0)
-                    + status_counts.get("protocol_error", 0)
-                )
-            ),
+            "total":
+            len(sessions),
+            "success":
+            status_counts.get("success", 0),
+            "failed":
+            status_counts.get("failed", 0),
+            "timeout":
+            status_counts.get("timeout", 0),
+            "protocol_error":
+            status_counts.get("protocol_error", 0),
+            "other": (len(sessions) - (status_counts.get("success", 0) + status_counts.get("failed", 0) +
+                                       status_counts.get("timeout", 0) + status_counts.get("protocol_error", 0))),
         },
         "chunk_gap_ms": {
             **chunk_gap_stats,
@@ -606,51 +536,39 @@ def print_summary(
     bucket_bw = bandwidth["bucketed_1s"]
 
     print("=== LTX2 Realtime Stress Test Summary ===")
-    print(
-        "Run: "
-        f"url={run_info['url']} clients={run_info['clients']} "
-        f"processes={run_info['processes']} "
-        f"preset={run_info['preset_id']} "
-        f"curated_limit={run_info['curated_limit']}"
-    )
-    print(
-        "Sessions: "
-        f"total={sessions['total']} success={sessions['success']} "
-        f"failed={sessions['failed']} timeout={sessions['timeout']} "
-        f"protocol_error={sessions['protocol_error']}"
-    )
-    print(
-        "Chunk gap ms: "
-        f"min={format_num(chunk_gap['min'])} "
-        f"p50={format_num(chunk_gap['p50'])} "
-        f"p95={format_num(chunk_gap['p95'])} "
-        f"p99={format_num(chunk_gap['p99'])} "
-        f"max={format_num(chunk_gap['max'])} "
-        f"threshold={format_num(chunk_gap['threshold_ms'])} "
-        f"breaches={chunk_gap['breach_count']}"
-    )
-    print(
-        "Queue wait ms: "
-        f"min={format_num(queue_wait['min'])} "
-        f"p50={format_num(queue_wait['p50'])} "
-        f"p95={format_num(queue_wait['p95'])} "
-        f"max={format_num(queue_wait['max'])}"
-    )
+    print("Run: "
+          f"url={run_info['url']} clients={run_info['clients']} "
+          f"processes={run_info['processes']} "
+          f"preset={run_info['preset_id']} "
+          f"curated_limit={run_info['curated_limit']}")
+    print("Sessions: "
+          f"total={sessions['total']} success={sessions['success']} "
+          f"failed={sessions['failed']} timeout={sessions['timeout']} "
+          f"protocol_error={sessions['protocol_error']}")
+    print("Chunk gap ms: "
+          f"min={format_num(chunk_gap['min'])} "
+          f"p50={format_num(chunk_gap['p50'])} "
+          f"p95={format_num(chunk_gap['p95'])} "
+          f"p99={format_num(chunk_gap['p99'])} "
+          f"max={format_num(chunk_gap['max'])} "
+          f"threshold={format_num(chunk_gap['threshold_ms'])} "
+          f"breaches={chunk_gap['breach_count']}")
+    print("Queue wait ms: "
+          f"min={format_num(queue_wait['min'])} "
+          f"p50={format_num(queue_wait['p50'])} "
+          f"p95={format_num(queue_wait['p95'])} "
+          f"max={format_num(queue_wait['max'])}")
     ratio = progressive["ratio"]
     ratio_text = "n/a" if ratio is None else f"{ratio * 100:.2f}%"
-    print(
-        "Progressive streaming: "
-        f"{progressive['success_sessions']}/"
-        f"{progressive['eligible_sessions']} ({ratio_text})"
-    )
-    print(
-        "Bandwidth Mbps: "
-        f"per_session_avg={format_num(per_session_bw['avg'])} "
-        f"per_session_p95={format_num(per_session_bw['p95'])} "
-        f"global={format_num(bandwidth['global_goodput_mbps'])} "
-        f"bucket_avg={format_num(bucket_bw['avg_mbps'])} "
-        f"bucket_peak={format_num(bucket_bw['peak_mbps'])}"
-    )
+    print("Progressive streaming: "
+          f"{progressive['success_sessions']}/"
+          f"{progressive['eligible_sessions']} ({ratio_text})")
+    print("Bandwidth Mbps: "
+          f"per_session_avg={format_num(per_session_bw['avg'])} "
+          f"per_session_p95={format_num(per_session_bw['p95'])} "
+          f"global={format_num(bandwidth['global_goodput_mbps'])} "
+          f"bucket_avg={format_num(bucket_bw['avg_mbps'])} "
+          f"bucket_peak={format_num(bucket_bw['peak_mbps'])}")
     print(f"VERDICT: {'PASS' if summary['passed'] else 'FAIL'}")
     if summary["fail_reasons"]:
         print("Fail reasons:")
@@ -670,10 +588,8 @@ def distribute_sessions(total_clients: int, process_count: int) -> list[int]:
 
 def run_stress(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
     if websockets is None:
-        raise RuntimeError(
-            "Missing dependency: websockets. Install it before running this "
-            "stress test."
-        )
+        raise RuntimeError("Missing dependency: websockets. Install it before running this "
+                           "stress test.")
 
     preset_file = Path(args.preset_file).expanduser().resolve()
     selected_preset_id, curated_prompts, total_prompt_count = load_curated_prompts(
@@ -735,13 +651,8 @@ def run_stress(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
 
         start_event.set()
 
-        result_deadline = (
-            time.monotonic()
-            + args.connect_timeout_s
-            + args.session_timeout_s
-            + args.post_complete_wait_s
-            + 180.0
-        )
+        result_deadline = (time.monotonic() + args.connect_timeout_s + args.session_timeout_s +
+                           args.post_complete_wait_s + 180.0)
         worker_results: list[dict[str, Any]] = []
         while len(worker_results) < len(processes):
             timeout_s = max(0.1, result_deadline - time.monotonic())
@@ -765,24 +676,20 @@ def run_stress(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
             if result.get("status") == "ok":
                 sessions.extend(result.get("sessions", []))
             else:
-                worker_errors.append(
-                    {
-                        "worker_id": result.get("worker_id"),
-                        "error": result.get("error"),
-                        "traceback": result.get("traceback"),
-                    }
-                )
+                worker_errors.append({
+                    "worker_id": result.get("worker_id"),
+                    "error": result.get("error"),
+                    "traceback": result.get("traceback"),
+                })
 
         received_workers = {result.get("worker_id") for result in worker_results}
         expected_workers = set(range(len(processes)))
         missing_workers = sorted(expected_workers - received_workers)
         for worker_id in missing_workers:
-            worker_errors.append(
-                {
-                    "worker_id": worker_id,
-                    "error": "No worker result received.",
-                }
-            )
+            worker_errors.append({
+                "worker_id": worker_id,
+                "error": "No worker result received.",
+            })
 
         run_end_epoch = time.time()
         run_end_iso = iso_from_epoch(run_end_epoch)
@@ -795,9 +702,8 @@ def run_stress(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
 
         if worker_errors:
             summary["passed"] = False
-            summary["fail_reasons"] = list(summary["fail_reasons"]) + [
-                f"{len(worker_errors)} worker error(s) occurred."
-            ]
+            summary["fail_reasons"] = list(
+                summary["fail_reasons"]) + [f"{len(worker_errors)} worker error(s) occurred."]
 
         output_payload = {
             "run_info": {
@@ -833,9 +739,7 @@ def run_stress(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
 
 
 def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(
-        description="Multiprocess realtime stress test for LTX2 streaming.",
-    )
+    parser = argparse.ArgumentParser(description="Multiprocess realtime stress test for LTX2 streaming.", )
     parser.add_argument(
         "-u",
         "--url",

apps/dreamverse/dreamverse/tests/test_session_init_image.py

@@ -47,13 +47,11 @@ def test_persist_session_init_image_returns_none_when_missing_data():
 
 def test_persist_session_init_image_rejects_unsupported_mime():
     with pytest.raises(ValueError, match="PNG, JPEG, or WebP"):
-        persist_session_init_image(
-            {
-                "name": "frame.gif",
-                "mime_type": "image/gif",
-                "data_url": "data:image/gif;base64,R0lGODlhAQABAAAAACw=",
-            }
-        )
+        persist_session_init_image({
+            "name": "frame.gif",
+            "mime_type": "image/gif",
+            "data_url": "data:image/gif;base64,R0lGODlhAQABAAAAACw=",
+        })
 
 
 def test_persist_session_init_image_rejects_large_payload(monkeypatch):
@@ -66,10 +64,8 @@ def test_persist_session_init_image_rejects_large_payload(monkeypatch):
     monkeypatch.setattr(base64, "b64decode", fake_b64decode)
 
     with pytest.raises(ValueError, match="15 MB or smaller"):
-        persist_session_init_image(
-            {
-                "name": "frame.png",
-                "mime_type": "image/png",
-                "data_url": data_url,
-            }
-        )
+        persist_session_init_image({
+            "name": "frame.png",
+            "mime_type": "image/png",
+            "data_url": data_url,
+        })

apps/dreamverse/dreamverse/worker_ipc.py

@@ -15,6 +15,8 @@ from __future__ import annotations
 
 from dataclasses import dataclass
 
+from dreamverse.generation_inputs import GenerationInputs
+
 # ---- User-scoped events (carry user_id) ------------------------------------
 
 
@@ -147,6 +149,7 @@ class UserStepPayload:
     segment_idx: int
     image_path: str | None
     reset_conditioning: bool
+    generation_inputs: GenerationInputs | None = None
 
 
 @dataclass(frozen=True)

apps/dreamverse/gpu-pool.drawio

@@ -70,7 +70,7 @@
                 <mxCell id="dispatcher" value="command dispatcher&#xa;&#xa;gpu_worker_process() branches on&#xa;CommandType; asserts payload type&#xa;&#xa;INIT / WARMUP / RELOAD_MODEL&#xa;USER_JOIN / USER_STEP / USER_LEAVE&#xa;SHUTDOWN" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffe6cc;strokeColor=#d79b00;fontSize=11;align=left;spacingLeft=10;spacingTop=8;fontStyle=1;" parent="1" vertex="1">
                     <mxGeometry x="120" y="1120" width="240" height="120" as="geometry"/>
                 </mxCell>
-                <mxCell id="do_step" value="VideoGenerationWorker.generate_step()&#xa;video_generation.py:380&#xa;&#xa;reads + updates ContinuationState,&#xa;calls generator" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;align=left;spacingLeft=10;spacingTop=8;fontStyle=1;" parent="1" vertex="1">
+                <mxCell id="do_step" value="VideoGenerationWorker.generate_step()&#xa;ltx2_generation.py:380&#xa;&#xa;reads + updates ContinuationState,&#xa;calls generator" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;align=left;spacingLeft=10;spacingTop=8;fontStyle=1;" parent="1" vertex="1">
                     <mxGeometry x="460" y="1120" width="240" height="120" as="geometry"/>
                 </mxCell>
                 <mxCell id="stream_av" value="stream_fmp4()&#xa;av_streaming.py:121&#xa;&#xa;trims overlap, pipes to ffmpeg,&#xa;publishes StreamInit / StreamChunk /&#xa;StreamComplete via callback" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#b1d8d7;strokeColor=#23445d;fontSize=11;align=left;spacingLeft=10;spacingTop=8;fontStyle=1;" parent="1" vertex="1">
@@ -79,13 +79,13 @@
                 <mxCell id="Ot8BU52QTIb4EhyRSe7I-2" value="" style="edgeStyle=none;html=1;" parent="1" source="generator" target="Ot8BU52QTIb4EhyRSe7I-1" edge="1">
                     <mxGeometry relative="1" as="geometry"/>
                 </mxCell>
-                <mxCell id="generator" value="VideoGenerator  (fastvideo)&#xa;&#xa;LTX2 DiT + refine upsampler&#xa;FP4 quant, torch.compile&#xa;&#xa;owned by VideoGenerationWorker&#xa;video_generation.py:211" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;" parent="1" vertex="1">
+                <mxCell id="generator" value="VideoGenerator  (fastvideo)&#xa;&#xa;LTX2 DiT + refine upsampler&#xa;FP4 quant, torch.compile&#xa;&#xa;owned by VideoGenerationWorker&#xa;ltx2_generation.py:211" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;" parent="1" vertex="1">
                     <mxGeometry x="460" y="1300" width="240" height="100" as="geometry"/>
                 </mxCell>
                 <mxCell id="ffmpeg" value="ffmpeg subprocess&#xa;&#xa;libx264 / *_nvenc&#xa;fragmented mp4" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffffff;strokeColor=#d79b00;fontSize=11;" parent="1" vertex="1">
                     <mxGeometry x="800" y="1300" width="260" height="100" as="geometry"/>
                 </mxCell>
-                <mxCell id="caches" value="ContinuationState&#xa;video_generation.py:89&#xa;&#xa;• video_images: list[PIL.Image]&#xa;• audio_latents: torch.Tensor (CPU)&#xa;&#xa;carried across segments" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;align=left;spacingLeft=10;spacingTop=8;" parent="1" vertex="1">
+                <mxCell id="caches" value="ContinuationState&#xa;ltx2_generation.py:89&#xa;&#xa;• video_images: list[PIL.Image]&#xa;• audio_latents: torch.Tensor (CPU)&#xa;&#xa;carried across segments" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;fontSize=11;align=left;spacingLeft=10;spacingTop=8;" parent="1" vertex="1">
                     <mxGeometry x="120" y="1300" width="240" height="100" as="geometry"/>
                 </mxCell>
                 <mxCell id="e_cp" value="acquire" style="edgeStyle=orthogonalEdgeStyle;rounded=0;html=1;strokeColor=#6c8ebf;endArrow=classic;fontSize=11;exitX=0.5;exitY=1;exitDx=0;exitDy=0;entryX=0.5;entryY=0;entryDx=0;entryDy=0;" parent="1" source="client" target="pool" edge="1">
@@ -250,7 +250,7 @@
                         <mxPoint x="690" y="880"/>
                     </Array>
                 </mxCell>
-                <mxCell id="legend" value="Legend&#xa;&#xa;■ blue  client / external&#xa;■ green  main-process pool/slot&#xa;    (methods — italic label)&#xa;■ yellow  containers (routing state)&#xa;■ red  IPC primitives (mp.Queue, mp.RawArray)&#xa;&#xa;Worker subprocess modules:&#xa;■ orange  gpu_pool.py (dispatcher)&#xa;■ lavender  video_generation.py&#xa;■ teal  av_streaming.py&#xa;■ gray  worker_ipc.py (shared types)&#xa;&#xa;Flow:&#xa;  client → pool → slot&#xa;  → _send_command(_tagged) → command_queue&#xa;  → dispatcher → generate_step()&#xa;  → stream_fmp4() → ffmpeg&#xa;  → shared_buf + response_queue&#xa;  → _response_reader → futures / stream_queues&#xa;  → client awaits (via main.py AV loop)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#999999;fontSize=11;align=left;spacingLeft=10;spacingTop=8;" parent="1" vertex="1">
+                <mxCell id="legend" value="Legend&#xa;&#xa;■ blue  client / external&#xa;■ green  main-process pool/slot&#xa;    (methods — italic label)&#xa;■ yellow  containers (routing state)&#xa;■ red  IPC primitives (mp.Queue, mp.RawArray)&#xa;&#xa;Worker subprocess modules:&#xa;■ orange  gpu_pool.py (dispatcher)&#xa;■ lavender  ltx2_generation.py&#xa;■ teal  av_streaming.py&#xa;■ gray  worker_ipc.py (shared types)&#xa;&#xa;Flow:&#xa;  client → pool → slot&#xa;  → _send_command(_tagged) → command_queue&#xa;  → dispatcher → generate_step()&#xa;  → stream_fmp4() → ffmpeg&#xa;  → shared_buf + response_queue&#xa;  → _response_reader → futures / stream_queues&#xa;  → client awaits (via main.py AV loop)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#999999;fontSize=11;align=left;spacingLeft=10;spacingTop=8;" parent="1" vertex="1">
                     <mxGeometry x="39" y="-200" width="270" height="380" as="geometry"/>
                 </mxCell>
                 <mxCell id="Ot8BU52QTIb4EhyRSe7I-1" value="FastVideo video_generator" style="whiteSpace=wrap;html=1;fontSize=11;fillColor=#e1d5e7;strokeColor=#9673a6;rounded=1;" parent="1" vertex="1">
@@ -389,10 +389,10 @@
                 <mxCell id="cw2" value="from fastvideo.entrypoints.video_generator import VideoGenerator&#xa;from fastvideo.models.dits.ltx2 import DEFAULT_LTX2_AUDIO_*&#xa;&#xa;** Dreamverse reaches into fastvideo internals here **" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffe0b2;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
                     <mxGeometry x="675" y="695" width="550" height="60" as="geometry"/>
                 </mxCell>
-                <mxCell id="cw3" value="on Command(INIT):&#xa;   VideoGenerationWorker.initialize()    (video_generation.py:247)&#xa;      maybe_download_model(model_id)&#xa;      VideoGenerator.from_pretrained(path, FP4Config, PipelineConfig)&#xa;      load audio VAE, resolve refine upsampler&#xa;      resp_q.put(InitAck(success=True))" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffffff;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
+                <mxCell id="cw3" value="on Command(INIT):&#xa;   VideoGenerationWorker.initialize()    (ltx2_generation.py:247)&#xa;      maybe_download_model(model_id)&#xa;      VideoGenerator.from_pretrained(path, FP4Config, PipelineConfig)&#xa;      load audio VAE, resolve refine upsampler&#xa;      resp_q.put(InitAck(success=True))" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffffff;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
                     <mxGeometry x="675" y="765" width="550" height="95" as="geometry"/>
                 </mxCell>
-                <mxCell id="cw4" value="on Command(WARMUP) with WarmupPayload:&#xa;   VideoGenerationWorker.warmup(payload.prompt)    (video_generation.py:518)&#xa;      two synthetic segments prime caches + torch.compile&#xa;   resp_q.put(WarmupComplete(timings=...))" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffffff;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
+                <mxCell id="cw4" value="on Command(WARMUP) with WarmupPayload:&#xa;   VideoGenerationWorker.warmup(payload.prompt)    (ltx2_generation.py:518)&#xa;      two synthetic segments prime caches + torch.compile&#xa;   resp_q.put(WarmupComplete(timings=...))" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffffff;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
                     <mxGeometry x="675" y="870" width="550" height="55" as="geometry"/>
                 </mxCell>
                 <mxCell id="cw5" value="enter main worker loop → waits for JOIN_USER / USER_STEP / LEAVE" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#c8e6c9;strokeColor=#388e3c;fontSize=11;fontStyle=1;fontFamily=monospace;" parent="1" vertex="1">
@@ -534,7 +534,7 @@
                         <mxPoint x="1040" y="1610" as="targetPoint"/>
                     </mxGeometry>
                 </mxCell>
-                <mxCell id="dm11a" value="10a. worker runs:&#xa;VideoGenerationWorker.generate_step()&#xa;    (video_generation.py:380)&#xa;  → generator.generate_video()&#xa;  → updates ContinuationState&#xa;then stream_fmp4() (av_streaming.py:121)&#xa;  → ffmpeg (rawvideo+wav → fmp4)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffe0b2;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
+                <mxCell id="dm11a" value="10a. worker runs:&#xa;VideoGenerationWorker.generate_step()&#xa;    (ltx2_generation.py:380)&#xa;  → generator.generate_video()&#xa;  → updates ContinuationState&#xa;then stream_fmp4() (av_streaming.py:121)&#xa;  → ffmpeg (rawvideo+wav → fmp4)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ffe0b2;strokeColor=#d79b00;fontSize=10;align=left;spacingLeft=8;fontFamily=monospace;" parent="1" vertex="1">
                     <mxGeometry x="955" y="1640" width="180" height="70" as="geometry"/>
                 </mxCell>
                 <mxCell id="dm11" value="10b. resp_q.put(MediaInit / MediaChunk / MediaComplete / StepComplete)" style="endArrow=classic;html=1;strokeColor=#b85450;fontSize=10;labelBackgroundColor=#ffffff;" parent="1" edge="1">

apps/dreamverse/scripts/launch/launch_backend_dreamverse.sh

@@ -39,6 +39,17 @@ export FASTVIDEO_GENERATION_SEGMENT_CAP="${FASTVIDEO_GENERATION_SEGMENT_CAP:-6}"
 export FASTVIDEO_PROMPT_AUTO_SLEEP_MS="${FASTVIDEO_PROMPT_AUTO_SLEEP_MS:-120}"
 export FASTVIDEO_PROMPT_AUTO_TIMEOUT_MS="${FASTVIDEO_PROMPT_AUTO_TIMEOUT_MS:-1800}"
 
+if [[ "${ENABLE_TORCH_COMPILE}" == "1" ]]; then
+  # Persist Inductor, AOTAutograd, and Triton artifacts across launches.
+  export DREAMVERSE_TORCH_COMPILE_CACHE_ROOT="${DREAMVERSE_TORCH_COMPILE_CACHE_ROOT:-${HOME}/.cache/dreamverse/torch_compile}"
+  export TORCHINDUCTOR_CACHE_DIR="${TORCHINDUCTOR_CACHE_DIR:-${DREAMVERSE_TORCH_COMPILE_CACHE_ROOT}/inductor}"
+  export TRITON_CACHE_DIR="${TRITON_CACHE_DIR:-${DREAMVERSE_TORCH_COMPILE_CACHE_ROOT}/triton}"
+  export TORCHINDUCTOR_FX_GRAPH_CACHE="${TORCHINDUCTOR_FX_GRAPH_CACHE:-1}"
+  export TORCHINDUCTOR_AUTOGRAD_CACHE="${TORCHINDUCTOR_AUTOGRAD_CACHE:-1}"
+  mkdir -p "${TORCHINDUCTOR_CACHE_DIR}" "${TRITON_CACHE_DIR}"
+  echo "[launch-demo] torch.compile cache: ${DREAMVERSE_TORCH_COMPILE_CACHE_ROOT}"
+fi
+
 cd "${DREAMVERSE_ROOT}"
 
 if ! command -v dreamverse-server >/dev/null 2>&1; then

apps/dreamverse/scripts/modal/modal_app.py

@@ -8,12 +8,10 @@ import modal
 
 IMAGE = os.environ.get("DREAMVERSE_IMAGE")
 if not IMAGE:
-    raise RuntimeError(
-        "DREAMVERSE_IMAGE is required. Set it to a published SHA-specific Dreamverse image, "
-        "for example a dreamverse-backend-cuda13.0.0-sha-* tag or a "
-        "dreamverse-ui-cuda13.0.0-sha-* tag if serving the static UI. "
-        "CUDA 12 / cu126 images use the corresponding cuda12.6.3 tag."
-    )
+    raise RuntimeError("DREAMVERSE_IMAGE is required. Set it to a published SHA-specific Dreamverse image, "
+                       "for example a dreamverse-backend-cuda13.0.0-sha-* tag or a "
+                       "dreamverse-ui-cuda13.0.0-sha-* tag if serving the static UI. "
+                       "CUDA 12 / cu126 images use the corresponding cuda12.6.3 tag.")
 
 # ``@modal.web_server`` invokes ``serve()`` directly and bypasses the image
 # ENTRYPOINT (``docker/docker_entrypoint.sh``).  That entrypoint normally
@@ -65,14 +63,10 @@ def serve():
     # ``or ""`` collapses ``None`` (unset) into an empty string, ``.strip()``
     # collapses whitespace-only values (e.g. ``"   "``) — both should be
     # treated as missing.
-    missing = [
-        k for k in _REQUIRED_SECRET_KEYS
-        if not (os.environ.get(k) or "").strip()
-    ]
+    missing = [k for k in _REQUIRED_SECRET_KEYS if not (os.environ.get(k) or "").strip()]
     if missing:
-        raise RuntimeError(
-            "dreamverse-api-keys secret is missing required entries: "
-            f"{', '.join(missing)}. Add them with `modal secret create "
-            "dreamverse-api-keys ... --force` and redeploy "
-            "(see apps/dreamverse/scripts/modal/README.md).")
+        raise RuntimeError("dreamverse-api-keys secret is missing required entries: "
+                           f"{', '.join(missing)}. Add them with `modal secret create "
+                           "dreamverse-api-keys ... --force` and redeploy "
+                           "(see apps/dreamverse/scripts/modal/README.md).")
     subprocess.Popen(["dreamverse-server", "--host", "0.0.0.0", "--port", "8009"])

apps/dreamverse/scripts/slurm/README.md

@@ -0,0 +1,107 @@
+# Dreamverse on Slurm
+
+Run Full H3 inside a one-node, four-GPU allocation. The maintained H3 examples
+default to four GPUs; this is a starting configuration, not a measured minimum.
+The full checkpoint supports T2VA, FL2VA, and Ref2VA. The FastH3 Preview profile
+is a separate T2VA configuration.
+
+`launch_backend.sh` checks that it is inside an `srun` step, preserves
+`CUDA_VISIBLE_DEVICES`, and replaces itself with the backend process. It does
+not allocate GPUs, kill existing processes, or source a personal credentials
+file. The local `dreamverse-deploy` helper is not suitable for a shared Slurm
+cluster because it kills processes by physical GPU and port.
+
+## Prepare and allocate
+
+Keep the checkout, weights, outputs, and logs on storage visible to the compute
+node. Source installation is documented in the [GPU guide](../../../../docs/getting_started/installation/gpu.md).
+On ARM64 GB200 use CUDA 13, a matching PyTorch build, and kernels built for
+`sm_100`; the DGX Spark `sm_121` kernel image is not the GB200 image.
+
+The repository's image workflow publishes an ARM64 GB200 variant under
+`ghcr.io/hao-ai-lab/fastvideo/fastvideo-dev:py3.12-cuda13.0.0-sm100-latest`.
+Resolve that tag to a digest for reproducible runs. If your compute nodes use
+Pyxis/Enroot, pass the approved image or a prepared SquashFS file to
+`srun --container-image`, with explicit mounts for your checkout and model cache.
+The Dreamverse-specific Docker images are currently AMD64-only.
+
+For the Slinky customer partition, a bounded allocation is:
+
+```bash
+salloc --account=customer --qos=normal --partition=hpc-rack-1 \
+  --nodes=1 --ntasks=1 --cpus-per-task=72 --gres=gpu:nvidia_gb200:4 \
+  --mem=800G --time=02:00:00 --job-name=dreamverse
+srun --ntasks=1 --pty bash
+```
+
+Wait for Slurm to grant the allocation before entering the compute step. A
+successful SSH login does not grant GPU resources. Inspect pending capacity
+with `squeue -u "$USER" --start`; do not attach to another user's job.
+
+The checkpoint includes duplicate release layouts. Download the diffusers
+components needed by both base and reference pipelines, rather than the whole
+repository (about 210 GB versus about 498 GB at revision
+`42ed227ee7df40d41602854ae760620d6eb651fe`):
+
+```bash
+hf download MiniMaxAI/MiniMax-H3 \
+  --revision 42ed227ee7df40d41602854ae760620d6eb651fe \
+  --include model_index.json --include modular_model_index.json \
+  --include 'audio_scheduler/*' --include 'audio_vae/*' \
+  --include 'processor/*' --include 'scheduler/*' \
+  --include 'text_encoder/*' --include 'tokenizer/*' \
+  --include 'transformer/*' --include 'transformer_ref/*' --include 'vae/*' \
+  --local-dir /path/to/models/MiniMax-H3
+```
+
+The GPU environment needs `fastvideo[dreamverse]`, the Dreamverse workspace
+package, and FFmpeg with H.264/AAC encoders. In a prepared FastVideo image,
+install the checked-out code and its Dreamverse dependencies in that image's
+Python environment. Keep its matching CUDA/PyTorch/kernel stack intact.
+
+## Start and connect
+
+From the checked-out repository inside the allocated step:
+
+```bash
+export DREAMVERSE_PYTHON=/path/to/environment/bin/python
+export DREAMVERSE_MODEL_PATH=/path/to/models/MiniMax-H3
+export FASTVIDEO_DREAMVERSE_HOME=/path/to/persistent/dreamverse-state
+bash apps/dreamverse/scripts/slurm/launch_backend.sh
+```
+
+The default backend binds port 8009 on the private compute node. Connect through
+the login node from your laptop, replacing `COMPUTE_NODE_IP` with the allocated
+node's `NodeAddr` from `scontrol show node`:
+
+```bash
+ssh -N -L 8009:COMPUTE_NODE_IP:8009 USER@LOGIN_NODE
+```
+
+In another laptop terminal, run the frontend from your local checkout:
+
+```bash
+cd apps/dreamverse/web
+BACKEND_HOST=127.0.0.1 BACKEND_PORT=8009 npm run dev
+```
+
+Open `http://localhost:5299`. `/healthz` reports the server process; `/readyz`
+reports model readiness. Full H3 loads and generates more slowly than the
+Preview adapter. Keep prompt enhancement disabled in the UI unless the
+runtime has the selected provider's credentials.
+
+## Verify and stop
+
+Check all three modes with small, valid user-owned assets. Capture the selected
+mode and assets, WebSocket errors or completion events, the generated video and
+audio, and GPU memory usage. Also verify actionable validation errors and
+backward compatibility with clients that omit `generation_mode`.
+
+Use the frontend Playwright instructions in the
+[Dreamverse development guide](../../../../docs/contributing/dreamverse-development.md)
+against the forwarded backend. A mock-server demo validates UI and protocol
+behavior; it is not evidence of GPU generation.
+
+Stop the backend with Ctrl-C, exit the compute step, and release your allocation.
+For a detached allocation, use `scancel YOUR_JOB_ID`. Cancel a pending demo job
+when it is no longer needed; do not leave an unattended reservation queued.

apps/dreamverse/scripts/slurm/launch_backend.sh

@@ -0,0 +1,49 @@
+#!/usr/bin/env bash
+# Run inside an existing Slurm step. Slurm owns the GPU visibility and lifetime.
+set -euo pipefail
+
+if [[ -z "${SLURM_JOB_ID:-}" || -z "${SLURM_STEP_ID:-}" ]]; then
+  echo "Run this launcher inside an allocated Slurm step (srun), not on the login node." >&2
+  exit 2
+fi
+
+script_dir="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+repo_root="$(cd -- "${script_dir}/../../../.." && pwd)"
+python_bin="${DREAMVERSE_PYTHON:-${repo_root}/.venv/bin/python}"
+if [[ ! -x "${python_bin}" ]]; then
+  echo "Set DREAMVERSE_PYTHON to a Python environment with fastvideo[dreamverse] installed." >&2
+  exit 2
+fi
+
+export DREAMVERSE_MODEL_ID="${DREAMVERSE_MODEL_ID:-full-h3}"
+export DREAMVERSE_SP_SIZE="${DREAMVERSE_SP_SIZE:-4}"
+export FASTVIDEO_GPU_COUNT="${FASTVIDEO_GPU_COUNT:-${DREAMVERSE_SP_SIZE}}"
+export FASTVIDEO_ENABLE_STARTUP_WARMUP="${FASTVIDEO_ENABLE_STARTUP_WARMUP:-0}"
+export ENABLE_TORCH_COMPILE="${ENABLE_TORCH_COMPILE:-0}"
+export STREAM_MODE="${STREAM_MODE:-av_fmp4}"
+export PYTHONPATH="${repo_root}/apps/dreamverse:${repo_root}${PYTHONPATH:+:${PYTHONPATH}}"
+export PYTHONUNBUFFERED=1
+
+"${python_bin}" - <<'PY'
+import os
+import shutil
+
+import torch
+
+expected = int(os.environ["DREAMVERSE_SP_SIZE"])
+visible = torch.cuda.device_count()
+if expected < 1 or visible < expected:
+    raise SystemExit(f"The Slurm step exposes {visible} GPUs; DREAMVERSE_SP_SIZE requires {expected}.")
+ffmpeg = os.environ.get("FASTVIDEO_FFMPEG_BIN", "ffmpeg")
+if not shutil.which(ffmpeg):
+    raise SystemExit("FFmpeg is missing; install it in the compute environment or set FASTVIDEO_FFMPEG_BIN.")
+print(f"Slurm job {os.environ['SLURM_JOB_ID']}: {visible} visible GPUs; using {expected} per worker")
+for index in range(expected):
+    properties = torch.cuda.get_device_properties(index)
+    print(f"  GPU {index}: {properties.name}, {properties.total_memory / 2**30:.1f} GiB")
+PY
+
+cd "${repo_root}"
+exec "${python_bin}" -m dreamverse.server_entry \
+  --host "${DREAMVERSE_BIND_HOST:-0.0.0.0}" \
+  --port "${DREAMVERSE_BACKEND_PORT:-8009}" "$@"

apps/dreamverse/web/e2e/generation-modes.spec.ts

@@ -0,0 +1,126 @@
+import { execFileSync } from "node:child_process";
+import { readFile } from "node:fs/promises";
+import path from "node:path";
+import { test, expect } from "@playwright/test";
+
+const imagePath = path.resolve("public/k2.png");
+const framePrompt = "A paper fox walks through a sunlit forest, gentle birdsong.";
+
+function makeAudio(sampleRate = 8000, seconds = 1): Buffer {
+	const sampleCount = sampleRate * seconds;
+	const bytes = Buffer.alloc(44 + sampleCount * 2);
+	bytes.write("RIFF", 0); bytes.writeUInt32LE(bytes.length - 8, 4); bytes.write("WAVEfmt ", 8);
+	bytes.writeUInt32LE(16, 16); bytes.writeUInt16LE(1, 20); bytes.writeUInt16LE(1, 22);
+	bytes.writeUInt32LE(sampleRate, 24); bytes.writeUInt32LE(sampleRate * 2, 28);
+	bytes.writeUInt16LE(2, 32); bytes.writeUInt16LE(16, 34); bytes.write("data", 36);
+	bytes.writeUInt32LE(sampleCount * 2, 40);
+	for (let i = 0; i < sampleCount; i++) bytes.writeInt16LE(Math.round(Math.sin(i * 440 * 2 * Math.PI / sampleRate) * 1000), 44 + i * 2);
+	return bytes;
+}
+
+test.describe("generation modes through the mock runtime", () => {
+	for (const mode of ["t2va", "fl2va", "ref2va"] as const) {
+		test(`${mode} sends validated assets and plays a clearly labeled sample`, async ({ page, request }, testInfo) => {
+			const response = await request.get("/generation-capabilities");
+			const capabilities = response.ok() ? await response.json() : {};
+			test.skip(capabilities.mock !== true, "This test uses the CPU mock runtime; it must not silently allocate a real GPU.");
+			const sent: Record<string, any>[] = [];
+			const received: Record<string, any>[] = [];
+			page.on("websocket", (socket) => {
+				socket.on("framesent", ({ payload }) => { if (typeof payload === "string") { try { sent.push(JSON.parse(payload)); } catch {} } });
+				socket.on("framereceived", ({ payload }) => { if (typeof payload === "string") { try { received.push(JSON.parse(payload)); } catch {} } });
+			});
+			await page.goto("/");
+			await expect(page.getByText(/Demo runtime · Sample playback only/)).toBeVisible();
+			const modeSelect = page.getByRole("combobox", { name: "Generation mode" });
+			const modeLabel = mode === "ref2va" ? "Ref2VA" : mode.toUpperCase();
+			await modeSelect.click();
+			await page.getByRole("option", { name: modeLabel, exact: true }).click();
+			await expect(modeSelect).toHaveText(modeLabel);
+			await page.getByLabel("Continuation prompt").fill(framePrompt);
+			const uploadedIds: string[] = [];
+			page.on("response", async (uploadResponse) => {
+				if (uploadResponse.request().method() === "POST" && uploadResponse.url().endsWith("/assets") && uploadResponse.ok()) {
+					const asset = await uploadResponse.json().catch(() => null);
+					if (asset?.asset_id) uploadedIds.push(asset.asset_id);
+				}
+			});
+			try {
+				if (mode === "fl2va") {
+					await expect(page.getByRole("button", { name: "Generate", exact: true })).toBeDisabled();
+					await page.locator('input[type="file"]').setInputFiles([
+						{ name: "first-frame.png", mimeType: "image/png", buffer: await readFile(imagePath) },
+						{ name: "last-frame.png", mimeType: "image/png", buffer: await readFile(imagePath) },
+					]);
+					await expect(page.getByRole("option", { name: "first-frame.png", exact: true }).first()).toBeAttached();
+					await page.getByRole("combobox", { name: "First frame", exact: true }).selectOption({ label: "first-frame.png" });
+					await expect(page.getByRole("button", { name: "Generate", exact: true })).toBeEnabled();
+					await page.getByRole("combobox", { name: "Last frame", exact: true }).selectOption({ label: "last-frame.png" });
+				}
+				if (mode === "ref2va") {
+					const video = execFileSync(process.env.FASTVIDEO_FFMPEG_BIN || "ffmpeg", ["-v", "error", "-f", "lavfi", "-i", "color=c=royalblue:s=64x64:r=8", "-t", "1", "-c:v", "libx264", "-pix_fmt", "yuv420p", "-movflags", "frag_keyframe+empty_moov", "-f", "mp4", "pipe:1"]);
+					await page.locator('input[type="file"]').setInputFiles([
+						{ name: "subject.png", mimeType: "image/png", buffer: await readFile(imagePath) },
+						{ name: "motion.mp4", mimeType: "video/mp4", buffer: video },
+						{ name: "sound.wav", mimeType: "audio/wav", buffer: makeAudio() },
+					]);
+					await expect(page.getByRole("button", { name: "Add sound.wav as reference" })).toBeEnabled();
+					await page.getByRole("button", { name: "Add sound.wav as reference" }).click();
+					await expect(page.getByRole("button", { name: "Generate", exact: true })).toBeDisabled();
+					await page.getByRole("button", { name: "Add subject.png as reference" }).click();
+					await page.getByRole("button", { name: "Add motion.mp4 as reference" }).click();
+					await page.getByRole("button", { name: "Move sound.wav down" }).click();
+					const names = await page.getByRole("list", { name: "Ordered references" }).locator("li p.font-medium").allTextContents();
+					expect(names).toEqual(["subject.png", "sound.wav", "motion.mp4"]);
+				}
+				await page.screenshot({ path: testInfo.outputPath(`${mode}-inputs.png`), fullPage: true });
+				await page.getByRole("button", { name: "Generate", exact: true }).click();
+				await expect.poll(() => sent.find((item) => item.type === "session_init_v2")?.generation_mode).toBe(mode);
+				const init = sent.find((item) => item.type === "session_init_v2")!;
+				expect(init.conditioning_assets.map((item: any) => item.role)).toEqual(mode === "t2va" ? [] : mode === "fl2va" ? ["first_frame", "last_frame"] : ["reference", "reference", "reference"]);
+				if (mode === "ref2va") expect(init.conditioning_assets.map((item: any) => item.asset_id)).toEqual([uploadedIds[0], uploadedIds[2], uploadedIds[1]]);
+				await expect.poll(() => received.find((item) => item.type === "gpu_assigned")?.generation_mode).toBe(mode);
+				await expect.poll(() => received.some((item) => item.type === "media_segment_complete")).toBe(true);
+				await expect(page.getByText(/Demo runtime · Sample playback only/)).toBeVisible();
+				await expect(modeSelect).toHaveCount(0);
+				await expect.poll(async () => page.locator("video:visible").first().evaluate((element: HTMLVideoElement) => element.readyState)).toBeGreaterThanOrEqual(2);
+				await page.screenshot({ path: testInfo.outputPath(`${mode}-playback.png`), fullPage: true });
+				if (mode === "ref2va") {
+					await page.getByRole("button", { name: "Toggle sidebar" }).click();
+					await page.getByRole("button", { name: "New project", exact: true }).click();
+					await expect(modeSelect).toHaveText("T2VA");
+					await modeSelect.click();
+					await page.getByRole("option", { name: "FL2VA", exact: true }).click();
+					await expect(modeSelect).toHaveText("FL2VA");
+					await page.getByRole("combobox", { name: "First frame", exact: true }).selectOption({ label: "subject.png" });
+					await expect(page.getByRole("combobox", { name: "Last frame", exact: true })).toHaveValue("");
+					await page.getByLabel("Continuation prompt").fill("The paper fox explores a new scene.");
+					await page.getByRole("button", { name: "Generate", exact: true }).click();
+					await expect.poll(() => sent.find((item) => item.type === "project_init_v1")?.generation_mode).toBe("fl2va");
+					const secondProject = sent.find((item) => item.type === "project_init_v1")!;
+					expect(secondProject.conditioning_assets).toEqual([{ asset_id: uploadedIds[0], role: "first_frame" }]);
+					expect(sent.filter((item) => item.type === "session_init_v2")).toHaveLength(1);
+					await expect.poll(() => received.filter((item) => item.type === "media_segment_complete").length).toBeGreaterThan(1);
+				}
+			} finally {
+				await page.close();
+				for (const id of uploadedIds) await request.delete(`/assets/${id}`);
+			}
+		});
+	}
+
+	test("proxies a media upload larger than Next's default 10 MiB body limit", async ({ request }) => {
+		const response = await request.get("/generation-capabilities");
+		const capabilities = response.ok() ? await response.json() : {};
+		test.skip(capabilities.mock !== true, "Requires the local mock runtime.");
+		const audio = makeAudio(192000, 29);
+		expect(audio.length).toBeGreaterThan(10 * 1024 * 1024);
+		const upload = await request.post("/assets", {
+			headers: { "Content-Type": "audio/wav", "X-Asset-Name": "large-proxy-check.wav" },
+			data: audio,
+		});
+		expect(upload.status()).toBe(201);
+		const asset = await upload.json();
+		try { expect(asset.size).toBe(audio.length); } finally { await request.delete(`/assets/${asset.asset_id}`); }
+	});
+});

apps/dreamverse/web/next.config.ts

@@ -9,6 +9,8 @@ const configDir = path.dirname(fileURLToPath(import.meta.url));
 const staticExport = process.env.NEXT_OUTPUT_EXPORT === '1';
 
 const nextConfig: NextConfig = {
+  // Next 15.5 name for the dev rewrite-proxy body limit; Next 16 renames it to `proxyClientMaxBodySize`.
+  experimental: { middlewareClientMaxBodySize: 100 * 1024 * 1024 },
   ...(staticExport ? { output: 'export' as const } : {}),
   ...(staticExport ? { images: { unoptimized: true } } : {}),
   outputFileTracingRoot: path.join(configDir, '..', '..', '..'),
@@ -38,6 +40,18 @@ const nextConfig: NextConfig = {
         source: '/router/:path*', 
         destination: `${backendUrl}/router/:path*` 
       },
+      {
+        source: '/generation-capabilities',
+        destination: `${backendUrl}/generation-capabilities`,
+      },
+      {
+        source: '/assets',
+        destination: `${backendUrl}/assets`,
+      },
+      {
+        source: '/assets/:path*',
+        destination: `${backendUrl}/assets/:path*`,
+      },
       {
         source: '/prompt-system-config',
         destination: `${backendUrl}/prompt-system-config`,

apps/dreamverse/web/src/app/page.integration.test.tsx

@@ -589,6 +589,7 @@ describe.skip('App websocket integration', () => {
     });
 
     const initMessage = outbound.find((message) => message.type === 'session_init_v2');
+    expect(initMessage.generation_mode).toBe('t2va');
     expect(initMessage.preset_id).toBe('test_preset');
     expect(initMessage.curated_prompts).toEqual(['segment one', 'segment two']);
     expect(initMessage.enhancement_enabled).toBe(true);
@@ -597,6 +598,42 @@ describe.skip('App websocket integration', () => {
     expect(initMessage.initial_rollout_prompt).toBe('');
   });
 
+  it('sends the selected generation mode and locks it after session start', async () => {
+    const outbound: any[] = [];
+    server.on('connection', (socket) => {
+      socket.on('message', (rawMessage) => {
+        outbound.push(JSON.parse(rawMessage as string));
+      });
+    });
+
+    const user = userEvent.setup();
+    render(<Page />);
+
+    const modeSelect = await screen.findByRole('combobox', { name: 'Generation mode' });
+    expect(modeSelect).toHaveTextContent('T2VA');
+
+    await user.click(modeSelect);
+    await user.click(await screen.findByRole('option', { name: 'FL2VA' }));
+    expect(modeSelect).toHaveTextContent('FL2VA');
+    expect(modeSelect).toHaveAttribute(
+      'title',
+      'First/last frames to video + audio. Start from a first frame image. Add an optional last frame to guide the ending.',
+    );
+
+    const generateButton = await screen.findByRole('button', { name: 'Generate' });
+    await waitFor(() => expect(generateButton).toBeEnabled());
+    await user.click(generateButton);
+
+    await waitFor(() => {
+      expect(outbound.some((message) => message.type === 'session_init_v2')).toBe(true);
+    });
+
+    const initMessage = outbound.find((message) => message.type === 'session_init_v2');
+    expect(initMessage.generation_mode).toBe('fl2va');
+    expect(screen.queryByRole('combobox', { name: 'Generation mode' }))
+      .not.toBeInTheDocument();
+  });
+
   it('starts a streaming session from a custom initial prompt without using curated prompts', async () => {
     const outbound: any[] = [];
     server.on('connection', (socket) => {

apps/dreamverse/web/src/app/page.tsx

@@ -5,6 +5,7 @@ import { Download, Share2 } from "lucide-react";
 import DevtoolsShell from "@/components/devtools/DevtoolsShell";
 import MonitorPage from "@/components/MonitorPage";
 import ChatBar from "@/components/ChatBar";
+import AssetList from "@/components/AssetList";
 import SessionTimeoutModal from "@/components/SessionTimeoutModal";
 import Sidebar from "@/components/Sidebar";
 import Header from "@/components/Header";
@@ -13,10 +14,13 @@ import Workspace from "@/components/Workspace";
 import { saveProject, saveProjectMetadata, listProjects, loadProjectClips, deleteProject, pruneOldProjects, type StoredProject, type StoredClip } from "@/lib/projectStorage";
 import { isInfrastructureError } from "@/lib/ws/reducer";
 import { useStore } from "@/hooks/useStore";
+import { useAssetLibrary } from "@/hooks/useAssetLibrary";
+import { useGenerationCapabilities } from "@/hooks/useGenerationCapabilities";
 import { resolveDevtoolsMode } from "@/lib/devtoolsMode";
 import { createAvPipeline, DEFAULT_AV_MIME } from "@/lib/media/avPipeline";
 import { remuxArchivedFmp4Segments } from "@/lib/media/fmp4Remux";
 import { DEFAULT_CUSTOM_PRESET_ID, parseStoryPresets, sanitizePresetId } from "@/lib/presets";
+import { DEFAULT_GENERATION_MODE, buildGenerationInitFields, validateGenerationInputs, type GenerationMode, type GenerationInitFields, type GenerationAsset } from "@/lib/generationMode";
 import {
 	buildRewritePromptWindowSnapshot,
 	buildRewritePromptWindowSnapshotFromPrompts,
@@ -341,6 +345,19 @@ export default function Page() {
 	const [isMobileShareCapable, setIsMobileShareCapable] = useState(false);
 	const [videoMuted, setVideoMuted] = useState(true);
 	const [timeoutModalOpen, setTimeoutModalOpen] = useState(false);
+	const [generationMode, setGenerationMode] = useState<GenerationMode>(DEFAULT_GENERATION_MODE);
+	const assetLibrary = useAssetLibrary();
+	const { capabilities, capabilityNotice, refreshCapabilities } = useGenerationCapabilities();
+	const joiningRef = useRef(false);
+	const activeGenerationRef = useRef<{ fields: GenerationInitFields; assets: GenerationAsset[]; mock: boolean } | null>(null);
+	const generationInputError = validateGenerationInputs(generationMode, assetLibrary.conditioningAssets, assetLibrary.assets);
+	const generationSupported = capabilities.modes.includes(generationMode);
+	const generationInputsValid = !generationInputError && generationSupported && !assetLibrary.uploading;
+	function changeGenerationMode(mode: GenerationMode) {
+		if (sessionStore.get().sessionStarted || joiningRef.current || !capabilities.modes.includes(mode)) return;
+		setGenerationMode(mode);
+		assetLibrary.clearConditioning();
+	}
 	useEffect(() => {
 		setIsMobileShareCapable(typeof navigator.canShare === "function" && window.matchMedia("(pointer: coarse)").matches);
 	}, []);
@@ -391,7 +408,7 @@ export default function Page() {
 
 	// --- Derived values ---
 
-	const canStartSession = !projectResetPending && (canJoinSession || Boolean(normalizeInitialPrompt(livePromptDraft as string)));
+	const canStartSession = generationInputsValid && !projectResetPending && (canJoinSession || Boolean(normalizeInitialPrompt(livePromptDraft as string)));
 
 	const currentClipLabel = useMemo(() => {
 		if ((activeClip as Record<string, any>)?.label) return (activeClip as Record<string, any>).label;
@@ -747,7 +764,11 @@ export default function Page() {
 
 	function recoverFailedSessionStart(notice: string) {
 		const restoredDraft = normalizeInitialPrompt(pendingInitialPromptRef.current);
-		resetToLobbyState();
+		if (wsRef.current) {
+			detachAndCloseWebSocket(wsRef.current);
+			wsRef.current = null;
+		}
+		resetToLobbyState({ preserveSessionNotice: true });
 		clearPendingProjectPointers();
 		pendingInitialPromptRef.current = "";
 		sessionStore.patch({
@@ -1702,6 +1723,10 @@ export default function Page() {
 
 	function resetToLobbyState({ preserveSessionNotice = false, preservePlayback = false } = {}) {
 		setVideoMuted(true);
+		if (!preserveSessionNotice) {
+			setGenerationMode(DEFAULT_GENERATION_MODE);
+			assetLibrary.clearConditioning();
+		}
 		clearCountdownInterval();
 		pendingInitialPromptRef.current = "";
 		sessionStore.patch({
@@ -1736,6 +1761,8 @@ export default function Page() {
 
 	function resetToProjectLobbyState() {
 		setVideoMuted(true);
+		setGenerationMode(DEFAULT_GENERATION_MODE);
+		assetLibrary.clearConditioning();
 		pendingInitialPromptRef.current = "";
 		sessionStore.patch({
 			sessionStarted: false,
@@ -1764,6 +1791,7 @@ export default function Page() {
 		setSeedPrompts(segmentPrompts);
 		return {
 			type,
+			...(activeGenerationRef.current?.fields || buildGenerationInitFields(generationMode, assetLibrary.conditioningAssets, assetLibrary.assets)),
 			preset_id: getInitialPresetId(),
 			preset_label: getInitialPresetLabel(),
 			curated_prompts: segmentPrompts,
@@ -1812,6 +1840,12 @@ export default function Page() {
 			return;
 		}
 		if (decoded.kind !== "json") return;
+		if (decoded.data?.type === "error" && sessionStore.get().sessionStarted
+			&& (!sessionStore.get().gpuAssigned || decoded.data.error_code === "invalid_generation_input")) {
+			const message = typeof decoded.data.message === "string" ? decoded.data.message : "The generation inputs were rejected. Check the mode and selected assets.";
+			recoverFailedSessionStart(message);
+			return;
+		}
 		if (decoded.data?.type === "error" && isInfrastructureError(decoded.data)) {
 			const message = typeof decoded.data?.message === "string" && decoded.data.message.trim()
 				? decoded.data.message.trim()
@@ -1937,9 +1971,15 @@ export default function Page() {
 		}
 	}
 
-	function beginProjectLocally({ force = false } = {}) {
+	function beginProjectLocally({ force = false, mockRuntime = capabilities.mock === true } = {}) {
 		if (!force && !canStartSession) return;
+		if (!generationInputsValid) return false;
 		if (sessionStore.get().sessionStarted || sessionStore.get().projectResetPending) return false;
+		activeGenerationRef.current = {
+			fields: buildGenerationInitFields(generationMode, assetLibrary.conditioningAssets, assetLibrary.assets),
+			assets: assetLibrary.assets.filter((asset) => assetLibrary.conditioningAssets.some((item) => item.asset_id === asset.asset_id)),
+			mock: mockRuntime,
+		};
 		setTimeoutModalOpen(false);
 		// Unmute during the user gesture so iOS Safari permits audio playback.
 		setVideoMuted(false);
@@ -1999,12 +2039,43 @@ export default function Page() {
 	}
 
 	async function joinSession({ force = false } = {}) {
+		if (joiningRef.current || sessionStore.get().sessionStarted) return;
+		if (generationInputError || assetLibrary.uploading) {
+			showPreSessionNotice(generationInputError || "Wait for the asset upload to finish.");
+			return;
+		}
+		joiningRef.current = true;
+		try {
+			await startGenerationSession({ force });
+		} finally {
+			joiningRef.current = false;
+		}
+	}
+
+	async function startGenerationSession({ force = false } = {}) {
+		sessionStore.patch({ sessionNotice: "" });
+		streamStore.patch({ loadingAnimation: true });
+		const currentCapabilities = await refreshCapabilities();
+		if (!currentCapabilities.modes.includes(generationMode)) {
+			streamStore.patch({ loadingAnimation: false });
+			showPreSessionNotice(`${generationMode.toUpperCase()} is unavailable on this runtime. Connect a full H3 runtime or choose a supported mode.`);
+			return;
+		}
+		const assetProblem = await assetLibrary.verifySelectedAssets();
+		if (assetProblem) {
+			streamStore.patch({ loadingAnimation: false });
+			showPreSessionNotice(assetProblem);
+			return;
+		}
 		if (
 			wsRef.current
 			&& wsRef.current.readyState === WebSocket.OPEN
 			&& sessionStore.get().connected
 		) {
-			if (!beginProjectLocally({ force })) return;
+			if (!beginProjectLocally({ force, mockRuntime: currentCapabilities.mock === true })) {
+				streamStore.patch({ loadingAnimation: false });
+				return;
+			}
 			sendProjectInitMessage();
 			return;
 		}
@@ -2017,7 +2088,7 @@ export default function Page() {
 			showPreSessionNotice(probe.notice);
 			return;
 		}
-		if (!beginProjectLocally({ force })) {
+		if (!beginProjectLocally({ force, mockRuntime: currentCapabilities.mock === true })) {
 			streamStore.patch({ loadingAnimation: false });
 			sessionStore.patch({ connecting: false });
 			return;
@@ -2044,6 +2115,10 @@ export default function Page() {
 			createdAt: currentProjectCreatedAtRef.current || Date.now(),
 			lastThumbnail: currentThumbnail,
 			promptEvents: [...(rewriteStore.get().promptEvents as Record<string, unknown>[])],
+			generationMode: activeGenerationRef.current?.fields.generation_mode || DEFAULT_GENERATION_MODE,
+			conditioningAssets: activeGenerationRef.current?.fields.conditioning_assets || [],
+			assets: activeGenerationRef.current?.assets || [],
+			mock: activeGenerationRef.current?.mock === true,
 		};
 		const clips: StoredClip[] = (streamStore.get().completedClips as any[])
 			.filter((clip: any) => clip?.blob instanceof Blob)
@@ -2478,6 +2553,24 @@ export default function Page() {
 
 	// --- Render ---
 
+	const conditioningPanel = generationMode !== "t2va" && !sessionStarted && !sessionExpired ? (
+		<AssetList
+			mode={generationMode}
+			assets={assetLibrary.assets}
+			conditioning={assetLibrary.conditioningAssets}
+			locked={Boolean(loadingAnimation || projectResetPending)}
+			uploading={assetLibrary.uploading}
+			error={assetLibrary.assetError}
+			validationNotice={generationInputError}
+			onUpload={assetLibrary.uploadAssets}
+			onAssign={assetLibrary.assignAsset}
+			onRemove={assetLibrary.removeAsset}
+			onUnselect={assetLibrary.removeConditioning}
+			onMove={assetLibrary.moveConditioning}
+			onMissing={assetLibrary.checkAssetAvailability}
+		/>
+	) : null;
+
 	if (!runtimeReady) {
 		return null;
 	}
@@ -2501,13 +2594,17 @@ export default function Page() {
 				enhancementEnabled={enhancementEnabled as boolean}
 				autoExtensionEnabled={autoExtensionEnabled as boolean}
 				loopGenerationEnabled={loopGenerationEnabled as boolean}
-				canJoinSession={canJoinSession as boolean}
+				canJoinSession={canStartSession}
 				canSubmitContinuation={canSubmitContinuation}
 				editableMode={editableMode as boolean}
 				demoMode={demoMode as boolean}
 				editableCanJoin={editableCanJoin as boolean}
 				curatedPromptLimit={curatedPromptLimit as number}
 				maxCuratedPromptCount={maxCuratedPromptCount as number}
+				generationMode={generationMode}
+				supportedGenerationModes={capabilities.modes}
+				conditioningPanel={conditioningPanel}
+				onGenerationModeChange={changeGenerationMode}
 				onPresetChange={handlePresetSelectionChange}
 				onEnhancementToggle={handleEnhancementToggle}
 				onCuratedPromptLimitChange={handleCuratedPromptLimitChange}
@@ -2641,7 +2738,12 @@ export default function Page() {
 			/>
 			<Header timeLeft={headerTimeLeft} formatTime={formatTime} onToggleSidebar={() => setSidebarOpen((prev) => !prev)} />
 
-			<div className="relative flex flex-1 min-h-0 flex-col justify-center px-4 pb-2 sm:px-6 sm:pb-12">
+			<div className={cn(
+				"relative flex flex-1 min-h-0 flex-col px-4 pb-2 sm:px-6 sm:pb-12",
+				!isViewingMode && !showActiveProject && generationMode !== "t2va"
+					? "justify-start overflow-y-auto pt-4"
+					: "justify-center",
+			)}>
 				{isViewingMode && (
 					<>
 						{viewingSelectedClip && (
@@ -2680,6 +2782,8 @@ export default function Page() {
 							/>
 						</section>
 						<motion.div layout="position" className="mx-auto w-full max-w-2xl shrink-0" transition={{ type: "spring", stiffness: 200, damping: 25 }}>
+							{viewingProject?.project.mock && <p className="mb-2 text-center text-xs text-violet-600 dark:text-violet-300">Demo sample · This saved clip was not generated by an AI model.</p>}
+							{viewingProject?.project.generationMode && <p className="mb-3 text-center text-xs text-muted-foreground">{viewingProject.project.generationMode.toUpperCase()} · {viewingProject.project.conditioningAssets?.length || 0} saved references. Uploaded originals may expire; your saved video remains available.</p>}
 							<ChatBar sessionStarted={false} viewingReadOnly={true} onStartNewProject={handleStartNewProject} onBackFromViewing={closeViewingProject} />
 						</motion.div>
 					</>
@@ -2759,7 +2863,7 @@ export default function Page() {
 					</section>
 
 					<AnimatePresence>
-						{!showActiveProject && (
+						{!showActiveProject && generationMode === "t2va" && (
 							<motion.div
 								key="hero-tagline"
 								initial={{ opacity: 0 }}
@@ -2785,6 +2889,12 @@ export default function Page() {
 							sessionExpired={sessionExpired as boolean}
 							sessionNotice={sessionNotice as string}
 							projectResetPending={projectResetPending as boolean}
+							generationMode={generationMode}
+							supportedGenerationModes={capabilities.modes}
+							generationInputsValid={generationInputsValid}
+							capabilityNotice={!generationSupported ? `${generationMode.toUpperCase()} is unavailable on this runtime.` : capabilityNotice}
+							mockRuntime={capabilities.mock}
+							conditioningPanel={conditioningPanel}
 							onPresetGenerate={handlePresetGenerate}
 							onContinuationInput={handleLivePromptInput}
 							onContinuationKeydown={handleLivePromptKeydown}
@@ -2792,6 +2902,7 @@ export default function Page() {
 							onSubmitContinuation={submitLivePrompt}
 							onLeave={leaveSession}
 							onStartNewProject={handleStartNewProject}
+							onGenerationModeChange={changeGenerationMode}
 							onSpeechTranscript={handleLivePromptSpeechTranscript}
 							onSpeechInterimChange={handleLivePromptSpeechInterim}
 						/>