[ci] Cherrypick perf regression skills to component time commit

[ci] Add performance baseline reseed skill and failed-run artifacts

.agents/skills/index.jsonl

@@ -7,3 +7,4 @@
 {"name": "search-related-work", "description": "Query the related work index for relevant papers, repos, or comparisons", "path": "search-related-work/SKILL.md", "status": "draft", "trust": "low"}
 {"name": "seed-ssim-references", "description": "Run a new or updated fastvideo/tests/ssim/ test on Modal, pull generated videos, and upload them to FastVideo/ssim-reference-videos so the test has a regression baseline", "path": "seed-ssim-references/SKILL.md", "status": "draft", "trust": "low"}
 {"name": "reseed-ssim-references", "description": "Re-seed (overwrite) HF reference videos for an existing fastvideo/tests/ssim/ test and a single model id on Modal L40S. Always backs up current refs first, regenerates on Modal, pauses for the user to eyeball before-vs-after, then uploads with --force scoped to --model-id. Sister skill to seed-ssim-references; use when intentional code change has invalidated existing refs", "path": "reseed-ssim-references/SKILL.md", "status": "draft", "trust": "low"}
+{"name": "reseed-performance-baseline", "description": "Re-seed the HF performance-tracking baseline for an intentional runtime, dependency, or environment-caused benchmark shift. 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 must be advanced by replicating one reviewed shifted source result into three success=true records, or five records when explicitly requested", "path": "reseed-performance-baseline/SKILL.md", "status": "draft", "trust": "low"}

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

@@ -0,0 +1,426 @@
+---
+name: reseed-performance-baseline
+description: Re-seed the HF performance-tracking baseline for an intentional runtime, dependency, or environment-caused benchmark shift. 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 by replicating one reviewed shifted source result into three success=true records, or five records when explicitly requested.
+---
+
+# 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`.
+
+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`.
+
+For a 5-record median, one shifted record is not enough to move the median if
+the other four records are from the old runtime. This skill therefore creates
+3 reviewed `success=true` records from one accepted shifted source result by
+default. If the user explicitly asks for a full reset, create 5 records.
+
+These replicated records are an intentional operator-approved baseline reset,
+not independent measurements. Mark them clearly with provenance fields so the
+HF history remains auditable.
+
+Use this skill when a performance test fails for an intentional and reviewed
+reason, such as a torch/runtime/container upgrade that legitimately increases
+peak memory or changes timings. This is the performance equivalent of
+`reseed-ssim-references`: backup first, scope tightly, require explicit human
+approval, then upload reviewed accepted baseline records.
+
+## When to use
+
+- A PR or main run failed the rolling performance comparison by more than the
+  allowed regression threshold, and maintainers agree the shift is caused by
+  an intentional runtime, dependency, hardware image, or benchmark environment
+  change rather than a FastVideo logic regression.
+- One shifted source result has been reviewed and accepted, and the operator
+  wants to replicate it into 3 successful records so the rolling median moves
+  immediately. Use 5 records only when the user explicitly asks to fully reset
+  the last-5 window.
+
+## When not to use
+
+- The benchmark failure might be a real code regression. Fix or investigate
+  the code path first.
+- The fixed benchmark thresholds in
+  `.buildkite/performance-benchmarks/tests/*.json` are too low. Those are a
+  separate gate from the rolling HF baseline and may need a code review change.
+- There is no clear source run, commit, and rationale. Baseline history is a
+  production signal; do not edit it without provenance.
+
+## Inputs
+
+| 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. |
+| `source_result` | Yes | Path or Buildkite artifact URL for one accepted shifted performance JSON. Prefer the normalized `normalized_perf_*.json` artifact emitted by `compare_baseline.py`. |
+| `replica_count` | No | Number of success records to create from `source_result`. Default: `3`. Only use `5` if the user explicitly asks for a full reset. |
+| `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. |
+
+Hardcoded defaults:
+
+- HF repo: `FastVideo/performance-tracking` (`HF_REPO_ID` override is
+  supported by the code, but use the default unless the user explicitly asks).
+- Local sync root: `/tmp/perf-tracking` or a timestamped local backup under
+  `performance_reseed_backup/`.
+- Baseline window: last 5 `success=true` records for the same
+  `(model_id, gpu_type)`.
+- Default reseed count: 3 replicated `success=true` records from one reviewed
+  source result. Explicit full-reset count: 5.
+
+## Steps
+
+### 1. Validate the target and source result
+
+If `source_result` is a Buildkite artifact URL, download it first into a
+local scratch directory such as `performance_reseed_source/` and use that
+downloaded JSON path for the rest of the workflow. If the agent cannot access
+the artifact because Buildkite authentication is missing, ask the user to
+download the artifact manually and provide the local path.
+
+Prefer the normalized Buildkite artifact emitted by `compare_baseline.py`:
+
+```text
+perf_reports/results/normalized_perf_*.json
+```
+
+That file is already in the HF tracking schema. Load it directly and confirm
+it has the expected baseline fields:
+
+```python
+import json
+
+with open(source_result, encoding="utf-8") as f:
+    record = json.load(f)
+```
+
+If only the older raw `fastvideo/tests/performance/results/perf_*.json`
+artifact is available, normalize it with `compare_baseline.py`'s shared helper
+before continuing. Run this from the repository root with
+`PYTHONPATH=fastvideo/tests/performance` so the script-local `hf_store` import
+resolves the same way it does in CI:
+
+```python
+import json
+from compare_baseline import normalize_performance_result
+
+with open(source_result, encoding="utf-8") as f:
+    record = normalize_performance_result(json.load(f))
+```
+
+The raw-to-normalized helper maps:
+
+- `model_id` comes from `benchmark_id`.
+- `gpu_type` comes from `device`.
+- `memory` comes from `max_peak_memory_mb`.
+- `latency` comes from `avg_generation_time_s`.
+- `throughput` comes from `throughput_fps`.
+- component timings come from the raw `text_encoder_time_s`, `dit_time_s`,
+  and `vae_decode_time_s` fields when present. If an older raw artifact lacks
+  those keys, they normalize to `None`; that source can still reseed latency,
+  throughput, and memory, but it cannot move component-time baselines.
+
+Stop if the normalized record's `model_id` or `gpu_type` does not match the
+requested `model_id` and `gpu_type`.
+
+The source record may have `success: false` when it came from a failed rolling
+baseline comparison. That is expected; only the reviewed reseed replicas become
+new `success: true` baseline records after explicit approval.
+
+Set `replica_count` to `3` by default. Set it to `5` only when the user
+explicitly asks to upload the same shifted source result 5 times for a full
+last-5 reset. Reject other counts unless the user gives a concrete reason.
+
+Check that `HF_API_KEY` is exported. The sync path may be public, but the
+upload path requires write access.
+
+### 1a. How to obtain `source_result` from CI
+
+The performance CI exports normalized source results for failed rolling
+baseline comparisons when `compare_baseline.py` ran. The preferred artifact
+comes from:
+
+```text
+perf_reports/results/normalized_perf_*.json
+```
+
+and is uploaded by Buildkite with the performance reports. The normal operator
+flow is:
+
+1. Open the failed Buildkite performance job.
+2. Download the `normalized_perf_*.json` artifact for the failed benchmark.
+3. Pass the local path or artifact URL as `source_result`.
+
+Do not scrape the Markdown performance summary to reconstruct the JSON. The
+normalized JSON artifact is the source of truth for reseed metrics and
+provenance. If only a raw `fastvideo/tests/performance/results/perf_*.json`
+artifact is present, normalize it with `normalize_performance_result()` before
+continuing. If no JSON artifact is present, the benchmark likely failed before
+writing results, so that run is not a valid source for baseline reseeding.
+
+### 2. Sync and back up existing HF records
+
+Use `fastvideo/tests/performance/hf_store.py` helpers directly. Do **not** use
+`compare_baseline.py` as a sync shortcut; on full main runs it can persist
+records, while this step must only fetch and back up existing history.
+
+The sync command pattern is:
+
+```bash
+export PERFORMANCE_TRACKING_ROOT="${PERFORMANCE_TRACKING_ROOT:-/tmp/perf-tracking}"
+export HF_REPO_ID="${HF_REPO_ID:-FastVideo/performance-tracking}"
+PYTHONPATH=fastvideo/tests/performance python -c 'from 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:
+
+```bash
+SHORT_COMMIT=$(git rev-parse --short=12 HEAD)
+TIMESTAMP=$(date -u +%Y%m%d_%H%M%S)
+MODEL_SAFE=$(python - <<'PY'
+from fastvideo.tests.performance.hf_store import sanitize
+print(sanitize("<model_id>"))
+PY
+)
+BACKUP_DIR="performance_reseed_backup/${TIMESTAMP}_${SHORT_COMMIT}_${MODEL_SAFE}"
+mkdir -p "$BACKUP_DIR"
+cp -R "${PERFORMANCE_TRACKING_ROOT}/${MODEL_SAFE}" "$BACKUP_DIR/" 2>/dev/null || true
+```
+
+Write provenance next to the backup:
+
+```bash
+cat > "$BACKUP_DIR/PROVENANCE.txt" <<EOF
+model_id: <model_id>
+gpu_type: <gpu_type>
+source_result: <source_result>
+replica_count: <3_or_5>
+head_commit: $(git rev-parse HEAD)
+timestamp_utc: $(date -u +%FT%TZ)
+reason: <intent_rationale>
+EOF
+```
+
+If the backup has no prior records, this is not a destructive reseed; it is a
+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:
+
+```python
+from fastvideo.tests.performance.hf_store import load_records_for_model
+
+records = load_records_for_model(
+    "/tmp/perf-tracking",
+    "<model_id>",
+    "<gpu_type>",
+    last_n=5,
+    successful_only=True,
+)
+```
+
+Print a small table showing the source result metrics, the replicated
+candidate median, and the old medians for:
+
+- `latency`
+- `throughput`
+- `memory`
+- `text_encoder_time_s`
+- `dit_time_s`
+- `vae_decode_time_s`
+
+Also print how many successful old records exist. Make clear:
+
+- 1 shifted record only seeds audit history and usually does not move the
+  median.
+- 3 replicated shifted records in a 5-record window move the median
+  immediately.
+- 5 replicated shifted records fully reset the rolling window to the source
+  result's runtime profile.
+- Replicated records are not independent measurements; they are an intentional
+  approved baseline reset and must be labeled that way.
+
+### 4. Confirm intent
+
+Require an explicit confirmation phrase before preparing the upload:
+
+> About to RE-SEED performance baseline for `<model_id>` on `<gpu_type>`.
+> This will upload `<N>` new `success=true` records to
+> `FastVideo/performance-tracking/<sanitize(model_id)>/`.
+>
+> Reason: `<intent_rationale>`
+> Source result: `<source_result>`
+> Replica count: `<replica_count>`
+> Note: these records replicate one reviewed measurement to force the rolling
+> median to the accepted runtime profile.
+> HEAD: `<git rev-parse --short=12 HEAD>`
+> Backup: `<BACKUP_DIR>`
+>
+> Reply `confirm performance reseed` to proceed, anything else to abort.
+
+Do not continue unless the user types exactly `confirm performance reseed`.
+
+### 5. Create the accepted seed records
+
+Create `replica_count` normalized records from the single source result. Use
+an explicit allowlist; do not copy the raw result JSON wholesale.
+
+Each record must include only these baseline fields plus the reseed provenance
+fields below:
+
+- `model_id`
+- `timestamp`
+- `commit_sha`
+- `gpu_type`
+- `latency`
+- `throughput`
+- `memory`
+- `text_encoder_time_s`
+- `dit_time_s`
+- `vae_decode_time_s`
+- `success: true`
+
+For normalized `normalized_perf_*.json` sources, these fields already exist.
+For older raw `perf_*.json` sources, map the raw fields exactly as
+`normalize_performance_result()` in `compare_baseline.py` does:
+
+| Normalized field | Raw source field |
+|------------------|------------------|
+| `model_id` | `benchmark_id` |
+| `gpu_type` | `device` |
+| `latency` | `avg_generation_time_s` |
+| `throughput` | `throughput_fps` |
+| `memory` | `max_peak_memory_mb` |
+| `text_encoder_time_s` | `text_encoder_time_s` |
+| `dit_time_s` | `dit_time_s` |
+| `vae_decode_time_s` | `vae_decode_time_s` |
+| `commit_sha` | `commit` |
+
+Do not upload raw-only fields such as `model_short_name`, `num_gpus`,
+`num_warmup_runs`, `num_measurement_runs`, `individual_times_s`,
+`individual_peak_memories_mb`, `thresholds`, or `pr_number`.
+
+Optional provenance fields are allowed and useful:
+
+- `baseline_reseed: true`
+- `baseline_reseed_reason`
+- `baseline_reseed_source_result`
+- `baseline_reseed_source_timestamp`
+- `baseline_reseed_replicated_source: true`
+- `baseline_reseed_batch_size`
+- `baseline_reseed_batch_index`
+- `baseline_reseed_operator`
+
+Use a fresh reseed timestamp for each replicated record, not the original
+source result timestamp. This is required because
+`load_records_for_model(..., last_n=5)` keeps the last records after loading
+the model directory; stale filenames/timestamps may not enter the last-5
+window and therefore may not move the median. Preserve the original source
+timestamp in `baseline_reseed_source_timestamp`.
+
+Use the existing filename convention from `_write_tracking_record()`:
+`<sanitize(timestamp)>_<sanitize(commit_sha)>.json` under the sanitized model
+directory, but include a deterministic suffix such as `_reseed_01`,
+`_reseed_02`, and `_reseed_03` before `.json` so the replicated files do not
+overwrite each other. For a 5-record full reset, continue through
+`_reseed_05`.
+
+If the source record already exists on HF with `success=false`, do not edit it
+in place unless the user explicitly asked for an audit-preserving correction.
+Prefer uploading new accepted seed records so failed history remains visible.
+
+### 6. Pause before upload
+
+Print:
+
+- Backup directory path.
+- HF paths that will receive the new records.
+- Old rolling medians.
+- Source metrics, replica count, and candidate median.
+- Rationale.
+
+Ask the user to reply exactly `upload`. Anything else aborts and leaves the
+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:
+
+```python
+from fastvideo.tests.performance.hf_store import upload_record
+
+upload_record("<local_record_path>", record, strict=True)
+```
+
+Run it once per prepared record. Each upload 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.
+
+### 8. Report outcome
+
+Report:
+
+- Uploaded HF paths.
+- Backup directory.
+- Old baseline window count and medians.
+- Source metrics, replica count, and candidate median.
+- Expected effect: 3 replicated shifted records move the 5-record median; 5
+  replicated shifted records fully reset the window to the accepted source
+  result.
+- Any separate threshold changes still needed in
+  `.buildkite/performance-benchmarks/tests/*.json`.
+
+Include the `intent_rationale` in the PR or follow-up comment so reviewers can
+distinguish an accepted baseline shift from a hidden regression.
+
+## Failure modes and handling
+
+- **`HF_API_KEY` unset.** Stop before upload. Do not create an untracked
+  process that appears to have reseeded but never reached HF.
+- **Source result does not match target.** Stop. The wrong benchmark or GPU
+  would poison a separate baseline.
+- **`replica_count` is 5 but the user did not explicitly ask for a full
+  reset.** Stop and use the default count of 3.
+- **The source result is noisy or suspicious.** Stop. Replicating one result
+  amplifies that measurement into the baseline, so it must be reviewed first.
+- **HF sync fails.** Stop for destructive reseeds. A stale or empty sync can
+  make the old baseline look missing.
+- **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.
+- **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.
+
+## References
+
+- `.agents/skills/reseed-ssim-references/SKILL.md` — safety pattern for
+  intentional baseline replacement.
+- `fastvideo/tests/performance/compare_baseline.py` — normalization, rolling
+  median comparison, and persistence rules.
+- `fastvideo/tests/performance/hf_store.py` — HF sync, record loading,
+  `sanitize()`, and `upload_record()`.
+- `fastvideo/tests/performance/test_inference_performance.py` — source result
+  JSON schema.
+- `.buildkite/performance-benchmarks/tests/*.json` — fixed absolute benchmark
+  thresholds, separate from rolling baseline comparisons.
+
+## Changelog
+
+| Date | Change |
+|------|--------|
+| 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 | Current 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. |

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

@@ -1,26 +1,41 @@
 ---
 name: seed-ssim-references
-description: Seed HF reference videos for a single newly-added SSIM test. Runs the test on Modal L40S, downloads the generated mp4s via `modal volume get`, pauses for the user to eyeball quality, then uploads only that test's files to `FastVideo/ssim-reference-videos`. Use when a new `fastvideo/tests/ssim/test_*_similarity.py` has just been added and has no references on HF yet.
+description: Seed HF reference artefacts for a single newly-added SSIM test (pixel `.mp4` for `run_text_to_video_similarity_test`-style tests, or latent `.pt` for `run_text_to_latent_similarity_test`-style tests). Runs the test on Modal L40S, downloads the generated artefacts via `modal volume get`, pauses for the user to verify (visual eyeball for mp4, numerics dump for pt), then uploads only that test's files to `FastVideo/ssim-reference-videos`. Use when a new `fastvideo/tests/ssim/test_*_similarity.py` has just been added and has no references on HF yet.
 ---
 
-# Seed SSIM Reference Videos
+# Seed SSIM Reference Artefacts (mp4 or pt)
 
 ## Purpose
 
 A brand-new SSIM test in `fastvideo/tests/ssim/` fails forever until its
-reference videos exist on the HF dataset (`FastVideo/ssim-reference-videos`).
+reference artefacts exist on the HF dataset
+(`FastVideo/ssim-reference-videos`). The dataset hosts two kinds of artefacts
+side-by-side per `(model_id, backend, prompt)`:
+
+- **`.mp4`** — pixel ground-truth for tests that call
+  `run_text_to_video_similarity_test` / `run_image_to_video_similarity_test`
+  in `inference_similarity_utils.py`. Compared via SSIM.
+- **`.pt`** — pre-VAE latent bundle (fp16 full latent + fp32 slice +
+  metadata + `slice_spec` + `format_version`) for tests that call
+  `run_text_to_latent_similarity_test` in `latent_similarity_utils.py`.
+  Compared via cosine distance on the slice and the full tensor.
+
 This skill:
 
-1. Runs the test on Modal's L40S pool to generate the videos.
-2. Downloads them to the local repo via `modal volume get`.
-3. Pauses so the user can eyeball the mp4s and confirm quality.
-4. Uploads only the new test's files to HF, with a guard that refuses to
+1. Detects which artefact type the test produces (pixel vs latent).
+2. Runs the test on Modal's L40S pool to generate the artefacts.
+3. Downloads them to the local repo via `modal volume get`.
+4. Pauses so the user can verify quality:
+   - **mp4**: visual eyeball in a video player.
+   - **pt**: numerics dump (shape, slice stats, NaN/Inf check, metadata).
+5. Uploads only the new test's files to HF, with a guard that refuses to
    overwrite anything already present.
 
 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 mp4 without crashing. The skill does not re-test locally; it
-goes straight to Modal L40S (which is what CI uses).
+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).
 
 ## When to use
 
@@ -69,7 +84,7 @@ Fail fast if the token env var is missing.
 
 ## Steps
 
-### 1. Ask for the test file
+### 1. Ask for the test file, then detect artefact type
 
 If the user didn't name one, ask: *"Which SSIM test file do you want to seed
 references for? (e.g. `fastvideo/tests/ssim/test_ltx2_similarity.py`)"*.
@@ -80,6 +95,22 @@ Validate:
 - File defines a `*_MODEL_TO_PARAMS` dict — grep it to extract the set of
   model ids. Those ids drive step 5.
 
+Detect artefact type by inspecting the file's imports / helper call:
+
+- **latent** (`.pt`) — file imports `run_text_to_latent_similarity_test`
+  from `fastvideo.tests.ssim.latent_similarity_utils` (or any other helper
+  that ends with `_latent_similarity_test`).
+- **pixel** (`.mp4`) — file imports
+  `run_text_to_video_similarity_test` / `run_image_to_video_similarity_test`
+  from `fastvideo.tests.ssim.inference_similarity_utils`, OR uses the
+  legacy custom-inline helper pattern (see `test_gamecraft`,
+  `test_longcat`, etc.). Default to pixel when both heuristics fail.
+
+Record `ARTEFACT_TYPE ∈ {pixel, latent}` for use in step 4. Steps 2, 3, 5,
+and 6 are artefact-type-agnostic — `_iter_reference_files`,
+`copy_generated_to_reference`, and `upload_reference_videos` already walk
+both `.mp4` and `.pt` (see `reference_videos_cli.py`).
+
 If either check fails, stop and tell the user what's wrong.
 
 ### 2. Run the test on Modal L40S
@@ -166,17 +197,59 @@ get` preserves that trailing `generated_videos/` segment.
 
 ### 4. PAUSE — user reviews quality
 
-Print the list of downloaded mp4s and their paths, then stop. Tell the user:
+Type-aware verification.
+
+**For `ARTEFACT_TYPE = pixel`** — list the downloaded mp4s and ask the user to
+open them in a video player:
 
 > "Generated videos downloaded to `./generated_videos_modal/default/generated_videos/L40S_reference_videos/`. Please open them and confirm the quality looks correct. Reply **`upload`** to continue, or anything else to abort."
 
+**For `ARTEFACT_TYPE = latent`** — `.pt` files are not human-watchable. Print
+a numerics dump for each `.pt` so the user can sanity-check shape, distribution,
+and metadata:
+
+```python
+import torch
+from pathlib import Path
+ROOT = Path("./generated_videos_modal/default/generated_videos/L40S_reference_videos")
+for p in sorted(ROOT.rglob("*.pt")):
+    d = torch.load(p, map_location="cpu", weights_only=False)
+    s = d["expected_slice"]
+    L = d["latent"].float()
+    print(f"=== {p.relative_to(ROOT)} ===")
+    print(f"  format_version: {d['format_version']}")
+    print(f"  shape:          {d['shape']}")
+    print(f"  dtype_original: {d['dtype_original']}")
+    print(f"  slice_spec:     {d['slice_spec']}")
+    print(f"  slice  shape={tuple(s.shape)}  mean={s.mean():+.4f}  std={s.std():.4f}  min={s.min():+.4f}  max={s.max():+.4f}")
+    print(f"  latent shape={tuple(L.shape)}  mean={L.mean():+.4f}  std={L.std():.4f}  min={L.min():+.4f}  max={L.max():+.4f}")
+    print(f"  finite: latent NaN={torch.isnan(L).any().item()} Inf={torch.isinf(L).any().item()}; "
+          f"slice NaN={torch.isnan(s).any().item()} Inf={torch.isinf(s).any().item()}")
+    print(f"  metadata: {d['metadata']}\n")
+```
+
+Sanity criteria:
+- `format_version == 1` (matches `LATENT_REFERENCE_FORMAT_VERSION`).
+- `shape` matches what the model produces (e.g. LTX-2 distilled =
+  `[1, 128, T_lat, H_lat, W_lat]`; Stable Audio Open 1.0 = `[1, 64, 1024]`).
+- `slice_spec.kind` matches a registered kind (`corner_3x3_first_frame`
+  for video, `audio_first_8_timesteps` for audio).
+- No `NaN`/`Inf`. `mean ≈ 0`, `std ≈ 1` (denoised latents stay close to
+  the initial Gaussian distribution; very wide deviations suggest
+  numerical drift).
+- `metadata.prompt` matches the test's prompt.
+
+Then ask:
+
+> "Numerics look right? Reply **`upload`** to continue, or anything else to abort."
+
 Do not proceed until the user explicitly says `upload`. If they abort, leave
 everything on disk so they can inspect further — no cleanup.
 
 ### 5. Copy into the local reference layout
 
-Scoped copy — only the new test's mp4s. Loop over each `<model_id>` extracted
-in step 1:
+Scoped copy — only the new test's artefacts. Single command works for both
+artefact types because `_iter_reference_files` walks `.mp4` and `.pt`:
 
 ```bash
 python fastvideo/tests/ssim/reference_videos_cli.py copy-local \
@@ -186,12 +259,13 @@ python fastvideo/tests/ssim/reference_videos_cli.py copy-local \
 ```
 
 (The `--generated-dir` points at the device-folder root inside the
-downloaded tree; `copy-local` walks all `<model>/<backend>/*.mp4`
+downloaded tree; `copy-local` walks all `<model>/<backend>/*.{mp4,pt}`
 underneath it. Since the Modal run was scoped to a single test file via
 `--test-files`, only that test's model(s) are present — so the copy is
 implicitly per-test.)
 
-Result: `fastvideo/tests/ssim/reference_videos/default/L40S_reference_videos/<model_id>/<backend>/<prompt>.mp4`.
+Result for pixel: `fastvideo/tests/ssim/reference_videos/default/L40S_reference_videos/<model_id>/<backend>/<prompt>.mp4`.
+Result for latent: same path with `.pt` extension.
 
 ### 6. Upload to HF — scoped per model_id, with overwrite guard
 
@@ -224,33 +298,54 @@ it will auto-download the refs they just uploaded.
 ## Failure modes and how to handle them
 
 - **`HF_API_KEY` unset.** Stop before step 2. The Modal run needs it (passed
-  via `--hf-api-key`), and step 6 needs it for upload.
-- **Modal run fails before generation.** No mp4s on the volume — nothing to
-  download. Fix the test locally (`pytest fastvideo/tests/ssim/<test_file>`)
+  via `--hf-api-key`), and step 6 needs it for upload. If the user
+  ran `hf auth login` instead of exporting an env var, read the cached
+  token via `huggingface_hub.get_token()` and forward it to Modal as
+  `--hf-api-key="$CACHED_TOKEN"`.
+- **Modal run fails before generation.** No artefacts on the volume — nothing
+  to download. Fix the test locally (`pytest fastvideo/tests/ssim/<test_file>`)
   and retry from step 2.
 - **`./generated_videos_modal/default/L40S_reference_videos/` missing after
-  `modal volume get`.** The run didn't produce videos (most likely the test
-  crashed before writing, or `REQUIRED_GPUS` exceeded the partition capacity
-  — see Modal logs).
+  `modal volume get`.** The run didn't produce artefacts (most likely the
+  test crashed before writing, or `REQUIRED_GPUS` exceeded the partition
+  capacity — see Modal logs).
+- **Latent test crashed with FSDP / inference_mode error
+  (`RuntimeError: Inference tensors do not track version counter`).** The
+  test must pass `init_kwargs_override={"use_fsdp_inference": False}` when
+  `sp_size == 1` — see `test_stable_audio_similarity.py` for the pattern.
+  Fix in the test, push, retry.
 - **Upload guard fires (files already exist).** The test name / model id
   collides with something already on HF. Verify the user actually wants to
   replace existing refs; if so, re-run the upload with `--force`. If not,
   rename the model id in `*_MODEL_TO_PARAMS` and re-seed.
-- **Quality looks wrong in step 4.** Abort. The mp4s stay on disk for
+- **Quality looks wrong in step 4.** Abort. The artefacts stay on disk for
   inspection. The fix is usually in the test's params (resolution, steps,
   seed) — edit the test, then re-run the skill.
+  - For latent: also check `slice_spec.kind` matches the latent rank
+    (`corner_3x3_first_frame` requires 5-D, `audio_first_8_timesteps`
+    requires 3-D); a rank/kind mismatch raises in `_extract_expected_slice`.
 
 ## Design notes (for future skill maintainers)
 
 - The skill deliberately runs on Modal, **not** locally, because the CI
   runner is L40S. Seeding from a different GPU SKU produces refs that CI's
-  L40S runs can't match (SSIM drifts across SKUs).
+  L40S runs can't match (pixel SSIM drifts across SKUs; latent cosine has
+  tighter cross-SKU bf16 drift but the configured tolerances assume
+  same-SKU seed → same-SKU verify).
 - The skill is default-tier only. `full_quality` refs are seeded by a
   separate, deliberate operation — they double runtime and aren't what CI
   gates on.
 - The overwrite guard in `reference_videos_cli.py upload` is default-on
   specifically because this skill exists. Re-seeding is a distinct operation
   that requires explicit `--force`.
+- Both artefact types share the same Modal flow: the orchestrator sets
+  `--skip-reference-download` + `--no-fail-fast`, runs pytest, the test's
+  helper writes the artefact (`.mp4` via `imageio` for pixel,
+  `save_latent_reference` → `torch.save` for latent) BEFORE the
+  missing-reference assertion raises. `_sync_generated_videos_to_volume` in
+  `ssim_test.py` does a `shutil.copytree` of the whole `generated_videos/`
+  tree, picking up `.mp4`, `.pt`, and the `*_ssim.json` / `*_latent.json`
+  metric files alongside.
 
 ## References
 
@@ -259,10 +354,17 @@ it will auto-download the refs they just uploaded.
   `--skip-reference-download`, `--no-fail-fast`.
 - `fastvideo/tests/ssim/reference_videos_cli.py` — `copy-local`, `upload`
   (with `--model-id`, `--force`), `download`, `ensure` subcommands.
+  Extension allowlist is `REFERENCE_EXTENSIONS = VIDEO_EXTENSIONS +
+  LATENT_EXTENSIONS` (`.pt`).
 - `fastvideo/tests/ssim/README.md` — reference layout, HF repo conventions.
-- `fastvideo/tests/ssim/inference_similarity_utils.py` —
-  `run_text_to_video_similarity_test` + `_build_init_kwargs`: what each test
-  config passes to `VideoGenerator.from_pretrained`.
+- `fastvideo/tests/ssim/inference_similarity_utils.py` — pixel helpers
+  (`run_text_to_video_similarity_test`,
+  `run_image_to_video_similarity_test`, `build_init_kwargs`).
+- `fastvideo/tests/ssim/latent_similarity_utils.py` — latent helper
+  (`run_text_to_latent_similarity_test`), slice spec dispatch
+  (`_extract_expected_slice`), reference schema
+  (`save_latent_reference` / `load_latent_reference`),
+  `LATENT_REFERENCE_FORMAT_VERSION`.
 
 ## Changelog
 
@@ -271,3 +373,4 @@ it will auto-download the refs they just uploaded.
 | 2026-04-17 | Initial version (Modal sync-to-volume flow). |
 | 2026-04-21 | Rewrite: single-test scope, explicit user-review pause, per-`model_id` upload, HF overwrite guard. Dropped `scripts/seed_ssim.sh`. |
 | 2026-04-21 | Post-first-run fixes: `modal volume get` needs `--force` when parent exists; download tree has an extra `generated_videos/` level so `--generated-dir` must reflect it. |
+| 2026-05-01 | Latent (`*.pt`) artefact support: artefact-type detection in step 1, type-aware verification (visual eyeball for mp4, numerics dump for pt) in step 4, FSDP+inference_mode failure-mode added, design notes for the unified Modal flow. Triggered by PR #1253 (LTX-2 latent migration + Stable Audio latent test). |

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

@@ -36,11 +36,17 @@
   "thresholds": {
     "L40S": {
       "max_generation_time_s": 34.0,
-      "max_peak_memory_mb": 11000.0
+      "max_peak_memory_mb": 11000.0,
+      "max_text_encoder_time_s": 5.0,
+      "max_dit_time_s": 10.0,
+      "max_vae_decode_time_s": 10.0
     },
     "default": {
       "max_generation_time_s": 120.0,
-      "max_peak_memory_mb": 30000.0
+      "max_peak_memory_mb": 30000.0,
+      "max_text_encoder_time_s": 5.0,
+      "max_dit_time_s": 10.0,
+      "max_vae_decode_time_s": 10.0
     }
   }
 }

.buildkite/scripts/pr_test.sh

@@ -121,6 +121,19 @@ upload_performance_artifacts() {
         fi
     }
 
+    _upload_normalized_perf_results() {
+        local found=0
+        while IFS= read -r -d '' target; do
+            found=1
+            log "Found normalized performance result: $target. Uploading to Buildkite..."
+            buildkite-agent artifact upload "$target"
+        done < <(find "$LOCAL_DIR" -path "*/results/normalized_perf_*.json" -print0)
+
+        if [ "$found" -eq 0 ]; then
+            log "No normalized performance result artifacts found. This is expected when the rolling performance comparison did not run."
+        fi
+    }
+
     _cleanup_modal_volume() {
         log "Cleaning up perf_reports/ from Modal Volume..."
         if modal volume rm hf-model-weights "perf_reports/" --recursive; then
@@ -139,6 +152,7 @@ upload_performance_artifacts() {
     _download_reports || { _cleanup_local; return 1; }
     _upload_dashboard
     _upload_perf_summary
+    _upload_normalized_perf_results
     _cleanup_modal_volume
     _cleanup_local
 }

fastvideo/entrypoints/cli/main.py

@@ -3,6 +3,8 @@
 from fastvideo.entrypoints.cli.cli_types import CLISubcommand
 from fastvideo.entrypoints.cli.generate import cmd_init as generate_cmd_init
 from fastvideo.utils import FlexibleArgumentParser
+from fastvideo.entrypoints.cli.router_serve import (
+    cmd_init as router_serve_cmd_init, )
 from fastvideo.entrypoints.cli.serve import cmd_init as serve_cmd_init
 from fastvideo.entrypoints.cli.bench import cmd_init as bench_cmd_init
 
@@ -12,6 +14,7 @@ def cmd_init() -> list[CLISubcommand]:
     commands = []
     commands.extend(generate_cmd_init())
     commands.extend(serve_cmd_init())
+    commands.extend(router_serve_cmd_init())
     commands.extend(bench_cmd_init())
     return commands
 

fastvideo/entrypoints/cli/router_serve.py

@@ -0,0 +1,115 @@
+# SPDX-License-Identifier: Apache-2.0
+"""``fastvideo router-serve`` CLI subcommand.
+
+Launches the streaming router from a YAML config. Separate from
+``fastvideo serve`` because the router is an orthogonal process: it
+fronts one or more running servers rather than hosting a generator
+itself.
+"""
+from __future__ import annotations
+
+import argparse
+import os
+from typing import cast
+
+from fastvideo.api.parser import load_raw_config
+from fastvideo.entrypoints.cli.cli_types import CLISubcommand
+from fastvideo.entrypoints.streaming.router.config import (
+    ReplicaEndpoint,
+    RouterConfig,
+)
+from fastvideo.logger import init_logger
+from fastvideo.utils import FlexibleArgumentParser
+
+logger = init_logger(__name__)
+
+
+class RouterServeSubcommand(CLISubcommand):
+    """Start the multi-replica WebSocket router."""
+
+    def __init__(self) -> None:
+        self.name = "router-serve"
+        super().__init__()
+
+    def cmd(self, args: argparse.Namespace) -> None:
+        config = _load_router_config(args.config)
+        logger.info(
+            "router listening on %s:%d (%d replicas, %d primary)",
+            config.host,
+            config.port,
+            len(config.replicas),
+            sum(1 for r in config.replicas if r.primary),
+        )
+        from fastvideo.entrypoints.streaming.router.main import run_router
+
+        run_router(config)
+
+    def validate(self, args: argparse.Namespace) -> None:
+        if not args.config:
+            raise ValueError("fastvideo router-serve requires --config PATH")
+        if not os.path.exists(args.config):
+            raise ValueError(f"Router config file not found: {args.config}")
+
+    def subparser_init(
+        self,
+        subparsers: argparse._SubParsersAction,
+    ) -> FlexibleArgumentParser:
+        parser = subparsers.add_parser(
+            "router-serve",
+            help="Start the streaming router (multi-replica load balancer)",
+            usage="fastvideo router-serve --config ROUTER_CONFIG",
+        )
+        parser.add_argument(
+            "--config",
+            type=str,
+            default="",
+            required=False,
+            help="Path to a YAML/JSON router config. Required.",
+        )
+        return cast(FlexibleArgumentParser, parser)
+
+
+def _load_router_config(path: str) -> RouterConfig:
+    raw = load_raw_config(path)
+    router_raw = raw.get("router") if isinstance(raw, dict) else None
+    if not isinstance(router_raw, dict):
+        raise ValueError(f"Router config {path!r} must have a top-level `router:` block")
+
+    replicas_raw = router_raw.get("replicas", [])
+    if not isinstance(replicas_raw, list):
+        raise ValueError(f"router.replicas must be a list, got {type(replicas_raw).__name__}")
+    replicas = []
+    for i, r in enumerate(replicas_raw):
+        if not isinstance(r, dict):
+            raise ValueError(f"router.replicas[{i}] must be a mapping, got {type(r).__name__}")
+        url = r.get("url")
+        if not url:
+            raise ValueError(f"router.replicas[{i}] is missing required key 'url'")
+        replicas.append(
+            ReplicaEndpoint(
+                url=url,
+                name=r.get("name"),
+                primary=bool(r.get("primary", False)),
+                weight=float(r.get("weight", 1.0)),
+            ))
+    if not replicas:
+        raise ValueError("Router config must list at least one replica under `router.replicas`")
+
+    health_check = router_raw.get("health_check") or {}
+    return RouterConfig(
+        host=str(router_raw.get("host", "0.0.0.0")),
+        port=int(router_raw.get("port", 9000)),
+        replicas=replicas,
+        health_check_path=str(health_check.get("path", "/health")),
+        health_check_interval_seconds=float(health_check.get("interval_seconds", 5.0)),
+        health_check_timeout_seconds=float(health_check.get("timeout_seconds", 2.0)),
+        failure_threshold=int(health_check.get("failure_threshold", 3)),
+        recovery_threshold=int(health_check.get("recovery_threshold", 2)),
+    )
+
+
+def cmd_init() -> list[CLISubcommand]:
+    return [RouterServeSubcommand()]
+
+
+__all__ = ["RouterServeSubcommand", "cmd_init"]

fastvideo/entrypoints/streaming/__init__.py

@@ -11,6 +11,28 @@ from fastvideo.entrypoints.streaming.session_store import (
     InMemorySessionStore,
     SessionStore,
 )
+from fastvideo.entrypoints.streaming.gpu_pool import (
+    GpuPool,
+    InProcessGpuPool,
+    PoolAcquireTimeout,
+    SubprocessGpuPool,
+)
+from fastvideo.entrypoints.streaming.mock_server import (
+    MockGenerator,
+    build_mock_app,
+)
+from fastvideo.entrypoints.streaming.prompt import (
+    LLMProvider,
+    PromptEnhancer,
+)
+from fastvideo.entrypoints.streaming.prompt.safety import (
+    PromptSafetyFilter,
+    SafetyDecision,
+)
+from fastvideo.entrypoints.streaming.session_logger import (
+    SessionLogEvent,
+    SessionLogger,
+)
 from fastvideo.entrypoints.streaming.stream import (
     FragmentedMP4Chunk,
     FragmentedMP4Encoder,
@@ -20,12 +42,24 @@ __all__ = [
     "BlobStore",
     "FragmentedMP4Chunk",
     "FragmentedMP4Encoder",
+    "GpuPool",
     "InMemoryBlobStore",
     "InMemorySessionStore",
+    "InProcessGpuPool",
+    "LLMProvider",
+    "MockGenerator",
+    "PoolAcquireTimeout",
+    "PromptEnhancer",
+    "PromptSafetyFilter",
+    "SafetyDecision",
+    "SessionLogEvent",
+    "SessionLogger",
+    "build_mock_app",
     "Session",
     "SessionManager",
     "SessionState",
     "SessionStore",
+    "SubprocessGpuPool",
     "build_app",
     "run_server",
 ]

fastvideo/entrypoints/streaming/gpu_pool.py

@@ -0,0 +1,542 @@
+# SPDX-License-Identifier: Apache-2.0
+"""GPU pool manager for the streaming server.
+
+Replaces the single-generator path in PR 7.5 with a typed pool
+abstraction. Three implementations ship here:
+
+* :class:`InProcessGpuPool` — one in-process ``VideoGenerator``; used
+  by tests and single-GPU dev deployments.
+* :class:`SubprocessGpuPool` — one ``multiprocessing.Process`` per
+  GPU, each running :func:`worker_main` against a ``GeneratorConfig``.
+  Jobs are dispatched via ``multiprocessing.Queue``.
+* :class:`GpuPool` (abstract) — the interface both use.
+
+Session-to-GPU binding lives in the pool so continuation state stays
+on the GPU that generated the previous segment (matching the internal
+``gpu_pool.py``'s per-GPU cache behavior). Cross-GPU handoff is
+supported via :class:`SessionStore` snapshot + hydrate, which
+serializes the state before the migration and rehydrates it on the
+new worker.
+
+Typed config: workers start from a :class:`GeneratorConfig` (no flat
+LTX-2 kwargs), satisfying the PR 6 + PR 7 contracts that the public
+surface doesn't reintroduce the legacy kwarg bag.
+"""
+from __future__ import annotations
+
+import asyncio
+import multiprocessing as mp
+import queue
+import threading
+import time
+import uuid
+from abc import ABC, abstractmethod
+from concurrent.futures import Future
+from dataclasses import dataclass, field
+from typing import Any, Protocol
+
+from fastvideo.api.schema import (
+    GeneratorConfig,
+    GenerationRequest,
+    GpuPoolConfig,
+    WarmupConfig,
+)
+from fastvideo.entrypoints.streaming.session_store import (
+    InMemorySessionStore,
+    SessionStore,
+)
+from fastvideo.entrypoints.streaming.worker import worker_main
+from fastvideo.logger import init_logger
+
+logger = init_logger(__name__)
+
+# ---------------------------------------------------------------------------
+# Public interface
+# ---------------------------------------------------------------------------
+
+
+class _GeneratorLike(Protocol):
+    """Subset the pool calls on a worker-side generator."""
+
+    def generate(self, request: GenerationRequest) -> Any:
+        ...
+
+
+@dataclass
+class PoolAssignment:
+    """The worker a session is currently bound to."""
+
+    gpu_id: int
+    worker_id: str
+    pinned_at: float = field(default_factory=time.monotonic)
+
+
+class GpuPool(ABC):
+    """Abstract GPU pool.
+
+    ``acquire`` binds a session to a worker and holds that binding
+    across segments so continuation state can stay hot. ``run`` submits
+    a single ``GenerationRequest`` for a bound session.
+
+    Acquire / release are independent of run — a session can run many
+    segments on one acquired worker, and must release on disconnect.
+    """
+
+    @abstractmethod
+    async def acquire(
+        self,
+        session_id: str,
+        *,
+        timeout: float | None = None,
+    ) -> PoolAssignment:
+        ...
+
+    @abstractmethod
+    async def run(
+        self,
+        session_id: str,
+        request: GenerationRequest,
+    ) -> Any:
+        ...
+
+    @abstractmethod
+    async def release(self, session_id: str) -> None:
+        ...
+
+    @abstractmethod
+    async def shutdown(self) -> None:
+        ...
+
+    @abstractmethod
+    def health(self) -> PoolHealth:
+        ...
+
+
+@dataclass
+class PoolHealth:
+    total_workers: int
+    available_workers: int
+    active_sessions: int
+    queued_sessions: int = 0
+
+
+class PoolAcquireTimeout(RuntimeError):
+    """Raised when ``acquire`` times out waiting for a free worker."""
+
+
+# ---------------------------------------------------------------------------
+# In-process implementation (single-worker, test / dev)
+# ---------------------------------------------------------------------------
+
+
+class InProcessGpuPool(GpuPool):
+    """Single-process pool backed by one :class:`_GeneratorLike`.
+
+    This is what PR 7.5's server uses by default; PR 7.6 adds the real
+    ``SubprocessGpuPool`` alternative but keeps this one for tests and
+    small deployments.
+    """
+
+    def __init__(
+        self,
+        generator: _GeneratorLike,
+        *,
+        gpu_id: int = 0,
+        session_store: SessionStore | None = None,
+    ) -> None:
+        self._generator = generator
+        self._gpu_id = gpu_id
+        self._worker_id = f"inproc-{uuid.uuid4().hex[:6]}"
+        self._session_store = session_store or InMemorySessionStore()
+        self._active: dict[str, PoolAssignment] = {}
+        self._lock = asyncio.Lock()
+        self._gen_lock = asyncio.Lock()
+
+    async def acquire(
+        self,
+        session_id: str,
+        *,
+        timeout: float | None = None,
+    ) -> PoolAssignment:
+        async with self._lock:
+            existing = self._active.get(session_id)
+            if existing is not None:
+                return existing
+            assignment = PoolAssignment(gpu_id=self._gpu_id, worker_id=self._worker_id)
+            self._active[session_id] = assignment
+            return assignment
+
+    async def run(
+        self,
+        session_id: str,
+        request: GenerationRequest,
+    ) -> Any:
+        if session_id not in self._active:
+            raise RuntimeError(f"session {session_id!r} is not acquired on this pool")
+        # Serialize generator access so one GPU runs one request at a
+        # time, matching the internal gpu_pool's per-GPU lock.
+        async with self._gen_lock:
+            loop = asyncio.get_running_loop()
+            return await loop.run_in_executor(None, self._generator.generate, request)
+
+    async def release(self, session_id: str) -> None:
+        async with self._lock:
+            self._active.pop(session_id, None)
+
+    async def shutdown(self) -> None:
+        self._active.clear()
+
+    def health(self) -> PoolHealth:
+        return PoolHealth(
+            total_workers=1,
+            available_workers=1 if not self._active else 0,
+            active_sessions=len(self._active),
+        )
+
+
+# ---------------------------------------------------------------------------
+# Subprocess implementation (multi-worker, real deployment)
+# ---------------------------------------------------------------------------
+
+
+@dataclass
+class _WorkerHandle:
+    process: Any  # mp.Process or compatible handle with is_alive / join / kill
+    job_queue: mp.Queue
+    result_queue: mp.Queue
+    gpu_id: int
+    worker_id: str
+    ready: threading.Event
+    # ``ready`` flips on either successful boot or boot failure so the
+    # parent stops waiting; ``boot_ok`` is set only on a real ready
+    # acknowledgement and is what gates pool admission.
+    boot_ok: threading.Event
+    shutdown_event: Any  # mp.Event is a factory, not a type — Any keeps mypy sane
+
+
+@dataclass
+class _PendingJob:
+    job_id: str
+    future: Future
+    session_id: str
+    worker_id: str
+
+
+class SubprocessGpuPool(GpuPool):
+    """One ``multiprocessing.Process`` per GPU.
+
+    Each worker boots :class:`fastvideo.VideoGenerator` from a typed
+    :class:`GeneratorConfig` inside the child process (post-
+    ``CUDA_VISIBLE_DEVICES`` setup) and consumes jobs from an mp Queue.
+
+    This is the production shape: the parent process stays CPU-only, and
+    GPU state never crosses process boundaries. Continuation state is
+    serialized through :class:`SessionStore` for cross-GPU handoff.
+
+    PR 7.6 ships this as an opt-in; PR 7.5's in-process pool remains the
+    default until nightly runs validate the subprocess path.
+    """
+
+    def __init__(
+        self,
+        generator_config: GeneratorConfig,
+        *,
+        pool_config: GpuPoolConfig,
+        warmup_config: WarmupConfig | None = None,
+        session_store: SessionStore | None = None,
+        worker_factory: WorkerFactory | None = None,
+    ) -> None:
+        self._generator_config = generator_config
+        self._pool_config = pool_config
+        self._warmup_config = warmup_config or WarmupConfig()
+        self._session_store = session_store or InMemorySessionStore()
+        self._worker_factory = worker_factory or _default_worker_factory
+        self._workers: list[_WorkerHandle] = []
+        self._available: asyncio.Queue[int] = asyncio.Queue()
+        self._assignments: dict[str, PoolAssignment] = {}
+        self._worker_by_id: dict[str, _WorkerHandle] = {}
+        self._pending: dict[str, _PendingJob] = {}
+        self._lock = asyncio.Lock()
+        self._result_reader_tasks: list[asyncio.Task] = []
+
+    async def start(self) -> None:
+        """Spawn worker processes and wait for each to report ready."""
+        num_workers = self._pool_config.num_workers or 1
+        for gpu_id in range(num_workers):
+            handle = self._worker_factory(
+                gpu_id=gpu_id,
+                generator_config=self._generator_config,
+                warmup_config=self._warmup_config,
+            )
+            self._workers.append(handle)
+            self._worker_by_id[handle.worker_id] = handle
+
+        # Wait for each worker's ready event in a thread to avoid
+        # blocking the event loop.
+        loop = asyncio.get_running_loop()
+        await asyncio.gather(*[
+            loop.run_in_executor(None, handle.ready.wait, self._warmup_config.timeout_seconds)
+            for handle in self._workers
+        ])
+
+        # Start background result readers — one task per worker
+        # drains its result queue and resolves futures in _pending.
+        for handle in self._workers:
+            task = asyncio.create_task(self._drain_results(handle))
+            self._result_reader_tasks.append(task)
+
+        # Only admit workers that successfully booted. Anything that
+        # failed boot (timeout, crash, error sentinel) stays out of the
+        # available queue so we never assign a session to it.
+        for idx, handle in enumerate(self._workers):
+            if handle.boot_ok.is_set():
+                await self._available.put(idx)
+            else:
+                logger.error(
+                    "pool: worker %s failed to boot; skipping",
+                    handle.worker_id,
+                )
+
+    async def acquire(
+        self,
+        session_id: str,
+        *,
+        timeout: float | None = None,
+    ) -> PoolAssignment:
+        async with self._lock:
+            existing = self._assignments.get(session_id)
+            if existing is not None:
+                return existing
+        try:
+            idx = await asyncio.wait_for(self._available.get(), timeout=timeout)
+        except asyncio.TimeoutError as exc:
+            raise PoolAcquireTimeout(f"no worker available after {timeout}s") from exc
+        handle = self._workers[idx]
+        assignment = PoolAssignment(gpu_id=handle.gpu_id, worker_id=handle.worker_id)
+        async with self._lock:
+            self._assignments[session_id] = assignment
+        return assignment
+
+    async def run(
+        self,
+        session_id: str,
+        request: GenerationRequest,
+    ) -> Any:
+        assignment = self._assignments.get(session_id)
+        if assignment is None:
+            raise RuntimeError(f"session {session_id!r} not acquired on this pool")
+        handle = self._worker_by_id[assignment.worker_id]
+        job_id = uuid.uuid4().hex
+        future: Future = Future()
+        self._pending[job_id] = _PendingJob(
+            job_id=job_id,
+            future=future,
+            session_id=session_id,
+            worker_id=handle.worker_id,
+        )
+        # mp.Queue.put can block if the underlying pipe buffer is full;
+        # offload to a thread so the event loop keeps serving other
+        # sessions. If the put itself fails, drop the pending entry so
+        # _drain_results doesn't dangle a future forever.
+        loop = asyncio.get_running_loop()
+        try:
+            await loop.run_in_executor(
+                None,
+                handle.job_queue.put,
+                {
+                    "job_id": job_id,
+                    "request": request
+                },
+            )
+        except Exception:
+            self._pending.pop(job_id, None)
+            raise
+        return await asyncio.wrap_future(future)
+
+    async def release(self, session_id: str) -> None:
+        async with self._lock:
+            assignment = self._assignments.pop(session_id, None)
+        if assignment is None:
+            return
+        idx = next((i for i, h in enumerate(self._workers) if h.worker_id == assignment.worker_id), None)
+        if idx is None:
+            return
+        # Don't return a dead worker to the pool; otherwise the next
+        # acquire will hand a session to a process that can't run jobs.
+        if not self._workers[idx].process.is_alive():
+            logger.warning(
+                "pool: worker %s died; not returning to available queue",
+                self._workers[idx].worker_id,
+            )
+            return
+        await self._available.put(idx)
+
+    async def shutdown(self) -> None:
+        loop = asyncio.get_running_loop()
+
+        # Signal all workers in parallel; .put may block on a full pipe,
+        # so off-load it the same way run() does.
+        async def _signal(handle: _WorkerHandle) -> None:
+            try:
+                handle.shutdown_event.set()
+                await loop.run_in_executor(None, handle.job_queue.put, None)
+            except Exception:  # pragma: no cover - best-effort cleanup
+                pass
+
+        await asyncio.gather(*(_signal(h) for h in self._workers))
+        # Join in parallel so total shutdown is bounded by the slowest
+        # worker, not the sum of all timeouts.
+        await asyncio.gather(*(loop.run_in_executor(None, handle.process.join, 5.0) for handle in self._workers))
+        for handle in self._workers:
+            if handle.process.is_alive():
+                handle.process.kill()
+        for task in self._result_reader_tasks:
+            task.cancel()
+        self._result_reader_tasks.clear()
+        self._workers.clear()
+        self._worker_by_id.clear()
+
+    def health(self) -> PoolHealth:
+        return PoolHealth(
+            total_workers=len(self._workers),
+            available_workers=self._available.qsize(),
+            active_sessions=len(self._assignments),
+        )
+
+    async def _drain_results(self, handle: _WorkerHandle) -> None:
+        loop = asyncio.get_running_loop()
+        try:
+            while not handle.shutdown_event.is_set():
+                try:
+                    msg = await loop.run_in_executor(None, _safe_queue_get, handle.result_queue, 0.5)
+                except Exception:
+                    logger.exception("pool: worker %s result reader failed", handle.worker_id)
+                    return
+                if msg is None:
+                    continue
+                job_id = msg.get("job_id")
+                if job_id is None:
+                    continue
+                pending = self._pending.pop(job_id, None)
+                if pending is None:
+                    continue
+                if msg.get("kind") == "error":
+                    pending.future.set_exception(RuntimeError(msg["error"]))
+                else:
+                    pending.future.set_result(msg.get("result"))
+        finally:
+            # If we exit for any reason — shutdown, exception, cancel —
+            # surface that to any in-flight jobs on this worker so their
+            # await never hangs on a future no one will resolve.
+            for jid in [jid for jid, job in self._pending.items() if job.worker_id == handle.worker_id]:
+                pending = self._pending.pop(jid, None)
+                if pending is not None and not pending.future.done():
+                    pending.future.set_exception(
+                        RuntimeError(f"worker {handle.worker_id} result reader exited "
+                                     "with pending jobs"))
+
+
+def _safe_queue_get(q: mp.Queue, timeout: float) -> Any | None:
+    try:
+        return q.get(timeout=timeout)
+    except queue.Empty:
+        return None
+
+
+# ---------------------------------------------------------------------------
+# Worker process
+# ---------------------------------------------------------------------------
+
+
+class WorkerFactory(Protocol):
+
+    def __call__(
+        self,
+        *,
+        gpu_id: int,
+        generator_config: GeneratorConfig,
+        warmup_config: WarmupConfig,
+    ) -> _WorkerHandle:
+        ...
+
+
+def _default_worker_factory(
+    *,
+    gpu_id: int,
+    generator_config: GeneratorConfig,
+    warmup_config: WarmupConfig,
+) -> _WorkerHandle:
+    """Spawn a real multiprocessing worker.
+
+    The child process calls :func:`worker_main` which constructs a
+    :class:`VideoGenerator` from ``generator_config`` and runs a
+    blocking job loop. The ``ready`` event flips after the warmup
+    request completes.
+    """
+    ctx = mp.get_context("spawn")
+    job_queue: mp.Queue = ctx.Queue()
+    result_queue: mp.Queue = ctx.Queue()
+    ready = threading.Event()
+    boot_ok = threading.Event()
+    shutdown_event = ctx.Event()
+    worker_id = f"gpu{gpu_id}-{uuid.uuid4().hex[:6]}"
+    process = ctx.Process(
+        target=worker_main,
+        kwargs={
+            "gpu_id": gpu_id,
+            "worker_id": worker_id,
+            "generator_config": generator_config,
+            "warmup_config": warmup_config,
+            "job_queue": job_queue,
+            "result_queue": result_queue,
+            "shutdown_event": shutdown_event,
+        },
+        daemon=False,
+    )
+    process.start()
+
+    # Block the parent-side ``ready`` flag until the worker posts a
+    # ready acknowledgement on the result queue. We drain that single
+    # sentinel here; subsequent results belong to jobs. ``boot_ok``
+    # only flips on a real ready; on error we set ``ready`` to unblock
+    # the parent's wait but leave ``boot_ok`` clear so the pool keeps
+    # the worker out of the available queue.
+    def _await_ready() -> None:
+        while not shutdown_event.is_set():
+            try:
+                msg = result_queue.get(timeout=1.0)
+            except queue.Empty:
+                continue
+            if isinstance(msg, dict) and msg.get("kind") == "ready":
+                boot_ok.set()
+                ready.set()
+                return
+            if isinstance(msg, dict) and msg.get("kind") == "error":
+                logger.error("pool: worker %s failed to boot: %s", worker_id, msg.get("error"))
+                ready.set()
+                return
+
+    threading.Thread(target=_await_ready, daemon=True).start()
+
+    return _WorkerHandle(
+        process=process,
+        job_queue=job_queue,
+        result_queue=result_queue,
+        gpu_id=gpu_id,
+        worker_id=worker_id,
+        ready=ready,
+        boot_ok=boot_ok,
+        shutdown_event=shutdown_event,
+    )
+
+
+__all__ = [
+    "GpuPool",
+    "InProcessGpuPool",
+    "PoolAcquireTimeout",
+    "PoolAssignment",
+    "PoolHealth",
+    "SubprocessGpuPool",
+    "WorkerFactory",
+    "worker_main",
+]

fastvideo/entrypoints/streaming/mock_server.py

@@ -0,0 +1,122 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Mock streaming server — a frontend dev aid.
+
+Boots the same FastAPI app the real streaming server uses, but backs
+it with :class:`InProcessGpuPool` wrapping a synthetic generator that
+emits pre-baked RGB frames. No GPU or model weights required.
+
+Use cases:
+
+* Frontend development without a real model loaded.
+* Integration tests that exercise the WS protocol end-to-end.
+* Reproducing protocol bugs locally.
+
+Launch: ``python -m fastvideo.entrypoints.streaming.mock_server``.
+"""
+from __future__ import annotations
+
+import argparse
+import time
+from dataclasses import dataclass
+from typing import Any
+
+import numpy as np
+
+from fastvideo.api.schema import (
+    ContinuationState,
+    GenerationRequest,
+    GeneratorConfig,
+    SamplingConfig,
+    ServeConfig,
+    StreamingConfig,
+)
+from fastvideo.entrypoints.streaming.server import build_app
+
+
+@dataclass
+class MockGenerator:
+    """Generator stand-in that returns synthetic gradient frames.
+
+    Each call produces one segment worth of frames whose pixels vary by
+    a constant derived from the request seed and segment index. Latency
+    is configurable via ``sleep_ms`` so the caller can exercise slow-
+    generate scenarios without spinning a GPU.
+    """
+
+    sleep_ms: float = 0.0
+
+    def generate(self, request: GenerationRequest) -> dict[str, Any]:
+        if self.sleep_ms:
+            time.sleep(self.sleep_ms / 1000.0)
+        width = max(16, request.sampling.width)
+        height = max(16, request.sampling.height)
+        num_frames = max(1, request.sampling.num_frames)
+        frames = [_gradient_frame(height, width, idx, seed=request.sampling.seed) for idx in range(num_frames)]
+        state = ContinuationState(
+            kind="ltx2.v1",
+            payload={
+                "schema_version": 1,
+                "segment_index": 0,
+                "source_prompt": request.prompt,
+            },
+        )
+        return {
+            "frames": frames,
+            "audio_sample_rate": 24000,
+            "state": state,
+        }
+
+
+def _gradient_frame(height: int, width: int, idx: int, *, seed: int) -> np.ndarray:
+    base = (idx * 17 + seed * 3) % 256
+    row = np.linspace(base, (base + 64) % 256, width, dtype=np.uint8)
+    frame = np.tile(row, (height, 1))
+    stacked = np.stack([frame, np.roll(frame, 8, axis=1), np.roll(frame, 16, axis=1)], axis=-1)
+    return stacked.astype(np.uint8)
+
+
+def build_mock_app(*, sleep_ms: float = 0.0):
+    """Build a FastAPI app backed by :class:`MockGenerator`."""
+    serve_config = ServeConfig(
+        generator=GeneratorConfig(model_path="/models/mock"),
+        streaming=StreamingConfig(
+            session_timeout_seconds=120,
+            generation_segment_cap=6,
+        ),
+    )
+    serve_config.default_request.sampling = SamplingConfig(
+        num_frames=24,
+        height=256,
+        width=256,
+        fps=24,
+        num_inference_steps=1,
+    )
+    return build_app(serve_config, MockGenerator(sleep_ms=sleep_ms))
+
+
+def main() -> None:  # pragma: no cover - CLI entry
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--host", default="127.0.0.1")
+    parser.add_argument("--port", type=int, default=8000)
+    parser.add_argument(
+        "--sleep-ms",
+        type=float,
+        default=0.0,
+        help="Per-segment artificial latency for testing slow paths",
+    )
+    args = parser.parse_args()
+
+    import uvicorn
+
+    app = build_mock_app(sleep_ms=args.sleep_ms)
+    uvicorn.run(app, host=args.host, port=args.port)
+
+
+__all__ = [
+    "MockGenerator",
+    "build_mock_app",
+    "main",
+]
+
+if __name__ == "__main__":  # pragma: no cover - CLI entry
+    main()

fastvideo/entrypoints/streaming/prompt/__init__.py

@@ -0,0 +1,36 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Prompt pipeline for the streaming server.
+
+* :mod:`providers` — LLM backend abstraction + built-in adapters
+* :mod:`enhancer` — provider-agnostic enhance / auto-extend / rewrite
+  operations on top of the provider layer
+
+All of this is optional; the streaming server runs fine without it
+(PR 7.5's skeleton never invokes the enhancer). When the operator
+enables ``ServeConfig.streaming.prompt.enabled``, the server routes
+each ``session_init_v2`` curated prompt through ``enhance`` before the
+first segment.
+"""
+from fastvideo.entrypoints.streaming.prompt.enhancer import (
+    PromptEnhancer,
+    PromptOperation,
+)
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMMessage,
+    LLMProvider,
+    LLMProviderError,
+    LLMRequest,
+    LLMResponse,
+    LLMTimeoutError,
+)
+
+__all__ = [
+    "LLMMessage",
+    "LLMProvider",
+    "LLMProviderError",
+    "LLMRequest",
+    "LLMResponse",
+    "LLMTimeoutError",
+    "PromptEnhancer",
+    "PromptOperation",
+]

fastvideo/entrypoints/streaming/prompt/enhancer.py

@@ -0,0 +1,197 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Provider-agnostic prompt orchestration for the streaming server.
+
+Three operations the streaming server needs:
+
+* ``enhance`` — polish a user prompt (add cinematic detail, fix syntax)
+* ``auto_extend`` — generate a follow-on prompt for loop generation
+* ``rewrite`` — rewrite a seed prompt for a user-directed rewrite flow
+
+All three share the same orchestration: pick a provider in priority
+order, submit an ``LLMRequest``, fall back to the next provider on
+retryable errors, and surface a structured :class:`LLMResponse` back
+to the caller.
+
+System prompts are loaded from ``system_prompt_dir`` on construction
+and can be hot-reloaded via :meth:`PromptEnhancer.reload_system_prompts`.
+The streaming server's management endpoint calls that method in
+response to a ``rewrite_seed_prompts_started`` frame.
+"""
+from __future__ import annotations
+
+import enum
+import os
+from collections.abc import Sequence
+from dataclasses import dataclass, replace
+
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMMessage,
+    LLMProvider,
+    LLMProviderError,
+    LLMRequest,
+    LLMResponse,
+)
+from fastvideo.logger import init_logger
+
+logger = init_logger(__name__)
+
+
+class PromptOperation(enum.Enum):
+    ENHANCE = "enhance"
+    AUTO_EXTEND = "auto_extend"
+    REWRITE = "rewrite"
+
+
+@dataclass
+class _SystemPrompts:
+    enhance: str
+    auto_extend: str
+    rewrite: str
+
+
+_DEFAULT_SYSTEM_PROMPTS = _SystemPrompts(
+    enhance=("You are a prompt enhancer for cinematic video generation. Given "
+             "a user prompt, produce an enhanced prompt that is more vivid, "
+             "specific, and concrete. Keep the subject intact; add lighting, "
+             "camera, and motion detail. Reply with just the enhanced prompt."),
+    auto_extend=("You are a video continuation assistant. Given the current "
+                 "sequence of prompts, produce one new prompt that naturally "
+                 "continues the sequence. Reply with just the next prompt."),
+    rewrite=("You are a creative prompt rewriter. Given a seed prompt, produce "
+             "a set of alternative prompts that explore different angles, "
+             "styles, and moods. Reply with one prompt per line."),
+)
+
+
+class PromptEnhancer:
+    """Orchestrates prompt operations across a priority-ordered provider
+    list with structured fallback + hot-reloadable system prompts.
+
+    Usage::
+
+        enhancer = PromptEnhancer(
+            providers=[CerebrasProvider(), GroqProvider()],
+            model="gpt-oss-120b",
+            system_prompt_dir="/etc/fastvideo/prompts",
+        )
+        response = await enhancer.enhance("a fox running through snow")
+    """
+
+    def __init__(
+        self,
+        *,
+        providers: Sequence[LLMProvider],
+        model: str,
+        timeout_ms: int = 20000,
+        temperature: float = 0.7,
+        max_tokens: int | None = 256,
+        system_prompt_dir: str | None = None,
+    ) -> None:
+        if not providers:
+            raise ValueError("PromptEnhancer requires at least one LLMProvider")
+        self._providers = list(providers)
+        self._model = model
+        self._timeout_ms = timeout_ms
+        self._temperature = temperature
+        self._max_tokens = max_tokens
+        self._system_prompt_dir = system_prompt_dir
+        self._system_prompts = self._load_system_prompts()
+
+    @property
+    def providers(self) -> list[LLMProvider]:
+        return list(self._providers)
+
+    def register_provider(self, provider: LLMProvider, *, priority: int = -1) -> None:
+        """Insert an additional provider. ``priority=0`` makes it primary;
+        ``priority=-1`` (default) appends as a fallback."""
+        if priority < 0:
+            self._providers.append(provider)
+        else:
+            self._providers.insert(priority, provider)
+
+    def reload_system_prompts(self) -> None:
+        """Re-read the system prompt files from ``system_prompt_dir``.
+
+        The streaming server exposes this via a management endpoint so
+        operators can iterate on prompt templates without restarting
+        workers.
+        """
+        self._system_prompts = self._load_system_prompts()
+        logger.info("prompt enhancer: reloaded system prompts from %s", self._system_prompt_dir or "defaults")
+
+    async def enhance(self, prompt: str) -> LLMResponse:
+        return await self._run(
+            PromptOperation.ENHANCE,
+            system=self._system_prompts.enhance,
+            user=prompt,
+        )
+
+    async def auto_extend(self, prior_prompts: Sequence[str]) -> LLMResponse:
+        user = "\n".join(prior_prompts)
+        return await self._run(
+            PromptOperation.AUTO_EXTEND,
+            system=self._system_prompts.auto_extend,
+            user=user,
+        )
+
+    async def rewrite(self, seed_prompt: str) -> LLMResponse:
+        return await self._run(
+            PromptOperation.REWRITE,
+            system=self._system_prompts.rewrite,
+            user=seed_prompt,
+        )
+
+    async def _run(
+        self,
+        operation: PromptOperation,
+        *,
+        system: str,
+        user: str,
+    ) -> LLMResponse:
+        request = LLMRequest(
+            messages=[
+                LLMMessage(role="system", content=system),
+                LLMMessage(role="user", content=user),
+            ],
+            model=self._model,
+            max_tokens=self._max_tokens,
+            temperature=self._temperature,
+            timeout_ms=self._timeout_ms,
+        )
+        last_error: LLMProviderError | None = None
+        for idx, provider in enumerate(self._providers):
+            try:
+                response = await provider.complete(request)
+                if idx > 0:
+                    # Mark the fallback flag without losing any other
+                    # response fields the provider populated.
+                    response = replace(response, fallback_used=True)
+                return response
+            except LLMProviderError as exc:
+                logger.warning("prompt %s: provider %s failed: %s; trying next", operation.value, provider.name, exc)
+                last_error = exc
+                if not exc.retryable:
+                    break
+        assert last_error is not None
+        raise last_error
+
+    def _load_system_prompts(self) -> _SystemPrompts:
+        if not self._system_prompt_dir:
+            return _DEFAULT_SYSTEM_PROMPTS
+        return _SystemPrompts(
+            enhance=_read_prompt(self._system_prompt_dir, "enhance.txt", _DEFAULT_SYSTEM_PROMPTS.enhance),
+            auto_extend=_read_prompt(self._system_prompt_dir, "auto_extend.txt", _DEFAULT_SYSTEM_PROMPTS.auto_extend),
+            rewrite=_read_prompt(self._system_prompt_dir, "rewrite.txt", _DEFAULT_SYSTEM_PROMPTS.rewrite),
+        )
+
+
+def _read_prompt(dirname: str, filename: str, default: str) -> str:
+    path = os.path.join(dirname, filename)
+    if not os.path.exists(path):
+        return default
+    with open(path, encoding="utf-8") as f:
+        content = f.read().strip()
+    return content or default
+
+
+__all__ = ["PromptEnhancer", "PromptOperation"]

fastvideo/entrypoints/streaming/prompt/providers/__init__.py

@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: Apache-2.0
+"""LLM provider implementations used by the prompt enhancer."""
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMMessage,
+    LLMProvider,
+    LLMProviderError,
+    LLMRequest,
+    LLMResponse,
+    LLMTimeoutError,
+)
+from fastvideo.entrypoints.streaming.prompt.providers.cerebras import (
+    CerebrasProvider, )
+from fastvideo.entrypoints.streaming.prompt.providers.groq import GroqProvider
+
+__all__ = [
+    "CerebrasProvider",
+    "GroqProvider",
+    "LLMMessage",
+    "LLMProvider",
+    "LLMProviderError",
+    "LLMRequest",
+    "LLMResponse",
+    "LLMTimeoutError",
+]

fastvideo/entrypoints/streaming/prompt/providers/_openai_compat.py

@@ -0,0 +1,101 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Shared HTTP path for OpenAI-compatible ``/chat/completions`` providers.
+
+Cerebras and Groq both expose the OpenAI chat-completions schema, so
+the request shape, error mapping, and response decoding are identical
+between them. This module centralizes that logic; the per-provider
+modules stay thin (just defaults + env var wiring).
+"""
+from __future__ import annotations
+
+import time
+
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMProviderError,
+    LLMRequest,
+    LLMResponse,
+    LLMTimeoutError,
+)
+
+
+async def complete_openai_compatible(
+    *,
+    api_key: str | None,
+    api_key_hint: str,
+    base_url: str,
+    provider_name: str,
+    request: LLMRequest,
+) -> LLMResponse:
+    """Issue a chat-completions call and decode the OpenAI response."""
+    if not api_key:
+        raise LLMProviderError(
+            f"{provider_name} provider requires {api_key_hint} "
+            "(or explicit api_key=...)",
+            retryable=False,
+        )
+    try:
+        import httpx
+    except ImportError as exc:  # pragma: no cover - optional dep
+        raise LLMProviderError(
+            f"{provider_name} provider requires httpx; install httpx",
+            retryable=False,
+        ) from exc
+
+    timeout_s = (request.timeout_ms or 20000) / 1000.0
+    t0 = time.perf_counter()
+    try:
+        async with httpx.AsyncClient(timeout=timeout_s) as client:
+            response = await client.post(
+                f"{base_url}/chat/completions",
+                headers={
+                    "Authorization": f"Bearer {api_key}",
+                    "Content-Type": "application/json",
+                },
+                json={
+                    "model": request.model,
+                    "messages": [{
+                        "role": m.role,
+                        "content": m.content
+                    } for m in request.messages],
+                    "max_tokens": request.max_tokens,
+                    "temperature": request.temperature,
+                },
+            )
+    except httpx.TimeoutException as exc:
+        raise LLMTimeoutError(f"{provider_name} timed out after {timeout_s}s") from exc
+    except httpx.HTTPError as exc:
+        raise LLMProviderError(f"{provider_name} HTTP error: {exc}") from exc
+
+    if response.status_code >= 400:
+        # 5xx and 429 (rate-limit) are retryable: another provider may
+        # succeed. 4xx (auth, bad-request, etc.) are client errors —
+        # the enhancer should stop fallback traversal.
+        retryable = (response.status_code >= 500 or response.status_code == 429)
+        raise LLMProviderError(
+            f"{provider_name} returned {response.status_code}: "
+            f"{response.text[:200]}",
+            retryable=retryable,
+        )
+
+    try:
+        data = response.json()
+    except Exception as exc:
+        # Non-JSON body usually means a proxy / load-balancer error
+        # page; leave it retryable so a fallback provider can try.
+        raise LLMProviderError(f"{provider_name} returned non-JSON body: {exc}") from exc
+
+    choices = data.get("choices") or []
+    if not choices:
+        raise LLMProviderError(f"{provider_name} returned no choices")
+    content = choices[0].get("message", {}).get("content") or ""
+
+    latency_ms = (time.perf_counter() - t0) * 1000.0
+    return LLMResponse(
+        content=content.strip(),
+        provider=provider_name,
+        model=request.model,
+        latency_ms=latency_ms,
+    )
+
+
+__all__ = ["complete_openai_compatible"]

fastvideo/entrypoints/streaming/prompt/providers/base.py

@@ -0,0 +1,85 @@
+# SPDX-License-Identifier: Apache-2.0
+"""LLM provider protocol + DTOs used by the prompt enhancer.
+
+Third-party users add a new provider by implementing
+:class:`LLMProvider` and registering it with a prompt enhancer
+instance. The shipped providers live in sibling modules
+(``cerebras.py``, ``groq.py``) and each is ~100-200 LOC — the
+provider layer is intentionally thin so the enhancer stays
+provider-agnostic.
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Literal, Protocol, runtime_checkable
+
+
+@dataclass
+class LLMMessage:
+    role: Literal["system", "user", "assistant"]
+    content: str
+
+
+@dataclass
+class LLMRequest:
+    messages: list[LLMMessage]
+    model: str
+    max_tokens: int | None = None
+    temperature: float | None = None
+    timeout_ms: int | None = None
+
+
+@dataclass
+class LLMResponse:
+    content: str
+    provider: str
+    model: str
+    latency_ms: float
+    fallback_used: bool = False
+
+
+class LLMProviderError(RuntimeError):
+    """Raised when an LLM provider fails a request.
+
+    ``retryable`` controls whether the enhancer falls back to the next
+    provider. It is settable per-instance so the same exception type
+    can describe retryable transport errors (5xx, 429) and
+    non-retryable client errors (4xx auth/bad-request) without forcing
+    a separate subclass for every status family.
+    """
+
+    def __init__(self, message: str, *, retryable: bool = True) -> None:
+        super().__init__(message)
+        self.retryable = retryable
+
+
+class LLMTimeoutError(LLMProviderError):
+    """Raised when an LLM provider times out — always retryable."""
+
+    def __init__(self, message: str) -> None:
+        super().__init__(message, retryable=True)
+
+
+@runtime_checkable
+class LLMProvider(Protocol):
+    """Provider interface every LLM adapter implements.
+
+    Providers are async-first because every built-in implementation
+    talks to an HTTP API. Synchronous providers can wrap their call in
+    ``asyncio.to_thread`` internally.
+    """
+
+    name: str
+
+    async def complete(self, request: LLMRequest) -> LLMResponse:
+        ...
+
+
+__all__ = [
+    "LLMMessage",
+    "LLMProvider",
+    "LLMProviderError",
+    "LLMRequest",
+    "LLMResponse",
+    "LLMTimeoutError",
+]

fastvideo/entrypoints/streaming/prompt/providers/cerebras.py

@@ -0,0 +1,44 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Cerebras LLM provider (OpenAI-compatible chat endpoint)."""
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass
+
+from fastvideo.entrypoints.streaming.prompt.providers._openai_compat import (
+    complete_openai_compatible, )
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMRequest,
+    LLMResponse,
+)
+
+_DEFAULT_BASE_URL = "https://api.cerebras.ai/v1"
+_API_KEY_ENV = "CEREBRAS_API_KEY"
+
+
+@dataclass
+class CerebrasProvider:
+    """Cerebras inference adapter.
+
+    ``api_key`` falls back to ``CEREBRAS_API_KEY`` when unset.
+    """
+
+    api_key: str | None = None
+    base_url: str = _DEFAULT_BASE_URL
+    name: str = "cerebras"
+
+    def __post_init__(self) -> None:
+        if self.api_key is None:
+            self.api_key = os.environ.get(_API_KEY_ENV)
+
+    async def complete(self, request: LLMRequest) -> LLMResponse:
+        return await complete_openai_compatible(
+            api_key=self.api_key,
+            api_key_hint=_API_KEY_ENV,
+            base_url=self.base_url,
+            provider_name=self.name,
+            request=request,
+        )
+
+
+__all__ = ["CerebrasProvider"]

fastvideo/entrypoints/streaming/prompt/providers/groq.py

@@ -0,0 +1,46 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Groq LLM provider (OpenAI-compatible chat endpoint)."""
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass
+
+from fastvideo.entrypoints.streaming.prompt.providers._openai_compat import (
+    complete_openai_compatible, )
+from fastvideo.entrypoints.streaming.prompt.providers.base import (
+    LLMRequest,
+    LLMResponse,
+)
+
+_DEFAULT_BASE_URL = "https://api.groq.com/openai/v1"
+_API_KEY_ENV = "GROQ_API_KEY"
+
+
+@dataclass
+class GroqProvider:
+    """Groq inference adapter.
+
+    Identical wire format to :class:`CerebrasProvider`; both go through
+    :func:`complete_openai_compatible`. The two providers differ only
+    in base URL, env var, and model id conventions.
+    """
+
+    api_key: str | None = None
+    base_url: str = _DEFAULT_BASE_URL
+    name: str = "groq"
+
+    def __post_init__(self) -> None:
+        if self.api_key is None:
+            self.api_key = os.environ.get(_API_KEY_ENV)
+
+    async def complete(self, request: LLMRequest) -> LLMResponse:
+        return await complete_openai_compatible(
+            api_key=self.api_key,
+            api_key_hint=_API_KEY_ENV,
+            base_url=self.base_url,
+            provider_name=self.name,
+            request=request,
+        )
+
+
+__all__ = ["GroqProvider"]

fastvideo/entrypoints/streaming/prompt/rewrite.py

@@ -0,0 +1,82 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Rewrite payload builder.
+
+The UI's "rewrite seed prompts" flow asks the enhancer to produce a
+batch of alternative prompts given one seed. This module packages the
+seed + options into the payload the enhancer expects and unpacks the
+response back into a typed :class:`RewriteResult`.
+
+Separating this from :mod:`enhancer` keeps the enhancer provider-
+agnostic; anything UI-specific (how many alternatives to request, how
+to split the response, temperature) lives here.
+"""
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+from fastvideo.entrypoints.streaming.prompt.enhancer import PromptEnhancer
+
+_LEADING_MARKER_RE = re.compile(r"^(?:[-*•]\s*|\d+\s*[.)]\s*)+")
+
+
+@dataclass
+class RewriteOptions:
+    count: int = 3
+    """Number of alternative prompts to request."""
+    temperature: float | None = None
+
+
+@dataclass
+class RewriteResult:
+    seed_prompt: str
+    alternatives: list[str]
+    provider: str
+    model: str
+    latency_ms: float
+    fallback_used: bool = False
+
+
+async def build_rewrite(
+    enhancer: PromptEnhancer,
+    seed_prompt: str,
+    *,
+    options: RewriteOptions | None = None,
+) -> RewriteResult:
+    """Run a rewrite op through the enhancer and return a typed result."""
+    if not seed_prompt.strip():
+        raise ValueError("rewrite seed prompt must be non-empty")
+    options = options or RewriteOptions()
+    response = await enhancer.rewrite(seed_prompt)
+    alternatives = _split_response(response.content, limit=options.count)
+    return RewriteResult(
+        seed_prompt=seed_prompt,
+        alternatives=alternatives,
+        provider=response.provider,
+        model=response.model,
+        latency_ms=response.latency_ms,
+        fallback_used=response.fallback_used,
+    )
+
+
+def _split_response(content: str, *, limit: int) -> list[str]:
+    """Split the LLM response into discrete prompt candidates.
+
+    The shipped system prompt instructs the model to emit one prompt
+    per line; this function is forgiving about numbered lists or
+    leading bullets so user-supplied system prompts don't break it.
+    """
+    lines = [line.strip() for line in content.splitlines() if line.strip()]
+    cleaned: list[str] = []
+    for line in lines:
+        stripped = _LEADING_MARKER_RE.sub("", line).strip()
+        if stripped:
+            cleaned.append(stripped)
+    return cleaned[:max(1, limit)]
+
+
+__all__ = [
+    "RewriteOptions",
+    "RewriteResult",
+    "build_rewrite",
+]

fastvideo/entrypoints/streaming/prompt/safety.py

@@ -0,0 +1,146 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Optional prompt safety filter.
+
+Uses a fastText classifier to score prompts against a banned-content
+rubric. Only loaded when ``ServeConfig.streaming.safety.enabled`` is
+True and fastText is installed — users who don't need it see no
+runtime cost.
+
+Install: ``pip install fastvideo[prompt-safety]`` (ships fasttext as an
+optional extra) or install fasttext directly.
+"""
+from __future__ import annotations
+
+import enum
+import threading
+from dataclasses import dataclass
+from typing import Any
+
+from fastvideo.logger import init_logger
+
+logger = init_logger(__name__)
+
+
+class SafetyDecision(enum.Enum):
+    ALLOW = "allow"
+    BLOCK = "block"
+    UNAVAILABLE = "unavailable"
+    """Returned when the classifier can't run (not configured, fastText
+    missing). Safety is opt-in; the server treats ``UNAVAILABLE`` as
+    ``ALLOW`` but logs it so operators know the filter is off."""
+
+
+@dataclass
+class SafetyResult:
+    prompt: str
+    decision: SafetyDecision
+    score: float = 0.0
+    label: str | None = None
+    reason: str | None = None
+
+
+class PromptSafetyFilter:
+    """Minimal fastText-backed prompt safety filter.
+
+    Loads the classifier lazily on first use so the streaming server
+    can construct the filter eagerly at startup without paying the
+    model-load cost when safety is disabled.
+    """
+
+    def __init__(
+        self,
+        *,
+        classifier_path: str | None,
+        enabled: bool = True,
+        block_threshold: float = 0.5,
+    ) -> None:
+        self._classifier_path = classifier_path
+        self._enabled = enabled
+        self._block_threshold = block_threshold
+        self._model: Any | None = None
+        self._load_attempted = False
+        self._load_lock = threading.Lock()
+
+    @property
+    def enabled(self) -> bool:
+        return self._enabled and self._classifier_path is not None
+
+    def classify(self, prompt: str) -> SafetyResult:
+        if not self.enabled:
+            return SafetyResult(
+                prompt=prompt,
+                decision=SafetyDecision.UNAVAILABLE,
+                reason="safety filter not enabled",
+            )
+        model = self._ensure_loaded()
+        if model is None:
+            return SafetyResult(
+                prompt=prompt,
+                decision=SafetyDecision.UNAVAILABLE,
+                reason="fastText model unavailable",
+            )
+        try:
+            labels, probs = model.predict(prompt.replace("\n", " "), k=1)
+        except Exception as exc:  # pragma: no cover - defensive
+            logger.warning("safety: classifier failed: %s", exc)
+            return SafetyResult(
+                prompt=prompt,
+                decision=SafetyDecision.UNAVAILABLE,
+                reason=f"classifier error: {exc}",
+            )
+        label = labels[0].removeprefix("__label__") if labels else None
+        score = float(probs[0]) if len(probs) else 0.0
+        decision = (SafetyDecision.BLOCK if
+                    (label == "unsafe" and score >= self._block_threshold) else SafetyDecision.ALLOW)
+        return SafetyResult(
+            prompt=prompt,
+            decision=decision,
+            score=score,
+            label=label,
+        )
+
+    def _ensure_loaded(self) -> Any | None:
+        if self._model is not None:
+            return self._model
+        if self._load_attempted:
+            return None
+        with self._load_lock:
+            if self._model is not None:
+                return self._model
+            if self._load_attempted:
+                return None
+            self._load_attempted = True
+            if self._classifier_path is None:
+                return None
+            try:
+                import fasttext  # type: ignore[import-not-found]
+            except ImportError:
+                logger.warning("safety: fasttext not installed; safety filter disabled. "
+                               "Install fastvideo[prompt-safety] to enable.")
+                return None
+            try:
+                self._model = fasttext.load_model(self._classifier_path)
+            except Exception as exc:  # pragma: no cover - requires real model
+                logger.warning("safety: failed to load %s: %s", self._classifier_path, exc)
+                return None
+            return self._model
+
+
+def first_blocked(
+    filter_: PromptSafetyFilter,
+    prompts: list[str],
+) -> SafetyResult | None:
+    """Return the first prompt the filter blocks, or ``None``."""
+    for prompt in prompts:
+        result = filter_.classify(prompt)
+        if result.decision is SafetyDecision.BLOCK:
+            return result
+    return None
+
+
+__all__ = [
+    "PromptSafetyFilter",
+    "SafetyDecision",
+    "SafetyResult",
+    "first_blocked",
+]

fastvideo/entrypoints/streaming/router/__init__.py

@@ -0,0 +1,27 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Multi-replica load balancer + WebSocket proxy for the streaming server.
+
+Sits in front of one-or-more streaming-server replicas and forwards
+WebSocket sessions to a healthy primary, with failover to secondaries.
+Kept in-repo under ``fastvideo/entrypoints/streaming/router/`` per the
+PR plan's default; the alternative (separate package) is an open
+question deferred to review.
+"""
+from fastvideo.entrypoints.streaming.router.registry import (
+    Replica,
+    ReplicaHealth,
+    ReplicaRegistry,
+    ReplicaStatus,
+)
+from fastvideo.entrypoints.streaming.router.config import RouterConfig
+from fastvideo.entrypoints.streaming.router.main import build_router_app, run_router
+
+__all__ = [
+    "Replica",
+    "ReplicaHealth",
+    "ReplicaRegistry",
+    "ReplicaStatus",
+    "RouterConfig",
+    "build_router_app",
+    "run_router",
+]

fastvideo/entrypoints/streaming/router/config.py

@@ -0,0 +1,88 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Typed router configuration."""
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from urllib.parse import urlparse
+
+
+@dataclass
+class ReplicaEndpoint:
+    """One backend replica the router can route to."""
+
+    url: str
+    """HTTP base URL, e.g. ``http://host:8000``. WebSocket URL is
+    derived automatically by replacing the scheme."""
+    name: str | None = None
+    primary: bool = False
+    """``True`` = prefer this replica over others in steady state."""
+    weight: float = 1.0
+
+
+@dataclass
+class RouterConfig:
+    """Typed router config loaded from a YAML file.
+
+    Example::
+
+        router:
+          host: 0.0.0.0
+          port: 9000
+          replicas:
+            - url: http://streamer-a:8000
+              primary: true
+            - url: http://streamer-b:8000
+          health_check:
+            path: /health
+            interval_seconds: 5
+            failure_threshold: 3
+
+    Validation runs in ``__post_init__``: empty replicas, non-positive
+    intervals/timeouts, thresholds < 1, non-http(s) URLs, and more than
+    one primary all raise ``ValueError`` so misconfigurations surface at
+    load time rather than as confusing runtime failures.
+    """
+
+    host: str = "0.0.0.0"
+    port: int = 9000
+    replicas: list[ReplicaEndpoint] = field(default_factory=list)
+    health_check_path: str = "/health"
+    health_check_interval_seconds: float = 5.0
+    health_check_timeout_seconds: float = 2.0
+    failure_threshold: int = 3
+    recovery_threshold: int = 2
+
+    def __post_init__(self) -> None:
+        if not self.replicas:
+            raise ValueError("RouterConfig.replicas must list at least one replica")
+        if self.health_check_interval_seconds <= 0:
+            raise ValueError(f"health_check_interval_seconds must be > 0, got {self.health_check_interval_seconds}")
+        if self.health_check_timeout_seconds <= 0:
+            raise ValueError(f"health_check_timeout_seconds must be > 0, got {self.health_check_timeout_seconds}")
+        if self.failure_threshold < 1:
+            raise ValueError(f"failure_threshold must be >= 1, got {self.failure_threshold}")
+        if self.recovery_threshold < 1:
+            raise ValueError(f"recovery_threshold must be >= 1, got {self.recovery_threshold}")
+        seen_urls: set[str] = set()
+        for replica in self.replicas:
+            if not replica.url.startswith(("http://", "https://")):
+                raise ValueError(f"ReplicaEndpoint.url must start with http:// or https://, got {replica.url!r}")
+            parsed = urlparse(replica.url)
+            if parsed.path not in ("", "/"):
+                raise ValueError(f"ReplicaEndpoint.url must be a base host[:port] URL without a path; "
+                                 f"got {replica.url!r} with path {parsed.path!r}. The router appends "
+                                 "`/health` and `/v1/stream` itself.")
+            if parsed.query or parsed.fragment:
+                raise ValueError(f"ReplicaEndpoint.url must not include query/fragment; got {replica.url!r}")
+            if replica.url in seen_urls:
+                raise ValueError(f"Duplicate ReplicaEndpoint.url {replica.url!r}; "
+                                 "router selection keys by URL so duplicates would silently collapse")
+            seen_urls.add(replica.url)
+        primaries = sum(1 for r in self.replicas if r.primary)
+        if primaries > 1:
+            raise ValueError(f"RouterConfig allows at most one primary replica; got {primaries}. "
+                             "Multi-primary load distribution is deferred — promote one replica to "
+                             "primary and treat the rest as secondaries.")
+
+
+__all__ = ["ReplicaEndpoint", "RouterConfig"]

fastvideo/entrypoints/streaming/router/main.py

@@ -0,0 +1,218 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Router FastAPI entry point.
+
+Exposes the same ``/v1/stream`` WebSocket path the backend servers do,
+accepts a client, picks a healthy replica from the registry, and
+proxies frames bidirectionally.
+
+PR 7.9 ships the minimum-viable shape: explicit replica list, single
+primary, JSON + binary passthrough in both directions, and a
+``/status`` endpoint for operators. Sticky-session routing (so a
+reconnect lands on the same backend) is left for a follow-up.
+"""
+from __future__ import annotations
+
+import asyncio
+import contextlib
+from dataclasses import dataclass
+
+from fastapi import FastAPI, WebSocket, WebSocketDisconnect
+from fastapi.responses import JSONResponse
+
+from fastvideo.entrypoints.streaming.router.config import RouterConfig
+from fastvideo.entrypoints.streaming.router.registry import (
+    ReplicaRegistry,
+    run_health_check_loop,
+)
+from fastvideo.logger import init_logger
+
+logger = init_logger(__name__)
+
+
+@dataclass
+class _RouterState:
+    config: RouterConfig
+    registry: ReplicaRegistry
+    stop_event: asyncio.Event
+    health_task: asyncio.Task | None = None
+
+
+def build_router_app(
+    config: RouterConfig,
+    *,
+    registry: ReplicaRegistry | None = None,
+) -> FastAPI:
+    """Build the router FastAPI app.
+
+    ``registry`` can be injected for tests; defaults to one built from
+    ``config.replicas``.
+    """
+    registry = registry or ReplicaRegistry(config.replicas)
+    state = _RouterState(
+        config=config,
+        registry=registry,
+        stop_event=asyncio.Event(),
+    )
+
+    @contextlib.asynccontextmanager
+    async def _lifespan(_app: FastAPI):
+        state.health_task = asyncio.create_task(
+            run_health_check_loop(
+                registry=state.registry,
+                config=state.config,
+                stop_event=state.stop_event,
+            ))
+        try:
+            yield
+        finally:
+            state.stop_event.set()
+            if state.health_task is not None:
+                with contextlib.suppress(asyncio.CancelledError):
+                    await state.health_task
+
+    app = FastAPI(title="FastVideo Streaming Router", lifespan=_lifespan)
+
+    @app.get("/status")
+    async def _status() -> JSONResponse:
+        return JSONResponse({
+            "replicas": [{
+                "url": r.url,
+                "primary": r.primary,
+                "status": r.health.status.value,
+                "last_ok_at": r.health.last_ok_at,
+                "last_latency_ms": r.health.last_latency_ms,
+                "consecutive_failures": r.health.consecutive_failures,
+            } for r in state.registry.all()],
+        })
+
+    @app.websocket("/v1/stream")
+    async def _proxy(websocket: WebSocket) -> None:
+        await websocket.accept()
+        replica = state.registry.select()
+        if replica is None:
+            await websocket.send_json({
+                "type": "error",
+                "code": "gpu_unavailable",
+                "message": "router: no healthy replica available",
+                "retryable": True,
+            })
+            await websocket.close(code=1013, reason="no_healthy_replica")
+            return
+
+        ws_url = _websocket_url_for(replica.url)
+        try:
+            await _bridge_session(websocket, ws_url)
+        except WebSocketDisconnect:
+            logger.info("router: client disconnected")
+        except Exception as exc:
+            logger.exception("router: bridge failed: %s", exc)
+            with contextlib.suppress(RuntimeError):
+                await websocket.send_json({
+                    "type": "error",
+                    "code": "worker_failed",
+                    "message": f"router bridge failed: {exc}",
+                    "retryable": True,
+                })
+            with contextlib.suppress(RuntimeError):
+                await websocket.close(code=1011)
+
+    app.state.router_state = state
+    return app
+
+
+def run_router(config: RouterConfig) -> None:  # pragma: no cover - CLI
+    import uvicorn
+
+    app = build_router_app(config)
+    uvicorn.run(app, host=config.host, port=config.port)
+
+
+async def _bridge_session(
+    client_ws: WebSocket,
+    backend_ws_url: str,
+) -> None:
+    """Connect to backend and shuttle messages in both directions.
+
+    Uses ``websockets`` for the backend side; imported lazily to keep
+    the router's import graph small for users who only want the server.
+
+    Cancellation: when either direction completes (client disconnect,
+    backend close, exception), the other is cancelled explicitly and
+    both are drained before returning. Unexpected exceptions from the
+    direction that completed first are re-raised; normal disconnect
+    paths (``WebSocketDisconnect``, ``ConnectionClosed``,
+    ``CancelledError``) are swallowed.
+    """
+    try:
+        import websockets
+    except ImportError as exc:  # pragma: no cover - optional extra
+        raise RuntimeError("router requires the `websockets` package for backend proxying") from exc
+
+    async with websockets.connect(backend_ws_url + "/v1/stream") as backend_ws:
+        c2b = asyncio.create_task(_forward_client_to_backend(client_ws, backend_ws))
+        b2c = asyncio.create_task(_forward_backend_to_client(backend_ws, client_ws))
+        try:
+            done, _pending = await asyncio.wait(
+                {c2b, b2c},
+                return_when=asyncio.FIRST_COMPLETED,
+            )
+        finally:
+            for task in (c2b, b2c):
+                if not task.done():
+                    task.cancel()
+            await asyncio.gather(c2b, b2c, return_exceptions=True)
+        for task in done:
+            task_exc = task.exception()
+            if task_exc is not None and not _is_normal_disconnect(task_exc):
+                raise task_exc
+
+
+def _is_normal_disconnect(exc: BaseException) -> bool:
+    """Whether ``exc`` is a routine WebSocket teardown vs a real bridge fault."""
+    if isinstance(exc, asyncio.CancelledError | WebSocketDisconnect):
+        return True
+    name = type(exc).__name__
+    # websockets.exceptions.ConnectionClosed{,OK,Error} all subclass
+    # WebSocketException; check by name to avoid the lazy-import dance.
+    return name.startswith("ConnectionClosed")
+
+
+async def _forward_client_to_backend(client_ws: WebSocket, backend_ws) -> None:
+    try:
+        while True:
+            msg = await client_ws.receive()
+            if msg.get("type") == "websocket.disconnect":
+                break
+            if "text" in msg and msg["text"] is not None:
+                await backend_ws.send(msg["text"])
+            elif "bytes" in msg and msg["bytes"] is not None:
+                await backend_ws.send(msg["bytes"])
+    finally:
+        with contextlib.suppress(Exception):
+            await backend_ws.close()
+
+
+async def _forward_backend_to_client(backend_ws, client_ws: WebSocket) -> None:
+    try:
+        async for frame in backend_ws:
+            if isinstance(frame, bytes):
+                await client_ws.send_bytes(frame)
+            else:
+                await client_ws.send_text(frame)
+    finally:
+        with contextlib.suppress(Exception):
+            await client_ws.close()
+
+
+def _websocket_url_for(http_url: str) -> str:
+    if http_url.startswith("https://"):
+        return "wss://" + http_url[len("https://"):]
+    if http_url.startswith("http://"):
+        return "ws://" + http_url[len("http://"):]
+    return http_url
+
+
+__all__ = [
+    "build_router_app",
+    "run_router",
+]

fastvideo/entrypoints/streaming/router/registry.py

@@ -0,0 +1,268 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Replica registry + health-check loop.
+
+The registry tracks the set of known backend replicas and their live
+health. The router consults it for "pick a backend for this session"
+decisions and a background task updates it from periodic HTTP probes.
+
+State machine per replica::
+
+    HEALTHY ──(N consecutive failures)──▶ UNHEALTHY
+       ▲                                     │
+       └──────(M consecutive successes)──────┘
+
+Where N = :attr:`RouterConfig.failure_threshold` and
+M = :attr:`RouterConfig.recovery_threshold`.
+"""
+from __future__ import annotations
+
+import asyncio
+import contextlib
+import enum
+import time
+from collections.abc import AsyncIterator, Awaitable, Callable
+from dataclasses import dataclass, field
+from typing import Any
+
+from fastvideo.entrypoints.streaming.router.config import (
+    ReplicaEndpoint,
+    RouterConfig,
+)
+from fastvideo.logger import init_logger
+
+HttpProbe = Any
+"""Structural alias for health-probe callables. Concrete signature is
+``async def __call__(url: str, *, timeout: float) -> tuple[float,
+str | None]``; typing.Callable cannot express keyword-only parameters,
+so duck-typing is the pragmatic compromise."""
+
+logger = init_logger(__name__)
+
+
+class ReplicaStatus(enum.Enum):
+    UNKNOWN = "unknown"
+    HEALTHY = "healthy"
+    UNHEALTHY = "unhealthy"
+
+
+@dataclass
+class ReplicaHealth:
+    status: ReplicaStatus = ReplicaStatus.UNKNOWN
+    last_ok_at: float | None = None
+    last_failure_at: float | None = None
+    consecutive_failures: int = 0
+    consecutive_successes: int = 0
+    last_latency_ms: float | None = None
+
+
+@dataclass
+class Replica:
+    endpoint: ReplicaEndpoint
+    health: ReplicaHealth = field(default_factory=ReplicaHealth)
+
+    @property
+    def url(self) -> str:
+        return self.endpoint.url
+
+    @property
+    def primary(self) -> bool:
+        return self.endpoint.primary
+
+    @property
+    def is_healthy(self) -> bool:
+        return self.health.status is ReplicaStatus.HEALTHY
+
+
+class ReplicaRegistry:
+    """Stateful map of replica URL → :class:`Replica`.
+
+    Selection favors primary replicas when healthy; otherwise the first
+    healthy non-primary is returned. When none are healthy, the
+    registry returns ``None`` so the router can reject incoming
+    sessions with ``gpu_unavailable``.
+    """
+
+    def __init__(self, replicas: list[ReplicaEndpoint]) -> None:
+        if not replicas:
+            raise ValueError("ReplicaRegistry requires at least one replica")
+        self._replicas: dict[str, Replica] = {endpoint.url: Replica(endpoint=endpoint) for endpoint in replicas}
+        self._lock = asyncio.Lock()
+
+    def all(self) -> list[Replica]:
+        return list(self._replicas.values())
+
+    def get(self, url: str) -> Replica | None:
+        return self._replicas.get(url)
+
+    def primaries(self) -> list[Replica]:
+        return [r for r in self._replicas.values() if r.primary]
+
+    def select(self) -> Replica | None:
+        """Pick the best healthy replica.
+
+        Priority order:
+
+        1. The first healthy primary (insertion order).
+        2. The first healthy non-primary (insertion order).
+        3. ``None`` when nothing is healthy.
+
+        This MVP picks the first match within each tier; it does NOT
+        load-balance across multiple healthy replicas of the same tier.
+        Round-robin and weighted distribution are deferred until a real
+        N-way active deployment exists.
+        """
+        healthy_primaries = [r for r in self._replicas.values() if r.primary and r.is_healthy]
+        if healthy_primaries:
+            return healthy_primaries[0]
+        healthy = [r for r in self._replicas.values() if r.is_healthy]
+        if healthy:
+            return healthy[0]
+        return None
+
+    async def record_success(
+        self,
+        replica: Replica,
+        *,
+        recovery_threshold: int,
+        latency_ms: float,
+    ) -> None:
+        async with self._lock:
+            h = replica.health
+            h.last_ok_at = time.time()
+            h.last_latency_ms = latency_ms
+            h.consecutive_failures = 0
+            h.consecutive_successes += 1
+            # State machine: UNKNOWN -> HEALTHY is immediate; only the
+            # UNHEALTHY -> HEALTHY transition is gated by recovery_threshold.
+            if h.status is ReplicaStatus.UNKNOWN:
+                logger.info("router: replica %s initial probe ok, marking HEALTHY", replica.url)
+                h.status = ReplicaStatus.HEALTHY
+                h.consecutive_successes = 0
+            elif (h.status is ReplicaStatus.UNHEALTHY and h.consecutive_successes >= recovery_threshold):
+                logger.info("router: replica %s recovered to HEALTHY after %d successes", replica.url,
+                            h.consecutive_successes)
+                h.status = ReplicaStatus.HEALTHY
+                h.consecutive_successes = 0
+
+    async def record_failure(
+        self,
+        replica: Replica,
+        *,
+        failure_threshold: int,
+        reason: str,
+    ) -> None:
+        async with self._lock:
+            h = replica.health
+            h.last_failure_at = time.time()
+            h.consecutive_successes = 0
+            h.consecutive_failures += 1
+            if (h.status is not ReplicaStatus.UNHEALTHY and h.consecutive_failures >= failure_threshold):
+                logger.warning("router: replica %s marked UNHEALTHY after %d failures: %s", replica.url,
+                               h.consecutive_failures, reason)
+                h.status = ReplicaStatus.UNHEALTHY
+
+
+async def run_health_check_loop(
+    registry: ReplicaRegistry,
+    config: RouterConfig,
+    *,
+    stop_event: asyncio.Event,
+    http_get: HttpProbe | None = None,
+) -> None:
+    """Poll all replicas' health endpoints in parallel on a fixed interval.
+
+    ``http_get`` is pluggable so unit tests can inject a deterministic
+    probe without hitting the network. The default builds a single
+    ``httpx.AsyncClient`` shared across the loop's lifetime so the
+    common case (steady polling against a stable replica set) reuses
+    TCP/TLS connections instead of paying handshake cost per probe.
+
+    Probes within one polling cycle run concurrently via ``asyncio.gather``
+    so a slow replica doesn't push the cycle past
+    ``health_check_interval_seconds``.
+    """
+    if http_get is not None:
+        await _run_loop(registry, config, stop_event, http_get)
+        return
+    async with _build_default_probe(config) as probe:
+        await _run_loop(registry, config, stop_event, probe)
+
+
+async def _run_loop(
+    registry: ReplicaRegistry,
+    config: RouterConfig,
+    stop_event: asyncio.Event,
+    http_get: Callable[..., Awaitable[tuple[float, str | None]]],
+) -> None:
+    while not stop_event.is_set():
+        replicas = registry.all()
+        results = await asyncio.gather(
+            *[
+                http_get(replica.url + config.health_check_path, timeout=config.health_check_timeout_seconds)
+                for replica in replicas
+            ],
+            return_exceptions=True,
+        )
+        for replica, result in zip(replicas, results, strict=True):
+            if isinstance(result, BaseException):
+                await registry.record_failure(
+                    replica,
+                    failure_threshold=config.failure_threshold,
+                    reason=f"{type(result).__name__}: {result}",
+                )
+                continue
+            status_ms, error = result
+            if error is None:
+                await registry.record_success(
+                    replica,
+                    recovery_threshold=config.recovery_threshold,
+                    latency_ms=status_ms,
+                )
+            else:
+                await registry.record_failure(
+                    replica,
+                    failure_threshold=config.failure_threshold,
+                    reason=error,
+                )
+        try:
+            await asyncio.wait_for(
+                stop_event.wait(),
+                timeout=config.health_check_interval_seconds,
+            )
+        except asyncio.TimeoutError:
+            continue
+
+
+@contextlib.asynccontextmanager
+async def _build_default_probe(
+    config: RouterConfig, ) -> AsyncIterator[Callable[..., Awaitable[tuple[float, str | None]]]]:
+    try:
+        import httpx
+    except ImportError as exc:  # pragma: no cover - optional extra
+        raise RuntimeError("router health checks require httpx; install with "
+                           "`pip install fastvideo[streaming]` or `pip install httpx`") from exc
+
+    async with httpx.AsyncClient(timeout=config.health_check_timeout_seconds) as client:
+
+        async def probe(url: str, *, timeout: float) -> tuple[float, str | None]:
+            start = time.perf_counter()
+            try:
+                response = await client.get(url, timeout=timeout)
+            except Exception as exc:
+                return 0.0, f"{type(exc).__name__}: {exc}"
+            latency_ms = (time.perf_counter() - start) * 1000.0
+            if response.status_code >= 400:
+                return latency_ms, f"HTTP {response.status_code}"
+            return latency_ms, None
+
+        yield probe
+
+
+__all__ = [
+    "HttpProbe",
+    "Replica",
+    "ReplicaHealth",
+    "ReplicaRegistry",
+    "ReplicaStatus",
+    "run_health_check_loop",
+]

fastvideo/entrypoints/streaming/server.py

@@ -51,6 +51,11 @@ from fastvideo.entrypoints.streaming.session import (
 )
 from fastvideo.entrypoints.streaming.session_init_image import (
     persist_session_init_image, )
+from fastvideo.entrypoints.streaming.gpu_pool import (
+    GpuPool,
+    InProcessGpuPool,
+    PoolAcquireTimeout,
+)
 from fastvideo.entrypoints.streaming.session_store import (
     InMemorySessionStore,
     SessionStore,
@@ -75,25 +80,37 @@ class _GeneratorProto(Protocol):
 @dataclass
 class ServerState:
     serve_config: ServeConfig
-    generator: _GeneratorProto
+    pool: GpuPool
     sessions: SessionManager
     session_store: SessionStore
 
 
 def build_app(
     serve_config: ServeConfig,
-    generator: _GeneratorProto,
+    generator: _GeneratorProto | None = None,
     *,
+    pool: GpuPool | None = None,
     session_store: SessionStore | None = None,
 ) -> FastAPI:
     """Build the FastAPI app used by :func:`run_server`.
 
     Exposed so tests can drive the WebSocket endpoint in-process via
     ``starlette.testclient.TestClient(app).websocket_connect(...)``.
+
+    Exactly one of ``generator`` (backed by :class:`InProcessGpuPool`)
+    or ``pool`` (for the subprocess-backed production shape) must be
+    given.
     """
     if serve_config.streaming is None:
         raise ValueError("ServeConfig.streaming must be set to launch the streaming "
                          "server; got None. Add a `streaming:` block to your serve config.")
+    if (generator is None) == (pool is None):
+        raise ValueError("build_app requires exactly one of `generator` or `pool`")
+
+    store = session_store or InMemorySessionStore()
+    if pool is None:
+        assert generator is not None
+        pool = InProcessGpuPool(generator, session_store=store)
 
     sessions = SessionManager(
         segment_cap=serve_config.streaming.generation_segment_cap,
@@ -101,9 +118,9 @@ def build_app(
     )
     state = ServerState(
         serve_config=serve_config,
-        generator=generator,
+        pool=pool,
         sessions=sessions,
-        session_store=session_store or InMemorySessionStore(),
+        session_store=store,
     )
 
     app = FastAPI(title="FastVideo Streaming")
@@ -135,6 +152,8 @@ def build_app(
             with contextlib.suppress(InvalidSessionTransition):
                 session.transition(SessionState.ERROR)
         finally:
+            with contextlib.suppress(Exception):
+                await state.pool.release(session.id)
             _cleanup_session(session, state)
 
     app.state.server_state = state
@@ -178,10 +197,22 @@ async def _handle_session(
     await _apply_session_init(session, init, state)
     await _send_json(websocket, QueueStatus(position=0, queue_depth=0))
     session.transition(SessionState.GPU_BINDING)
-    await _send_json(websocket, GpuAssigned(
-        gpu_id=0,
-        session_timeout=state.sessions.session_timeout_seconds,
-    ))
+    try:
+        assignment = await state.pool.acquire(
+            session.id,
+            timeout=float(state.sessions.session_timeout_seconds),
+        )
+    except PoolAcquireTimeout as exc:
+        await _send_error(websocket, "gpu_unavailable", str(exc), retryable=True)
+        with contextlib.suppress(InvalidSessionTransition):
+            session.transition(SessionState.TIMEOUT)
+        return
+    session.gpu_id = assignment.gpu_id
+    await _send_json(websocket,
+                     GpuAssigned(
+                         gpu_id=assignment.gpu_id,
+                         session_timeout=state.sessions.session_timeout_seconds,
+                     ))
     session.transition(SessionState.ACTIVE)
     await _send_json(websocket, _build_stream_start(session, state))
 
@@ -327,15 +358,13 @@ async def _run_segment(
         ))
 
     start = time.perf_counter()
-    loop = asyncio.get_running_loop()
-    # TODO: executor-wrapped generate() cannot be cancelled, so a
-    # client disconnect mid-segment leaves the GPU work running to
-    # completion. Real cancellation needs the generate_async API.
+    # TODO: pool.run() runs to completion even if the client disconnects
+    # mid-segment. Real cancellation needs the generate_async API.
     try:
-        result = await loop.run_in_executor(None, state.generator.generate, request)
+        result = await state.pool.run(session.id, request)
     except Exception as exc:
-        logger.exception("session %s: generator failed", session.id[:8])
-        await _send_error(websocket, "worker_failed", f"generator.generate failed: {exc}", retryable=True)
+        logger.exception("session %s: pool.run failed", session.id[:8])
+        await _send_error(websocket, "worker_failed", f"pool.run failed: {exc}", retryable=True)
         with contextlib.suppress(InvalidSessionTransition):
             session.transition(SessionState.ERROR)
         return

fastvideo/entrypoints/streaming/session_logger.py

@@ -0,0 +1,113 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Per-session JSONL event logger.
+
+Each session gets its own JSONL file under the configured log root so
+post-hoc analytics (enhancer latency, GPU assignment, segment timings)
+can be recovered without a tracing backend. The internal UI uses this
+format; keeping the same shape makes log tooling portable.
+"""
+from __future__ import annotations
+
+import contextlib
+import json
+import os
+import re
+import threading
+import time
+from dataclasses import dataclass, field
+from typing import Any, TextIO
+
+_FILENAME_SANITIZE_RE = re.compile(r"[^A-Za-z0-9._-]")
+
+
+@dataclass
+class SessionLogEvent:
+    """One line in the session JSONL file."""
+
+    session_id: str
+    event: str
+    payload: dict[str, Any] = field(default_factory=dict)
+    ts: float = field(default_factory=time.time)
+
+
+class SessionLogger:
+    """Append-only JSONL logger keyed by session id.
+
+    Thread-safe; the server may be writing from multiple asyncio tasks
+    (fMP4 encoder thread + control-frame handler) for the same session.
+    """
+
+    def __init__(self, log_dir: str | None) -> None:
+        self._log_dir = log_dir
+        self._files: dict[str, TextIO] = {}
+        self._locks: dict[str, threading.Lock] = {}
+        self._registry_lock = threading.Lock()
+        self._ensure_dir()
+
+    def log(self, event: SessionLogEvent) -> None:
+        if self._log_dir is None:
+            return
+        opened = self._get_file(event.session_id)
+        if opened is None:
+            return
+        handle, lock = opened
+        line = json.dumps({
+            "session_id": event.session_id,
+            "event": event.event,
+            "ts": event.ts,
+            "payload": event.payload,
+        })
+        with lock, contextlib.suppress(ValueError):
+            handle.write(line + "\n")
+            handle.flush()
+
+    def close(self, session_id: str) -> None:
+        with self._registry_lock:
+            handle = self._files.pop(session_id, None)
+            lock = self._locks.pop(session_id, None)
+        if handle is None or lock is None:
+            return
+        with lock, contextlib.suppress(Exception):
+            handle.close()
+
+    def close_all(self) -> None:
+        with self._registry_lock:
+            sids = list(self._files)
+        for sid in sids:
+            self.close(sid)
+
+    def _ensure_dir(self) -> None:
+        if self._log_dir is None:
+            return
+        os.makedirs(self._log_dir, exist_ok=True)
+
+    def _get_file(self, session_id: str) -> tuple[TextIO, threading.Lock] | None:
+        if self._log_dir is None:
+            return None
+        with self._registry_lock:
+            handle = self._files.get(session_id)
+            lock = self._locks.get(session_id)
+            if handle is not None and lock is not None:
+                return handle, lock
+            # Defense-in-depth: session_id is server-generated UUID today,
+            # but sanitize against path traversal in case future code paths
+            # allow client-supplied ids.
+            safe_id = _FILENAME_SANITIZE_RE.sub("_", session_id) or "unknown"
+            path = os.path.join(
+                self._log_dir,
+                f"session-{safe_id}.jsonl",
+            )
+            try:
+                handle = open(path, "a", encoding="utf-8")  # noqa: SIM115
+            except OSError:
+                return None
+            lock = threading.Lock()
+            self._files[session_id] = handle
+            self._locks[session_id] = lock
+            return handle, lock
+
+
+__all__ = [
+    "SessionLogEvent",
+    "SessionLogger",
+]

fastvideo/entrypoints/streaming/worker.py

@@ -0,0 +1,133 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Per-GPU worker subprocess entry for :class:`SubprocessGpuPool`.
+
+The pool manages binding, lifecycle, and message dispatch in the parent
+process. The worker constructs its :class:`VideoGenerator` from a typed
+:class:`GeneratorConfig`, runs the two-segment warmup so both
+initial-segment and continuation-branch compile graphs are hot, and
+then loops on the job queue.
+"""
+from __future__ import annotations
+
+import multiprocessing as mp
+import queue
+from typing import Any
+
+from fastvideo.api.schema import (
+    GeneratorConfig,
+    GenerationRequest,
+    InputConfig,
+    OutputConfig,
+    SamplingConfig,
+    WarmupConfig,
+)
+from fastvideo.logger import init_logger
+
+logger = init_logger(__name__)
+
+# Synthetic warmup dimensions: small enough to keep boot fast, big enough
+# to exercise the real shape-dependent compile paths. Keep in sync with
+# WarmupConfig if these become user-tunable.
+_WARMUP_NUM_FRAMES = 8
+_WARMUP_HEIGHT = 256
+_WARMUP_WIDTH = 256
+_WARMUP_NUM_INFERENCE_STEPS = 1
+
+
+def worker_main(
+    *,
+    gpu_id: int,
+    worker_id: str,
+    generator_config: GeneratorConfig,
+    warmup_config: WarmupConfig,
+    job_queue: mp.Queue,
+    result_queue: mp.Queue,
+    shutdown_event: Any,
+) -> None:  # pragma: no cover - exercised via integration only
+    """Per-worker subprocess entry.
+
+    Runs inside the child spawned by ``SubprocessGpuPool``. Blocking
+    ``VideoGenerator`` construction + generation happens here, not in
+    the parent's event loop.
+    """
+    import os
+
+    os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu_id)
+    try:
+        from fastvideo import VideoGenerator
+
+        generator = VideoGenerator.from_pretrained(config=generator_config)
+        if warmup_config.enabled:
+            _warmup_worker(generator, warmup_config)
+        result_queue.put({"kind": "ready", "worker_id": worker_id})
+    except Exception as exc:
+        result_queue.put({"kind": "error", "error": repr(exc)})
+        return
+
+    while not shutdown_event.is_set():
+        try:
+            item = job_queue.get(timeout=0.5)
+        except queue.Empty:
+            continue
+        if item is None:
+            break
+        job_id = item["job_id"]
+        request = item["request"]
+        try:
+            result = generator.generate(request)
+            result_queue.put({
+                "kind": "result",
+                "job_id": job_id,
+                "result": result,
+            })
+        except Exception as exc:
+            result_queue.put({
+                "kind": "error",
+                "job_id": job_id,
+                "error": repr(exc),
+            })
+
+
+def _warmup_worker(
+    generator: Any,
+    warmup_config: WarmupConfig,
+) -> None:
+    """Run two synthetic generations so both compile branches are primed.
+
+    Segment 1 is a fresh start (no continuation state) and exercises
+    the initial-segment graph. Segment 2 feeds segment 1's continuation
+    state back in so the conditioning branch is also compiled before
+    the first user request lands.
+    """
+    sampling = SamplingConfig(
+        num_frames=_WARMUP_NUM_FRAMES,
+        height=_WARMUP_HEIGHT,
+        width=_WARMUP_WIDTH,
+        num_inference_steps=_WARMUP_NUM_INFERENCE_STEPS,
+    )
+    seg1 = GenerationRequest(
+        prompt=warmup_config.prompt,
+        sampling=sampling,
+        inputs=InputConfig(),
+        output=OutputConfig(save_video=False, return_frames=False, return_state=True),
+    )
+    seg1_result = generator.generate(seg1)
+
+    seg2 = GenerationRequest(
+        prompt=warmup_config.prompt,
+        sampling=sampling,
+        inputs=InputConfig(),
+        output=OutputConfig(save_video=False, return_frames=False),
+        state=_extract_continuation_state(seg1_result),
+    )
+    generator.generate(seg2)
+
+
+def _extract_continuation_state(result: Any) -> Any:
+    state = getattr(result, "state", None)
+    if state is None and isinstance(result, dict):
+        state = result.get("state")
+    return state
+
+
+__all__ = ["worker_main"]

fastvideo/entrypoints/video_generator.py

@@ -65,6 +65,7 @@ _FROM_PRETRAINED_CONVENIENCE_KWARGS = frozenset({
     "pin_cpu_memory",
     "enable_torch_compile",
     "torch_compile_kwargs",
+    "output_type",
 })
 
 
@@ -601,10 +602,20 @@ class VideoGenerator:
         thread = threading.Thread(target=execute_forward_thread)
         thread.start()
         latent_batch_size = _infer_latent_batch_size(batch)
-        samples = torch.empty(
-            (latent_batch_size, 3, sampling_param.num_frames, sampling_param.height, sampling_param.width),
-            device='cpu',
-            pin_memory=fastvideo_args.pin_cpu_memory)
+        # When ``output_type == "latent"`` the forward output has latent
+        # shape (e.g. ``[B, C_latent, T_latent, H_latent, W_latent]``)
+        # rather than the pre-allocation's pixel shape. Skip the pinned
+        # ~50 MB buffer entirely; we always fall through to the
+        # ``samples = output_batch.output.cpu()`` branch below in that
+        # mode. ``skip_pixel_prealloc`` also gates the slow-path warning.
+        skip_pixel_prealloc = fastvideo_args.output_type == "latent"
+        if skip_pixel_prealloc:
+            samples = torch.empty(0, device='cpu')
+        else:
+            samples = torch.empty(
+                (latent_batch_size, 3, sampling_param.num_frames, sampling_param.height, sampling_param.width),
+                device='cpu',
+                pin_memory=fastvideo_args.pin_cpu_memory)
         thread.join()
 
         if thread_error["error"] is not None:
@@ -619,29 +630,44 @@ class VideoGenerator:
         if output_batch.output.shape == samples.shape:
             samples.copy_(output_batch.output)
         else:
-            logger.warning("Output shape %s does not match expected shape %s; use slow path", output_batch.output.shape,
-                           samples.shape)
+            if not skip_pixel_prealloc:
+                logger.warning("Output shape %s does not match expected shape %s; use slow path",
+                               output_batch.output.shape, samples.shape)
             samples = output_batch.output.cpu()
         logging_info = output_batch.logging_info
 
         gen_time = time.perf_counter() - start_time
         logger.info("Generated successfully in %.2f seconds", gen_time)
 
-        # Process outputs (skip the make_grid loop for audio-only, where
-        # `samples` is a 1×3×1×8×8 placeholder no caller will use).
+        # Three mutually-exclusive output modes determine whether (a) we
+        # build an RGB frame buffer and (b) what file we write to disk:
+        #
+        #   1. `output_type == "latent"` — VAE is bypassed in DecodingStage
+        #      and `samples` holds raw latents (arbitrary channel count).
+        #      The RGB grid / uint8 / mp4 / png pipeline below cannot
+        #      consume those, so we skip it entirely and let callers work
+        #      with the latent tensor directly via `result["samples"]`.
+        #   2. Audio-only workload — `samples` is a 1×3×1×8×8 placeholder
+        #      no caller will use; skip the grid loop and save a `.wav`.
+        #   3. Pixel video / image — the historical happy path.
+        is_latent_output = fastvideo_args.output_type == "latent"
         audio_only = bool(output_batch.extra.get("audio_only"))
-        frames: list[np.ndarray] = []
-        if not audio_only:
+
+        frames: list[np.ndarray] | None
+        if is_latent_output or audio_only:
+            frames = None if is_latent_output else []
+        else:
             videos = rearrange(samples, "b c t h w -> t b c h w")
+            frames = []
             for x in videos:
                 x = torchvision.utils.make_grid(x, nrow=6)
                 x = x.permute(1, 2, 0).squeeze(-1)
                 x = (x * 255).to(torch.uint8)
                 frames.append(x.cpu().numpy())
 
-        # Save output if requested
-        if batch.save_video:
-            if output_batch.extra.get("audio_only"):
+        save_to_disk = batch.save_video and not is_latent_output
+        if save_to_disk:
+            if audio_only:
                 # Audio-only workload: write a standalone .wav rather than
                 # muxing the audio into a placeholder mp4 (which forces
                 # ffmpeg to round 8x8 placeholder frames up to 16x16).
@@ -654,9 +680,11 @@ class VideoGenerator:
                 logger.info("Saved audio to %s", output_path)
             elif self._is_image_workload():
                 # Image workloads (t2i, i2i, …): save the first frame as PNG.
+                assert frames is not None  # implied by save_to_disk and not audio_only
                 imageio.imwrite(output_path, frames[0])
                 logger.info("Saved image to %s", output_path)
             else:
+                assert frames is not None  # implied by save_to_disk and not audio_only
                 imageio.mimsave(output_path, frames, fps=batch.fps, format="mp4")
                 logger.info("Saved video to %s", output_path)
                 audio = output_batch.extra.get("audio")
@@ -680,7 +708,7 @@ class VideoGenerator:
             "trajectory": output_batch.trajectory_latents,
             "trajectory_timesteps": output_batch.trajectory_timesteps,
             "trajectory_decoded": output_batch.trajectory_decoded,
-            "video_path": output_path if batch.save_video else None,
+            "video_path": output_path if save_to_disk else None,
             "peak_memory_mb": output_batch.extra.get("peak_memory_mb"),
         }
 

fastvideo/pipelines/basic/stable_audio/stages/decoding.py

@@ -33,6 +33,15 @@ class StableAudioDecodingStage(PipelineStage):
         pc = fastvideo_args.pipeline_config
         latents = batch.latents
 
+        # Latent regression path: hand back the un-decoded denoised latent
+        # so the LatentSimilarityUtils harness can compare on pre-VAE
+        # numerics. Mirrors the bypass in `pipelines/stages/decoding.py`
+        # used by the video DiTs. Skips the `.to(device)` VAE move so we
+        # don't pay decoder load cost on this shortcut path.
+        if fastvideo_args.output_type == "latent":
+            batch.output = latents.detach().cpu()
+            return batch
+
         # VAE may be CPU-parked under `vae_cpu_offload=True`.
         from fastvideo.distributed.parallel_state import get_local_torch_device
         self.vae = self.vae.to(get_local_torch_device())

fastvideo/tests/entrypoints/streaming/test_auxiliaries.py

@@ -0,0 +1,261 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Tests for PR 7.8 streaming auxiliaries.
+
+Covers:
+
+* PromptSafetyFilter gracefully disables when fastText isn't installed
+* SafetyResult semantics (allow, block, unavailable)
+* RewriteOptions + _split_response parsing behavior
+* SessionLogger JSONL append semantics + close lifecycle
+* MockServer builds an app that drives the WS protocol end-to-end
+"""
+from __future__ import annotations
+
+import json
+import sys
+import types
+from pathlib import Path
+from typing import Any
+
+import pytest
+
+from fastvideo.entrypoints.streaming.prompt.safety import (
+    PromptSafetyFilter,
+    SafetyDecision,
+    first_blocked,
+)
+from fastvideo.entrypoints.streaming.prompt.rewrite import (
+    RewriteOptions,
+    _split_response,
+    build_rewrite,
+)
+from fastvideo.entrypoints.streaming.session_logger import (
+    SessionLogEvent,
+    SessionLogger,
+)
+
+
+# ----------------------------------------------------------------------
+# Safety
+# ----------------------------------------------------------------------
+
+
+class TestPromptSafetyFilter:
+
+    def test_disabled_by_default_when_no_path(self):
+        f = PromptSafetyFilter(classifier_path=None)
+        assert f.enabled is False
+        result = f.classify("hi")
+        assert result.decision is SafetyDecision.UNAVAILABLE
+
+    def test_disabled_when_enabled_false(self):
+        f = PromptSafetyFilter(classifier_path="/tmp/m.bin", enabled=False)
+        assert f.enabled is False
+
+    def test_unavailable_when_fasttext_missing(self, monkeypatch):
+        # Force `import fasttext` inside _ensure_loaded to fail.
+        monkeypatch.setitem(sys.modules, "fasttext", None)
+        f = PromptSafetyFilter(classifier_path="/tmp/m.bin", enabled=True)
+        result = f.classify("hi")
+        assert result.decision is SafetyDecision.UNAVAILABLE
+
+    def test_block_when_classifier_flags_unsafe(self, monkeypatch, tmp_path):
+        fake_model = types.SimpleNamespace(
+            predict=lambda text, k=1: (["__label__unsafe"], [0.95]))
+        stub = types.SimpleNamespace(load_model=lambda _p: fake_model)
+        monkeypatch.setitem(sys.modules, "fasttext", stub)
+        model_path = str(tmp_path / "m.bin")
+        Path(model_path).write_text("")
+        f = PromptSafetyFilter(classifier_path=model_path, enabled=True)
+        result = f.classify("please")
+        assert result.decision is SafetyDecision.BLOCK
+        assert result.label == "unsafe"
+        assert result.score == pytest.approx(0.95)
+
+    def test_allow_when_classifier_flags_safe(self, monkeypatch, tmp_path):
+        fake_model = types.SimpleNamespace(
+            predict=lambda text, k=1: (["__label__safe"], [0.99]))
+        stub = types.SimpleNamespace(load_model=lambda _p: fake_model)
+        monkeypatch.setitem(sys.modules, "fasttext", stub)
+        f = PromptSafetyFilter(classifier_path="ignored", enabled=True)
+        result = f.classify("hello")
+        assert result.decision is SafetyDecision.ALLOW
+
+    def test_below_threshold_allows_even_if_unsafe_label(self, monkeypatch):
+        fake_model = types.SimpleNamespace(
+            predict=lambda text, k=1: (["__label__unsafe"], [0.3]))
+        stub = types.SimpleNamespace(load_model=lambda _p: fake_model)
+        monkeypatch.setitem(sys.modules, "fasttext", stub)
+        f = PromptSafetyFilter(
+            classifier_path="m", enabled=True, block_threshold=0.5)
+        assert f.classify("x").decision is SafetyDecision.ALLOW
+
+    def test_first_blocked_returns_first_hit(self, monkeypatch):
+        responses = iter([
+            (["__label__safe"], [0.9]),
+            (["__label__unsafe"], [0.9]),
+            (["__label__safe"], [0.9]),
+        ])
+        fake_model = types.SimpleNamespace(
+            predict=lambda text, k=1: next(responses))
+        stub = types.SimpleNamespace(load_model=lambda _p: fake_model)
+        monkeypatch.setitem(sys.modules, "fasttext", stub)
+        f = PromptSafetyFilter(classifier_path="m", enabled=True)
+        blocked = first_blocked(f, ["ok", "bad", "also ok"])
+        assert blocked is not None
+        assert blocked.prompt == "bad"
+
+
+# ----------------------------------------------------------------------
+# Rewrite
+# ----------------------------------------------------------------------
+
+
+class TestRewriteSplit:
+
+    def test_plain_lines(self):
+        assert _split_response("one\ntwo\nthree", limit=3) == [
+            "one", "two", "three",
+        ]
+
+    def test_numbered_list(self):
+        assert _split_response("1. first\n2. second", limit=3) == [
+            "first", "second",
+        ]
+
+    def test_bulleted_list(self):
+        assert _split_response("- one\n* two\n• three", limit=3) == [
+            "one", "two", "three",
+        ]
+
+    def test_respects_limit(self):
+        assert _split_response("a\nb\nc\nd", limit=2) == ["a", "b"]
+
+    def test_limit_min_one(self):
+        assert _split_response("only one", limit=0) == ["only one"]
+
+
+class _StubEnhancer:
+
+    async def rewrite(self, seed):
+        from fastvideo.entrypoints.streaming.prompt.providers.base import LLMResponse
+
+        return LLMResponse(
+            content="1. alpha\n2. beta\n3. gamma",
+            provider="stub",
+            model="m",
+            latency_ms=1.0,
+        )
+
+
+class TestBuildRewrite:
+
+    def test_empty_seed_rejected(self):
+        import asyncio
+
+        with pytest.raises(ValueError):
+            asyncio.run(build_rewrite(_StubEnhancer(), "   "))
+
+    def test_returns_limited_alternatives(self):
+        import asyncio
+
+        result = asyncio.run(build_rewrite(
+            _StubEnhancer(), "seed",
+            options=RewriteOptions(count=2)))
+        assert result.seed_prompt == "seed"
+        assert result.alternatives == ["alpha", "beta"]
+        assert result.provider == "stub"
+
+
+# ----------------------------------------------------------------------
+# Session logger
+# ----------------------------------------------------------------------
+
+
+class TestSessionLogger:
+
+    def test_no_log_dir_is_noop(self):
+        logger = SessionLogger(None)
+        logger.log(SessionLogEvent(session_id="s", event="x"))  # no raise
+
+    def test_appends_jsonl(self, tmp_path):
+        logger = SessionLogger(str(tmp_path))
+        logger.log(SessionLogEvent(
+            session_id="s1",
+            event="start",
+            payload={"preset": "ltx2"},
+            ts=1.0,
+        ))
+        logger.log(SessionLogEvent(
+            session_id="s1",
+            event="segment",
+            payload={"idx": 0},
+            ts=2.0,
+        ))
+        logger.close("s1")
+        path = tmp_path / "session-s1.jsonl"
+        lines = path.read_text().splitlines()
+        assert len(lines) == 2
+        first = json.loads(lines[0])
+        assert first["event"] == "start"
+        assert first["payload"]["preset"] == "ltx2"
+
+    def test_separate_files_per_session(self, tmp_path):
+        logger = SessionLogger(str(tmp_path))
+        logger.log(SessionLogEvent(session_id="a", event="e"))
+        logger.log(SessionLogEvent(session_id="b", event="e"))
+        logger.close_all()
+        assert (tmp_path / "session-a.jsonl").exists()
+        assert (tmp_path / "session-b.jsonl").exists()
+
+
+# ----------------------------------------------------------------------
+# Mock server
+# ----------------------------------------------------------------------
+
+
+class TestMockServer:
+
+    def test_build_mock_app_returns_fastapi(self):
+        from fastapi import FastAPI
+
+        from fastvideo.entrypoints.streaming.mock_server import build_mock_app
+
+        app = build_mock_app()
+        assert isinstance(app, FastAPI)
+
+    def test_mock_generator_produces_frames(self):
+        from fastvideo.api.schema import GenerationRequest, SamplingConfig
+        from fastvideo.entrypoints.streaming.mock_server import MockGenerator
+
+        gen = MockGenerator()
+        result = gen.generate(GenerationRequest(
+            prompt="x",
+            sampling=SamplingConfig(
+                num_frames=3, height=32, width=32, num_inference_steps=1),
+        ))
+        assert len(result["frames"]) == 3
+        assert result["frames"][0].shape == (32, 32, 3)
+        assert result["state"].kind == "ltx2.v1"
+
+    def test_mock_app_health_endpoint(self):
+        from starlette.testclient import TestClient
+
+        from fastvideo.entrypoints.streaming.mock_server import build_mock_app
+
+        app = build_mock_app()
+        client = TestClient(app)
+        assert client.get("/health").json()["status"] == "ok"
+
+    def test_mock_app_ws_handshake(self):
+        from starlette.testclient import TestClient
+
+        from fastvideo.entrypoints.streaming.mock_server import build_mock_app
+
+        app = build_mock_app()
+        client = TestClient(app)
+        with client.websocket_connect("/v1/stream") as ws:
+            ws.send_json({"type": "session_init_v2"})
+            assert ws.receive_json()["type"] == "queue_status"
+            assert ws.receive_json()["type"] == "gpu_assigned"
+            assert ws.receive_json()["type"] == "ltx2_stream_start"

fastvideo/tests/entrypoints/streaming/test_gpu_pool.py

@@ -0,0 +1,463 @@
+# SPDX-License-Identifier: Apache-2.0
+"""GPU pool tests.
+
+InProcessGpuPool is exercised end-to-end. SubprocessGpuPool is driven
+with an injected ``worker_factory`` that stands up a fake worker inside
+a thread (not a subprocess) so the test suite stays CPU-only.
+"""
+from __future__ import annotations
+
+import asyncio
+import multiprocessing as mp
+import queue
+import threading
+import time
+from dataclasses import dataclass
+from typing import Any
+
+import pytest
+
+from fastvideo.api.schema import (
+    GeneratorConfig,
+    GenerationRequest,
+    GpuPoolConfig,
+    WarmupConfig,
+)
+from fastvideo.entrypoints.streaming.gpu_pool import (
+    GpuPool,
+    InProcessGpuPool,
+    PoolAcquireTimeout,
+    SubprocessGpuPool,
+    _WorkerHandle,
+)
+
+
+# ----------------------------------------------------------------------
+# In-process pool
+# ----------------------------------------------------------------------
+
+
+@dataclass
+class _MockGenerator:
+    sleep_s: float = 0.0
+
+    def generate(self, request: GenerationRequest) -> dict[str, Any]:
+        if self.sleep_s:
+            time.sleep(self.sleep_s)
+        return {
+            "frames": [],
+            "prompt_echo": request.prompt,
+        }
+
+
+class TestInProcessGpuPool:
+
+    def test_is_gpu_pool(self):
+        assert isinstance(
+            InProcessGpuPool(_MockGenerator()), GpuPool)
+
+    def test_acquire_returns_deterministic_assignment(self):
+        pool = InProcessGpuPool(_MockGenerator(), gpu_id=7)
+
+        async def run():
+            a = await pool.acquire("sess-a")
+            return a
+
+        a = asyncio.run(run())
+        assert a.gpu_id == 7
+        assert a.worker_id.startswith("inproc-")
+
+    def test_acquire_is_sticky_across_calls(self):
+        pool = InProcessGpuPool(_MockGenerator())
+
+        async def run():
+            a = await pool.acquire("sess-a")
+            b = await pool.acquire("sess-a")
+            return a, b
+
+        a, b = asyncio.run(run())
+        assert a == b
+
+    def test_run_without_acquire_raises(self):
+        pool = InProcessGpuPool(_MockGenerator())
+
+        async def run():
+            with pytest.raises(RuntimeError):
+                await pool.run(
+                    "sess-a", GenerationRequest(prompt="hi"))
+
+        asyncio.run(run())
+
+    def test_run_returns_generator_output(self):
+        pool = InProcessGpuPool(_MockGenerator())
+
+        async def run():
+            await pool.acquire("sess-a")
+            return await pool.run(
+                "sess-a", GenerationRequest(prompt="hi"))
+
+        result = asyncio.run(run())
+        assert result["prompt_echo"] == "hi"
+
+    def test_release_frees_binding(self):
+        pool = InProcessGpuPool(_MockGenerator())
+
+        async def run():
+            await pool.acquire("sess-a")
+            await pool.release("sess-a")
+            with pytest.raises(RuntimeError):
+                await pool.run(
+                    "sess-a", GenerationRequest(prompt="hi"))
+
+        asyncio.run(run())
+
+    def test_health_reports_active_sessions(self):
+        pool = InProcessGpuPool(_MockGenerator())
+
+        async def run():
+            await pool.acquire("sess-a")
+            health = pool.health()
+            assert health.total_workers == 1
+            assert health.active_sessions == 1
+            assert health.available_workers == 0
+
+        asyncio.run(run())
+
+
+# ----------------------------------------------------------------------
+# Subprocess pool (driven by a thread-backed fake worker factory)
+# ----------------------------------------------------------------------
+
+
+class _ThreadWorker:
+    """Stand-in for a subprocess worker.
+
+    Runs a Python thread that pulls jobs from ``job_queue`` and invokes
+    a supplied mock generator. The control flow matches
+    :func:`worker_main` exactly (ready + result dict shapes) so the
+    parent-side pool under test exercises the same code paths.
+    """
+
+    def __init__(
+        self,
+        generator: _MockGenerator,
+        *,
+        job_queue: mp.Queue,
+        result_queue: mp.Queue,
+        shutdown_event: threading.Event,
+        warmup_ms: float = 0.0,
+    ) -> None:
+        self._generator = generator
+        self._job_queue = job_queue
+        self._result_queue = result_queue
+        self._shutdown_event = shutdown_event
+        self._warmup_ms = warmup_ms
+        self._thread = threading.Thread(target=self._run, daemon=True)
+
+    def start(self) -> None:
+        self._thread.start()
+
+    def join(self, timeout: float | None = None) -> None:
+        self._thread.join(timeout)
+
+    def _run(self) -> None:
+        if self._warmup_ms:
+            time.sleep(self._warmup_ms / 1000.0)
+        self._result_queue.put({"kind": "ready"})
+        while not self._shutdown_event.is_set():
+            try:
+                item = self._job_queue.get(timeout=0.1)
+            except queue.Empty:
+                continue
+            if item is None:
+                return
+            try:
+                result = self._generator.generate(item["request"])
+                self._result_queue.put({
+                    "kind": "result",
+                    "job_id": item["job_id"],
+                    "result": result,
+                })
+            except Exception as exc:  # pragma: no cover - defensive
+                self._result_queue.put({
+                    "kind": "error",
+                    "job_id": item["job_id"],
+                    "error": repr(exc),
+                })
+
+
+def _thread_worker_factory(generator_builder):
+    """Return a WorkerFactory that uses thread workers instead of procs."""
+
+    def factory(
+        *,
+        gpu_id: int,
+        generator_config: GeneratorConfig,
+        warmup_config: WarmupConfig,
+    ) -> _WorkerHandle:
+        ctx = mp.get_context("spawn")
+        job_queue: mp.Queue = ctx.Queue()
+        result_queue: mp.Queue = ctx.Queue()
+        shutdown_event = threading.Event()
+        ready = threading.Event()
+        boot_ok = threading.Event()
+        mp_shutdown = ctx.Event()
+
+        generator = generator_builder(gpu_id)
+        worker = _ThreadWorker(
+            generator,
+            job_queue=job_queue,
+            result_queue=result_queue,
+            shutdown_event=shutdown_event,
+        )
+        worker.start()
+
+        # Drain the ready sentinel from the queue in the same way the
+        # real factory does (thread waiter populates ``ready`` /
+        # ``boot_ok``).
+        def _await_ready() -> None:
+            while True:
+                try:
+                    msg = result_queue.get(timeout=1.0)
+                except queue.Empty:
+                    if shutdown_event.is_set():
+                        return
+                    continue
+                if msg.get("kind") == "ready":
+                    boot_ok.set()
+                    ready.set()
+                    return
+                if msg.get("kind") == "error":
+                    ready.set()
+                    return
+
+        threading.Thread(target=_await_ready, daemon=True).start()
+
+        class _FakeProcess:
+            def __init__(self, stop: threading.Event, worker: _ThreadWorker):
+                self._stop = stop
+                self._worker = worker
+
+            def is_alive(self) -> bool:
+                return self._worker._thread.is_alive()
+
+            def join(self, timeout: float | None = None) -> None:
+                self._stop.set()
+                self._worker.join(timeout)
+
+            def kill(self) -> None:
+                self._stop.set()
+
+        fake_process = _FakeProcess(shutdown_event, worker)
+
+        return _WorkerHandle(
+            process=fake_process,  # type: ignore[arg-type]
+            job_queue=job_queue,
+            result_queue=result_queue,
+            gpu_id=gpu_id,
+            worker_id=f"gpu{gpu_id}-fake",
+            ready=ready,
+            boot_ok=boot_ok,
+            shutdown_event=mp_shutdown,
+        )
+
+    return factory
+
+
+@pytest.fixture
+def pool_factory():
+    """Provide a SubprocessGpuPool built against thread workers."""
+
+    async def _build(num_workers: int = 2):
+        pool = SubprocessGpuPool(
+            generator_config=GeneratorConfig(model_path="/models/fake"),
+            pool_config=GpuPoolConfig(num_workers=num_workers),
+            warmup_config=WarmupConfig(enabled=False),
+            worker_factory=_thread_worker_factory(
+                lambda gpu_id: _MockGenerator()),
+        )
+        await pool.start()
+        return pool
+
+    return _build
+
+
+class TestSubprocessGpuPool:
+
+    def test_start_spawns_requested_workers(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=3)
+            try:
+                h = pool.health()
+                assert h.total_workers == 3
+                assert h.available_workers == 3
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_acquire_decrements_available(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=2)
+            try:
+                a = await pool.acquire("sess-a")
+                assert a.worker_id.endswith("-fake")
+                assert pool.health().available_workers == 1
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_acquire_timeout_when_all_busy(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=1)
+            try:
+                await pool.acquire("sess-a")
+                with pytest.raises(PoolAcquireTimeout):
+                    await pool.acquire("sess-b", timeout=0.1)
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_run_returns_worker_result(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=1)
+            try:
+                await pool.acquire("sess-a")
+                result = await pool.run(
+                    "sess-a", GenerationRequest(prompt="hello"))
+                assert result["prompt_echo"] == "hello"
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_release_returns_worker_to_pool(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=1)
+            try:
+                await pool.acquire("sess-a")
+                await pool.release("sess-a")
+                # Now a second acquire should succeed without timeout.
+                a = await pool.acquire("sess-b", timeout=1.0)
+                assert a.worker_id.endswith("-fake")
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_sticky_binding_across_multiple_runs(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=2)
+            try:
+                a1 = await pool.acquire("sess-a")
+                a2 = await pool.acquire("sess-a")
+                assert a1.worker_id == a2.worker_id
+                # Two runs land on the same worker.
+                await pool.run("sess-a", GenerationRequest(prompt="1"))
+                await pool.run("sess-a", GenerationRequest(prompt="2"))
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_run_without_acquire_raises(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=1)
+            try:
+                with pytest.raises(RuntimeError):
+                    await pool.run(
+                        "sess-x", GenerationRequest(prompt="x"))
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_shutdown_is_idempotent(self, pool_factory):
+        async def run():
+            pool = await pool_factory(num_workers=2)
+            await pool.shutdown()
+            await pool.shutdown()  # no raise
+
+        asyncio.run(run())
+
+
+class TestSubprocessGpuPoolFailureModes:
+    """Coverage for boot/runtime failures the pool has to absorb."""
+
+    def test_failed_boot_excluded_from_available(self):
+        """A worker whose factory leaves ``boot_ok`` unset must not be
+        handed out by ``acquire``. Otherwise a session lands on a dead
+        worker and ``run`` blocks forever on the missing result."""
+
+        def factory_with_one_failure(*, gpu_id, generator_config, warmup_config):
+            ctx = mp.get_context("spawn")
+            job_queue: mp.Queue = ctx.Queue()
+            result_queue: mp.Queue = ctx.Queue()
+            ready = threading.Event()
+            boot_ok = threading.Event()
+            mp_shutdown = ctx.Event()
+            ready.set()
+            # gpu_id 0 boots fine; gpu_id 1 fails (boot_ok stays clear).
+            if gpu_id == 0:
+                boot_ok.set()
+
+            class _AliveProcess:
+                def is_alive(self) -> bool:
+                    return True
+                def join(self, timeout: float | None = None) -> None:
+                    return
+                def kill(self) -> None:
+                    return
+
+            return _WorkerHandle(
+                process=_AliveProcess(),  # type: ignore[arg-type]
+                job_queue=job_queue,
+                result_queue=result_queue,
+                gpu_id=gpu_id,
+                worker_id=f"gpu{gpu_id}",
+                ready=ready,
+                boot_ok=boot_ok,
+                shutdown_event=mp_shutdown,
+            )
+
+        async def run():
+            pool = SubprocessGpuPool(
+                generator_config=GeneratorConfig(model_path="/m"),
+                pool_config=GpuPoolConfig(num_workers=2),
+                warmup_config=WarmupConfig(enabled=False),
+                worker_factory=factory_with_one_failure,
+            )
+            await pool.start()
+            try:
+                # Only worker 0 booted; only one slot should be available.
+                assert pool.health().available_workers == 1
+                a = await pool.acquire("sess-a", timeout=0.1)
+                assert a.worker_id == "gpu0"
+                with pytest.raises(PoolAcquireTimeout):
+                    await pool.acquire("sess-b", timeout=0.1)
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())
+
+    def test_release_skips_dead_worker(self, pool_factory):
+        """If a worker died while bound, releasing the session must not
+        return its slot to the available queue — a later acquire would
+        hand the dead slot to a new session."""
+
+        async def run():
+            pool = await pool_factory(num_workers=1)
+            try:
+                await pool.acquire("sess-a")
+                # Simulate the worker dying mid-session.
+                pool._workers[0].process._stop.set()
+                pool._workers[0].process._worker.join(timeout=1.0)
+                await pool.release("sess-a")
+                # Available queue must remain empty.
+                assert pool.health().available_workers == 0
+            finally:
+                await pool.shutdown()
+
+        asyncio.run(run())

fastvideo/tests/entrypoints/streaming/test_prompt_enhancer.py

@@ -0,0 +1,207 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Tests for the provider-agnostic prompt enhancer."""
+from __future__ import annotations
+
+import asyncio
+from dataclasses import dataclass
+
+import pytest
+
+from fastvideo.entrypoints.streaming.prompt import (
+    LLMProvider,
+    LLMProviderError,
+    LLMRequest,
+    LLMResponse,
+    PromptEnhancer,
+)
+
+
+@dataclass
+class _StaticProvider:
+    name: str
+    content: str
+
+    async def complete(self, request: LLMRequest) -> LLMResponse:
+        return LLMResponse(
+            content=self.content,
+            provider=self.name,
+            model=request.model,
+            latency_ms=1.0,
+        )
+
+
+@dataclass
+class _FailingProvider:
+    name: str
+    message: str = "boom"
+    retryable: bool = True
+
+    async def complete(self, request: LLMRequest) -> LLMResponse:
+        raise LLMProviderError(self.message, retryable=self.retryable)
+
+
+class TestConstruction:
+
+    def test_requires_at_least_one_provider(self):
+        with pytest.raises(ValueError):
+            PromptEnhancer(providers=[], model="m")
+
+    def test_registers_system_prompts_from_defaults(self, tmp_path):
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("p", "ok")],
+            model="m",
+        )
+        # Defaults live inside _DEFAULT_SYSTEM_PROMPTS; no hot-reload file
+        # means the enhancer falls back to the shipped values.
+        assert "prompt enhancer" in enh._system_prompts.enhance.lower()
+
+    def test_reads_override_files_when_present(self, tmp_path):
+        (tmp_path / "enhance.txt").write_text("customized enhance")
+        (tmp_path / "auto_extend.txt").write_text("")  # empty -> default
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("p", "ok")],
+            model="m",
+            system_prompt_dir=str(tmp_path),
+        )
+        assert enh._system_prompts.enhance == "customized enhance"
+        # Empty file fell back to the default.
+        assert "continuation assistant" in enh._system_prompts.auto_extend
+
+
+class TestEnhance:
+
+    def test_enhance_returns_primary_provider_response(self):
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("primary", "enhanced!")],
+            model="m",
+        )
+        response = asyncio.run(enh.enhance("a fox"))
+        assert response.content == "enhanced!"
+        assert response.provider == "primary"
+        assert response.fallback_used is False
+
+    def test_enhance_falls_back_on_provider_error(self):
+        enh = PromptEnhancer(
+            providers=[
+                _FailingProvider("a"),
+                _StaticProvider("b", "fallback!"),
+            ],
+            model="m",
+        )
+        response = asyncio.run(enh.enhance("x"))
+        assert response.provider == "b"
+        assert response.content == "fallback!"
+        assert response.fallback_used is True
+
+    def test_enhance_stops_on_non_retryable_error(self):
+        enh = PromptEnhancer(
+            providers=[
+                _FailingProvider("a", message="hard-fail", retryable=False),
+                _StaticProvider("b", "should-not-be-reached"),
+            ],
+            model="m",
+        )
+        with pytest.raises(LLMProviderError, match="hard-fail"):
+            asyncio.run(enh.enhance("x"))
+
+    def test_enhance_raises_when_all_providers_fail(self):
+        enh = PromptEnhancer(
+            providers=[
+                _FailingProvider("a"),
+                _FailingProvider("b"),
+            ],
+            model="m",
+        )
+        with pytest.raises(LLMProviderError):
+            asyncio.run(enh.enhance("x"))
+
+
+class TestAutoExtendAndRewrite:
+
+    def test_auto_extend_joins_prior_prompts(self):
+        captured: list[str] = []
+
+        class _Capturer:
+            name = "cap"
+
+            async def complete(self, request: LLMRequest) -> LLMResponse:
+                user = next(m.content for m in request.messages
+                            if m.role == "user")
+                captured.append(user)
+                return LLMResponse(
+                    content="next",
+                    provider=self.name,
+                    model=request.model,
+                    latency_ms=1.0,
+                )
+
+        enh = PromptEnhancer(providers=[_Capturer()], model="m")
+        asyncio.run(enh.auto_extend(["first", "second"]))
+        assert captured[0] == "first\nsecond"
+
+    def test_rewrite_passes_seed_through(self):
+        captured: list[str] = []
+
+        class _Capturer:
+            name = "cap"
+
+            async def complete(self, request: LLMRequest) -> LLMResponse:
+                user = next(m.content for m in request.messages
+                            if m.role == "user")
+                captured.append(user)
+                return LLMResponse(
+                    content="one\ntwo\nthree",
+                    provider=self.name,
+                    model=request.model,
+                    latency_ms=1.0,
+                )
+
+        enh = PromptEnhancer(providers=[_Capturer()], model="m")
+        response = asyncio.run(enh.rewrite("seed prompt"))
+        assert captured == ["seed prompt"]
+        assert response.content.splitlines() == ["one", "two", "three"]
+
+
+class TestRegisterProvider:
+
+    def test_register_appends_by_default(self):
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("primary", "a")],
+            model="m",
+        )
+        enh.register_provider(_StaticProvider("extra", "b"))
+        assert [p.name for p in enh.providers] == ["primary", "extra"]
+
+    def test_register_priority_zero_makes_primary(self):
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("old", "a")],
+            model="m",
+        )
+        enh.register_provider(
+            _StaticProvider("new-primary", "b"), priority=0)
+        assert enh.providers[0].name == "new-primary"
+
+    def test_registered_provider_is_used_in_enhance(self):
+        enh = PromptEnhancer(
+            providers=[_FailingProvider("broken")],
+            model="m",
+        )
+        enh.register_provider(_StaticProvider("fallback", "ok"))
+        response = asyncio.run(enh.enhance("x"))
+        assert response.content == "ok"
+        assert response.fallback_used is True
+
+
+class TestHotReload:
+
+    def test_reload_picks_up_new_file(self, tmp_path):
+        (tmp_path / "enhance.txt").write_text("first version")
+        enh = PromptEnhancer(
+            providers=[_StaticProvider("p", "ok")],
+            model="m",
+            system_prompt_dir=str(tmp_path),
+        )
+        assert enh._system_prompts.enhance == "first version"
+        (tmp_path / "enhance.txt").write_text("second version")
+        enh.reload_system_prompts()
+        assert enh._system_prompts.enhance == "second version"

fastvideo/tests/entrypoints/streaming/test_prompt_providers.py

@@ -0,0 +1,298 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Tests for the LLM provider protocol + built-in adapters.
+
+Real Cerebras/Groq HTTP calls are stubbed with a fake ``httpx`` module
+inserted into ``sys.modules`` so the unit tests don't depend on
+external API availability or paid keys.
+"""
+from __future__ import annotations
+
+import asyncio
+import sys
+import types
+from dataclasses import dataclass
+from typing import Any
+
+import pytest
+
+from fastvideo.entrypoints.streaming.prompt.providers import (
+    CerebrasProvider,
+    GroqProvider,
+    LLMMessage,
+    LLMProvider,
+    LLMProviderError,
+    LLMRequest,
+    LLMTimeoutError,
+)
+
+
+@dataclass
+class _FakeResponse:
+    status_code: int
+    payload: dict[str, Any]
+    text: str = ""
+
+    def json(self) -> dict[str, Any]:
+        return self.payload
+
+
+class _FakeAsyncClient:
+
+    def __init__(self, response_or_exc, *, captured: list) -> None:
+        self._response_or_exc = response_or_exc
+        self._captured = captured
+
+    async def __aenter__(self):
+        return self
+
+    async def __aexit__(self, exc_type, exc, tb):
+        return None
+
+    async def post(self, url, **kwargs):
+        self._captured.append({"url": url, **kwargs})
+        if isinstance(self._response_or_exc, Exception):
+            raise self._response_or_exc
+        return self._response_or_exc
+
+
+def _install_fake_httpx(
+    monkeypatch,
+    *,
+    response: _FakeResponse | None = None,
+    exception: Exception | None = None,
+) -> list[dict]:
+    """Replace ``sys.modules['httpx']`` with a stub exposing the bits
+    providers touch: ``AsyncClient``, ``HTTPError``, ``TimeoutException``.
+
+    Returns a list the test can inspect to see what requests went out.
+    """
+    captured: list[dict] = []
+
+    class _HTTPError(Exception):
+        pass
+
+    class _TimeoutException(_HTTPError):
+        pass
+
+    payload = response if exception is None else exception
+
+    def _client_factory(*_args, **_kwargs):
+        return _FakeAsyncClient(payload, captured=captured)
+
+    stub = types.SimpleNamespace(
+        AsyncClient=_client_factory,
+        HTTPError=_HTTPError,
+        TimeoutException=_TimeoutException,
+    )
+    monkeypatch.setitem(sys.modules, "httpx", stub)
+    return captured
+
+
+# ----------------------------------------------------------------------
+# Cerebras
+# ----------------------------------------------------------------------
+
+
+class TestCerebrasProvider:
+
+    def test_is_llm_provider(self):
+        assert isinstance(CerebrasProvider(api_key="x"), LLMProvider)
+
+    def test_requires_api_key(self, monkeypatch):
+        monkeypatch.delenv("CEREBRAS_API_KEY", raising=False)
+        provider = CerebrasProvider()
+        with pytest.raises(LLMProviderError, match="CEREBRAS_API_KEY"):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+    def test_api_key_from_env(self, monkeypatch):
+        monkeypatch.setenv("CEREBRAS_API_KEY", "from-env")
+        provider = CerebrasProvider()
+        assert provider.api_key == "from-env"
+
+    def test_explicit_api_key_wins(self, monkeypatch):
+        monkeypatch.setenv("CEREBRAS_API_KEY", "from-env")
+        provider = CerebrasProvider(api_key="explicit")
+        assert provider.api_key == "explicit"
+
+    def test_success_path(self, monkeypatch):
+        captured = _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=200,
+                payload={
+                    "choices": [{"message": {"content": "enhanced"}}],
+                },
+            ),
+        )
+        provider = CerebrasProvider(api_key="secret")
+        result = asyncio.run(provider.complete(
+            LLMRequest(
+                messages=[LLMMessage(role="user", content="a fox")],
+                model="gpt-oss-120b",
+                max_tokens=64,
+                temperature=0.7,
+            )))
+        assert result.content == "enhanced"
+        assert result.provider == "cerebras"
+        assert result.model == "gpt-oss-120b"
+        # Verify the HTTP request was shaped correctly.
+        body = captured[0]["json"]
+        assert body["model"] == "gpt-oss-120b"
+        assert body["messages"] == [{"role": "user", "content": "a fox"}]
+        assert captured[0]["headers"]["Authorization"] == "Bearer secret"
+
+    def test_http_error_wrapped(self, monkeypatch):
+        # Stage the stub first, then raise that stub's own HTTPError so
+        # the provider's ``except httpx.HTTPError`` catches it.
+        stub = types.SimpleNamespace()
+
+        class _HTTPError(Exception):
+            pass
+
+        class _TimeoutException(_HTTPError):
+            pass
+
+        stub.HTTPError = _HTTPError
+        stub.TimeoutException = _TimeoutException
+        stub.AsyncClient = lambda *a, **k: _FakeAsyncClient(
+            _HTTPError("boom"), captured=[])
+        monkeypatch.setitem(sys.modules, "httpx", stub)
+
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="HTTP"):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+    def test_timeout_wrapped(self, monkeypatch):
+        _install_fake_httpx(monkeypatch)
+        # Install timeout exception after stubbing.
+        stub = sys.modules["httpx"]
+
+        def raising_factory(*_a, **_kw):
+            raise stub.TimeoutException("timed out")  # type: ignore[attr-defined]
+
+        stub.AsyncClient = raising_factory  # type: ignore[attr-defined]
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMTimeoutError):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+    def test_4xx_raises_non_retryable_provider_error(self, monkeypatch):
+        _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=401,
+                payload={},
+                text="unauthorized",
+            ),
+        )
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="401") as excinfo:
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+        assert excinfo.value.retryable is False
+
+    def test_429_raises_retryable_provider_error(self, monkeypatch):
+        _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=429,
+                payload={},
+                text="rate limited",
+            ),
+        )
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="429") as excinfo:
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+        assert excinfo.value.retryable is True
+
+    def test_5xx_raises_retryable_provider_error(self, monkeypatch):
+        _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=503,
+                payload={},
+                text="service unavailable",
+            ),
+        )
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="503") as excinfo:
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+        assert excinfo.value.retryable is True
+
+    def test_non_json_body_raises_provider_error(self, monkeypatch):
+        # Simulate a proxy/load-balancer HTML error page slipping in
+        # with a 200 status: response.json() raises, and the provider
+        # must wrap it in an LLMProviderError instead of bubbling.
+        class _BadJsonResponse:
+            status_code = 200
+            text = "<html>oops</html>"
+
+            def json(self):
+                raise ValueError("Expecting value: line 1 column 1 (char 0)")
+
+        _install_fake_httpx(
+            monkeypatch,
+            response=_BadJsonResponse(),  # type: ignore[arg-type]
+        )
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="non-JSON"):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+    def test_empty_choices_raises(self, monkeypatch):
+        _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=200, payload={"choices": []}),
+        )
+        provider = CerebrasProvider(api_key="k")
+        with pytest.raises(LLMProviderError, match="no choices"):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+
+# ----------------------------------------------------------------------
+# Groq
+# ----------------------------------------------------------------------
+
+
+class TestGroqProvider:
+
+    def test_is_llm_provider(self):
+        assert isinstance(GroqProvider(api_key="x"), LLMProvider)
+
+    def test_requires_api_key(self, monkeypatch):
+        monkeypatch.delenv("GROQ_API_KEY", raising=False)
+        provider = GroqProvider()
+        with pytest.raises(LLMProviderError, match="GROQ_API_KEY"):
+            asyncio.run(provider.complete(
+                LLMRequest(messages=[], model="m")))
+
+    def test_api_key_from_env(self, monkeypatch):
+        monkeypatch.setenv("GROQ_API_KEY", "from-env")
+        provider = GroqProvider()
+        assert provider.api_key == "from-env"
+
+    def test_success_path(self, monkeypatch):
+        captured = _install_fake_httpx(
+            monkeypatch,
+            response=_FakeResponse(
+                status_code=200,
+                payload={
+                    "choices": [{"message": {"content": "groq out"}}],
+                },
+            ),
+        )
+        provider = GroqProvider(api_key="secret")
+        result = asyncio.run(provider.complete(
+            LLMRequest(
+                messages=[LLMMessage(role="user", content="a deer")],
+                model="llama-3.1-70b",
+            )))
+        assert result.content == "groq out"
+        assert result.provider == "groq"
+        assert captured[0]["headers"]["Authorization"] == "Bearer secret"

fastvideo/tests/entrypoints/streaming/test_router.py

@@ -0,0 +1,279 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Router tests — registry semantics + health loop behavior.
+
+Avoids real WebSocket proxying (the bridge requires the ``websockets``
+package and two running uvicorn processes); test_server.py already
+covers the end-to-end WS protocol against a direct backend.
+"""
+from __future__ import annotations
+
+import asyncio
+
+import pytest
+
+from fastvideo.entrypoints.streaming.router.config import (
+    ReplicaEndpoint,
+    RouterConfig,
+)
+from fastvideo.entrypoints.streaming.router.registry import (
+    ReplicaRegistry,
+    ReplicaStatus,
+    run_health_check_loop,
+)
+
+
+def _registry(
+    *,
+    num_primary: int = 1,
+    num_secondary: int = 1,
+) -> ReplicaRegistry:
+    replicas = [
+        ReplicaEndpoint(
+            url=f"http://primary-{i}:8000",
+            primary=True,
+        )
+        for i in range(num_primary)
+    ] + [
+        ReplicaEndpoint(
+            url=f"http://secondary-{i}:8000",
+            primary=False,
+        )
+        for i in range(num_secondary)
+    ]
+    return ReplicaRegistry(replicas)
+
+
+class TestReplicaRegistry:
+
+    def test_requires_replicas(self):
+        with pytest.raises(ValueError):
+            ReplicaRegistry([])
+
+    def test_select_none_when_all_unknown(self):
+        assert _registry().select() is None
+
+    def test_select_prefers_healthy_primary(self):
+        reg = _registry()
+
+        async def promote():
+            primary = reg.primaries()[0]
+            await reg.record_success(
+                primary, recovery_threshold=1, latency_ms=1.0)
+            # Also mark secondary healthy — primary should still win.
+            sec = next(r for r in reg.all() if not r.primary)
+            await reg.record_success(
+                sec, recovery_threshold=1, latency_ms=1.0)
+            return reg.select()
+
+        pick = asyncio.run(promote())
+        assert pick is not None
+        assert pick.primary
+
+    def test_falls_back_to_secondary_when_primary_unhealthy(self):
+        reg = _registry()
+
+        async def run():
+            primary = reg.primaries()[0]
+            sec = next(r for r in reg.all() if not r.primary)
+            # Fail primary past threshold, succeed secondary.
+            for _ in range(3):
+                await reg.record_failure(
+                    primary, failure_threshold=3, reason="mock")
+            await reg.record_success(
+                sec, recovery_threshold=1, latency_ms=1.0)
+            return reg.select()
+
+        pick = asyncio.run(run())
+        assert pick is not None
+        assert not pick.primary
+
+    def test_failure_threshold_transitions_to_unhealthy(self):
+        reg = _registry()
+        primary = reg.primaries()[0]
+
+        async def run():
+            for _ in range(2):
+                await reg.record_failure(
+                    primary, failure_threshold=3, reason="x")
+            assert primary.health.status is not ReplicaStatus.UNHEALTHY
+            await reg.record_failure(
+                primary, failure_threshold=3, reason="x")
+            return primary
+
+        result = asyncio.run(run())
+        assert result.health.status is ReplicaStatus.UNHEALTHY
+        assert result.health.consecutive_failures == 3
+
+    def test_recovery_threshold_returns_to_healthy(self):
+        reg = _registry()
+        primary = reg.primaries()[0]
+
+        async def run():
+            for _ in range(3):
+                await reg.record_failure(
+                    primary, failure_threshold=3, reason="x")
+            assert primary.health.status is ReplicaStatus.UNHEALTHY
+            for _ in range(2):
+                await reg.record_success(
+                    primary, recovery_threshold=2, latency_ms=5.0)
+            return primary
+
+        result = asyncio.run(run())
+        assert result.health.status is ReplicaStatus.HEALTHY
+        assert result.health.last_latency_ms == 5.0
+
+    def test_record_success_resets_failure_counter(self):
+        reg = _registry()
+        primary = reg.primaries()[0]
+
+        async def run():
+            await reg.record_failure(
+                primary, failure_threshold=10, reason="x")
+            await reg.record_success(
+                primary, recovery_threshold=1, latency_ms=1.0)
+            return primary
+
+        result = asyncio.run(run())
+        assert result.health.consecutive_failures == 0
+
+
+class TestHealthCheckLoop:
+
+    def test_loop_transitions_replicas_on_probe_results(self):
+        config = RouterConfig(
+            replicas=[
+                ReplicaEndpoint(url="http://a", primary=True),
+                ReplicaEndpoint(url="http://b"),
+            ],
+            health_check_interval_seconds=0.01,
+            failure_threshold=1,
+            recovery_threshold=1,
+        )
+        reg = ReplicaRegistry(config.replicas)
+        stop_event = asyncio.Event()
+
+        calls: list[str] = []
+
+        async def probe(url, *, timeout):
+            calls.append(url)
+            if "http://a" in url:
+                return 1.0, None
+            return 0.0, "mock failure"
+
+        async def run() -> None:
+            task = asyncio.create_task(run_health_check_loop(
+                registry=reg, config=config, stop_event=stop_event,
+                http_get=probe,
+            ))
+            await asyncio.sleep(0.05)
+            stop_event.set()
+            await task
+
+        asyncio.run(run())
+        a = reg.get("http://a")
+        b = reg.get("http://b")
+        assert a is not None
+        assert b is not None
+        assert a.health.status is ReplicaStatus.HEALTHY
+        assert b.health.status is ReplicaStatus.UNHEALTHY
+        assert any("/health" in c for c in calls)
+
+
+class TestRouterApp:
+
+    def test_status_endpoint_lists_replicas(self):
+        from starlette.testclient import TestClient
+
+        from fastvideo.entrypoints.streaming.router.main import build_router_app
+
+        config = RouterConfig(
+            replicas=[
+                ReplicaEndpoint(url="http://a", primary=True),
+                ReplicaEndpoint(url="http://b"),
+            ],
+            health_check_interval_seconds=60,  # don't actually poll
+        )
+        reg = ReplicaRegistry(config.replicas)
+        app = build_router_app(config, registry=reg)
+        client = TestClient(app)
+        response = client.get("/status")
+        body = response.json()
+        urls = {r["url"] for r in body["replicas"]}
+        assert urls == {"http://a", "http://b"}
+        # Initial status is UNKNOWN.
+        assert all(r["status"] == "unknown" for r in body["replicas"])
+
+    def test_ws_rejects_when_no_healthy_replica(self):
+        from starlette.testclient import TestClient
+
+        from fastvideo.entrypoints.streaming.router.main import build_router_app
+
+        config = RouterConfig(
+            replicas=[ReplicaEndpoint(url="http://a", primary=True)],
+            health_check_interval_seconds=60,
+        )
+        reg = ReplicaRegistry(config.replicas)
+        app = build_router_app(config, registry=reg)
+        client = TestClient(app)
+        with client.websocket_connect("/v1/stream") as ws:
+            err = ws.receive_json()
+            assert err["type"] == "error"
+            assert err["code"] == "gpu_unavailable"
+
+
+class TestUnknownToHealthyImmediate:
+    """Initial probe must promote UNKNOWN -> HEALTHY without waiting for recovery_threshold."""
+
+    def test_first_success_promotes_unknown(self):
+        reg = _registry(num_primary=1, num_secondary=0)
+        primary = reg.primaries()[0]
+        assert primary.health.status is ReplicaStatus.UNKNOWN
+
+        async def run():
+            await reg.record_success(primary, recovery_threshold=10, latency_ms=1.0)
+
+        asyncio.run(run())
+        assert primary.health.status is ReplicaStatus.HEALTHY
+
+    def test_unhealthy_recovery_still_gated_by_threshold(self):
+        reg = _registry(num_primary=1, num_secondary=0)
+        primary = reg.primaries()[0]
+
+        async def run():
+            for _ in range(3):
+                await reg.record_failure(primary, failure_threshold=3, reason="x")
+            assert primary.health.status is ReplicaStatus.UNHEALTHY
+            await reg.record_success(primary, recovery_threshold=2, latency_ms=1.0)
+            assert primary.health.status is ReplicaStatus.UNHEALTHY  # 1/2 successes
+            await reg.record_success(primary, recovery_threshold=2, latency_ms=1.0)
+            assert primary.health.status is ReplicaStatus.HEALTHY    # 2/2 successes
+
+        asyncio.run(run())
+
+
+class TestConfigValidation:
+    """RouterConfig.__post_init__ rejects malformed configs."""
+
+    def test_rejects_path_in_url(self):
+        with pytest.raises(ValueError, match="without a path"):
+            RouterConfig(replicas=[ReplicaEndpoint(url="http://host:8000/api")])
+
+    def test_rejects_query_in_url(self):
+        with pytest.raises(ValueError, match="query/fragment"):
+            RouterConfig(replicas=[ReplicaEndpoint(url="http://host:8000?x=1")])
+
+    def test_rejects_fragment_in_url(self):
+        with pytest.raises(ValueError, match="query/fragment"):
+            RouterConfig(replicas=[ReplicaEndpoint(url="http://host:8000#frag")])
+
+    def test_rejects_duplicate_urls(self):
+        with pytest.raises(ValueError, match="Duplicate"):
+            RouterConfig(replicas=[
+                ReplicaEndpoint(url="http://host:8000"),
+                ReplicaEndpoint(url="http://host:8000"),
+            ])
+
+    def test_accepts_trailing_slash(self):
+        # parsed.path == "/" should be allowed
+        cfg = RouterConfig(replicas=[ReplicaEndpoint(url="http://host:8000/")])
+        assert cfg.replicas[0].url == "http://host:8000/"

fastvideo/tests/entrypoints/streaming/test_worker.py

@@ -0,0 +1,86 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Unit coverage for :mod:`fastvideo.entrypoints.streaming.worker`.
+
+The full ``worker_main`` loop runs in a subprocess and is exercised via
+``test_gpu_pool.py``'s subprocess integration tests. This file covers
+the in-process pieces:
+
+* the two-segment warmup feeds segment 1's continuation state into
+  segment 2 so both compile branches are primed before the worker
+  reports ready
+* result-shape extractors handle both attribute-style and dict-style
+  generator returns
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any
+
+from fastvideo.api.schema import (
+    ContinuationState,
+    GenerationRequest,
+    WarmupConfig,
+)
+from fastvideo.entrypoints.streaming.worker import (
+    _extract_continuation_state,
+    _warmup_worker,
+)
+
+
+@dataclass
+class _RecordingGenerator:
+    """Captures the requests passed to ``generate`` and returns a
+    canned :class:`ContinuationState` on the first call so the warmup
+    can feed it into the second call.
+    """
+
+    state_to_return: ContinuationState | None = field(default_factory=lambda: ContinuationState(
+        kind="ltx2.v1",
+        payload={"schema_version": 1, "segment_index": 1},
+    ))
+    requests: list[GenerationRequest] = field(default_factory=list)
+
+    def generate(self, request: GenerationRequest) -> dict[str, Any]:
+        self.requests.append(request)
+        return {"frames": [], "state": self.state_to_return}
+
+
+class TestWarmupTwoSegment:
+
+    def test_warmup_runs_segment_one_then_segment_two_with_returned_state(self) -> None:
+        gen = _RecordingGenerator()
+        _warmup_worker(gen, WarmupConfig(enabled=True, prompt="warm"))
+
+        assert len(gen.requests) == 2
+
+        seg1 = gen.requests[0]
+        assert seg1.state is None
+        assert seg1.output.return_state is True
+
+        seg2 = gen.requests[1]
+        assert seg2.state is gen.state_to_return
+
+    def test_warmup_passes_through_when_no_state_returned(self) -> None:
+        gen = _RecordingGenerator(state_to_return=None)
+        _warmup_worker(gen, WarmupConfig(enabled=True, prompt="warm"))
+
+        assert len(gen.requests) == 2
+        assert gen.requests[1].state is None
+
+
+class TestExtractContinuationState:
+
+    def test_extracts_from_attribute(self) -> None:
+
+        class _R:
+            state = ContinuationState(kind="k", payload={})
+
+        assert _extract_continuation_state(_R()).kind == "k"
+
+    def test_extracts_from_dict(self) -> None:
+        state = ContinuationState(kind="k", payload={})
+        assert _extract_continuation_state({"state": state}) is state
+
+    def test_returns_none_for_missing(self) -> None:
+        assert _extract_continuation_state({}) is None
+        assert _extract_continuation_state(object()) is None

fastvideo/tests/modal/pr_test.py

@@ -240,17 +240,23 @@ def run_lora_extraction_tests():
               ],
               volumes={"/root/data": model_vol})
 def run_performance_tests():
-    # Dashboard runs after compare_baseline regardless of regression result so
-    # the trend view is always available when investigating a failed gate.
+    # compare_baseline.py emits normalized_perf_*.json artifacts for manual
+    # performance-baseline reseeds when the rolling comparison runs.
     run_test(
         "export HF_HOME='/root/data/.cache' && "
         "export PERFORMANCE_TRACKING_ROOT='/tmp/perf-tracking' && "
         "hf auth login --token $HF_API_KEY && "
-        "pytest ./fastvideo/tests/performance -vs && "
-        "{ python ./fastvideo/tests/performance/compare_baseline.py; "
+        "pytest ./fastvideo/tests/performance -vs; "
+        "PYTEST_RC=$?; "
+        "PERF_RC=0; "
+        "if [ $PYTEST_RC -eq 0 ]; then "
+        "python ./fastvideo/tests/performance/compare_baseline.py; "
         "PERF_RC=$?; "
+        "fi; "
         "python ./fastvideo/tests/performance/dashboard.py || true; "
-        "exit $PERF_RC; }")
+        "FINAL_RC=$PYTEST_RC; "
+        "if [ $FINAL_RC -eq 0 ]; then FINAL_RC=$PERF_RC; fi; "
+        "exit $FINAL_RC")
 
 
 @app.function(gpu="L40S:1",

fastvideo/tests/modal/ssim_test.py

@@ -485,6 +485,10 @@ def _prepare_ssim_workspace(
     ./build.sh
     cd ..
     uv pip install git+https://github.com/microsoft/MoGe.git
+    # Stable Audio Open 1.0 inference deps (optional in basic install,
+    # required by `StableAudioDenoisingStage`; consumed by
+    # `test_stable_audio_similarity.py`).
+    uv pip install k_diffusion einops_exts alias_free_torch torchsde
     export HF_HOME='/root/data/.cache'
     hf auth login --token "$HF_API_KEY"
     """

fastvideo/tests/performance/compare_baseline.py

@@ -36,7 +36,23 @@ TRACKING_ROOT = os.environ.get(
     "PERFORMANCE_TRACKING_ROOT",
     "/tmp/perf-tracking",
 )
+PERF_REPORTS_DIR = os.environ.get("PERF_REPORTS_DIR", "/root/data/perf_reports")
 MAX_REGRESSION = float(os.environ.get("PERF_MAX_REGRESSION", "0.05"))
+METRICS = (
+    ("latency", "Latency", 3),
+    ("throughput", "Throughput", 3),
+    ("memory", "Memory", 1),
+    ("text_encoder_time_s", "Text Enc", 3),
+    ("dit_time_s", "DiT", 3),
+    ("vae_decode_time_s", "VAE Decode", 3),
+)
+LOWER_IS_BETTER_METRICS = {
+    "latency",
+    "memory",
+    "text_encoder_time_s",
+    "dit_time_s",
+    "vae_decode_time_s",
+}
 
 
 def _should_persist_tracking() -> bool:
@@ -44,7 +60,6 @@ def _should_persist_tracking() -> bool:
     branch = os.environ.get("BUILDKITE_BRANCH", "")
     return test_scope == "full" and branch == "main"
 
-
 def _load_current_results() -> list[dict[str, Any]]:
     pattern = os.path.join(RESULTS_DIR, "perf_*.json")
     records: list[dict[str, Any]] = []
@@ -54,7 +69,14 @@ def _load_current_results() -> list[dict[str, Any]]:
     return records
 
 
-def _normalize_record(result: dict[str, Any]) -> dict[str, Any]:
+def normalize_performance_result(result: dict[str, Any]) -> dict[str, Any]:
+    """Normalize a raw perf_*.json result into the HF tracking schema.
+
+    The Buildkite artifact intentionally keeps the raw benchmark output from
+    test_inference_performance.py. Baseline comparison, main-branch persistence,
+    and manual baseline reseeds should all use this mapping so the stored HF
+    records do not drift from the artifact schema.
+    """
     benchmark_id = result.get("benchmark_id", "unknown")
     model_id = benchmark_id
 
@@ -66,6 +88,9 @@ def _normalize_record(result: dict[str, Any]) -> dict[str, Any]:
     latency = safe_float(result.get("avg_generation_time_s"))
     throughput = safe_float(result.get("throughput_fps"))
     memory = safe_float(result.get("max_peak_memory_mb"))
+    text_encoder_time = safe_float(result.get("text_encoder_time_s"))
+    dit_time = safe_float(result.get("dit_time_s"))
+    vae_decode_time = safe_float(result.get("vae_decode_time_s"))
 
     return {
         "model_id": model_id,
@@ -75,10 +100,17 @@ def _normalize_record(result: dict[str, Any]) -> dict[str, Any]:
         "latency": latency,
         "throughput": throughput,
         "memory": memory,
+        "text_encoder_time_s": text_encoder_time,
+        "dit_time_s": dit_time,
+        "vae_decode_time_s": vae_decode_time,
         "success": True,
     }
 
 
+def _normalize_record(result: dict[str, Any]) -> dict[str, Any]:
+    return normalize_performance_result(result)
+
+
 def _write_tracking_record(record: dict[str, Any]) -> str:
     model_dir = os.path.join(TRACKING_ROOT, sanitize(record["model_id"]))
     os.makedirs(model_dir, exist_ok=True)
@@ -93,6 +125,24 @@ def _write_tracking_record(record: dict[str, Any]) -> str:
     return out_path
 
 
+def _write_normalized_artifact(record: dict[str, Any]) -> None:
+    try:
+        results_dir = os.path.join(PERF_REPORTS_DIR, "results")
+        os.makedirs(results_dir, exist_ok=True)
+        timestamp = sanitize(record["timestamp"])
+        model_id = sanitize(record["model_id"])
+        commit = sanitize(record["commit_sha"] or "unknown")
+        path = os.path.join(
+            results_dir,
+            f"normalized_perf_{model_id}_{timestamp}_{commit}.json",
+        )
+        with open(path, "w", encoding="utf-8") as f:
+            json.dump(record, f, indent=2)
+        print(f"Normalized performance result written to {path}")
+    except Exception as e:
+        print(f"Failed to write normalized performance result artifact: {e}")
+
+
 def _baseline_metric(records: list[dict[str, Any]], key: str) -> float | None:
     values = [safe_float(r.get(key)) for r in records]
     values = [v for v in values if v is not None]
@@ -108,7 +158,8 @@ def _check_regressions(
 ) -> list[str]:
     failures: list[str] = []
 
-    for metric in ("latency", "memory"):
+    for metric in ("latency", "memory", "text_encoder_time_s", "dit_time_s",
+                   "vae_decode_time_s"):
         baseline = _baseline_metric(baseline_records, metric)
         curr = safe_float(current.get(metric))
         if baseline is None or curr is None or baseline <= 0:
@@ -141,7 +192,7 @@ def _metric_delta_percent(
     if curr is None or baseline is None or baseline <= 0:
         return None
 
-    if metric in ("latency", "memory"):
+    if metric in LOWER_IS_BETTER_METRICS:
         return (curr - baseline) / baseline * 100.0
     if metric == "throughput":
         return (baseline - curr) / baseline * 100.0
@@ -161,28 +212,27 @@ def _build_summary_row(
 ) -> dict[str, Any]:
     """Format a single benchmark result as a row for the Markdown table."""
 
-    latency_base = _baseline_metric(baseline_records, "latency")
-    throughput_base = _baseline_metric(baseline_records, "throughput")
-    memory_base = _baseline_metric(baseline_records, "memory")
+    metric_values: dict[str, dict[str, float | None]] = {}
+    regressions: list[float] = []
+    for metric, _label, _precision in METRICS:
+        curr = safe_float(record.get(metric))
+        baseline = _baseline_metric(baseline_records, metric)
+        regression = _metric_delta_percent(metric, record, baseline_records)
+        metric_values[metric] = {
+            "curr": curr,
+            "base": baseline,
+            "regression_pct": regression,
+        }
+        if regression is not None:
+            regressions.append(regression)
 
-    # Calculate percentages for the 'Worst Regression' column
-    latency_reg = _metric_delta_percent("latency", record, baseline_records)
-    throughput_reg = _metric_delta_percent("throughput", record, baseline_records)
-    memory_reg = _metric_delta_percent("memory", record, baseline_records)
-
-    regressions = [v for v in (latency_reg, throughput_reg, memory_reg) if v is not None]
     worst_regression_pct = max(regressions) if regressions else None
 
     return {
         "model_id": record["model_id"],
         "gpu_type": record["gpu_type"],
         "baseline_n": len(baseline_records),
-        "latency_curr": safe_float(record.get("latency")),
-        "latency_base": latency_base,
-        "throughput_curr": safe_float(record.get("throughput")),
-        "throughput_base": throughput_base,
-        "memory_curr": safe_float(record.get("memory")),
-        "memory_base": memory_base,
+        "metrics": metric_values,
         "worst_regression_pct": worst_regression_pct,
         "failed": has_failed,
     }
@@ -199,24 +249,24 @@ def _build_markdown_summary(
         "",
         ("| Model | GPU | Baseline N | Latency (curr/base) | "
          "Throughput (curr/base) | Memory (curr/base) | "
-         "Worst Regression | Status |"),
-        "|---|---|---:|---|---|---|---:|---|",
+         "Text Enc (curr/base) | DiT (curr/base) | "
+         "VAE Decode (curr/base) | Worst Regression | Status |"),
+        "|---|---|---:|---|---|---|---|---|---|---:|---|",
     ]
 
     for row in summary_rows:
-        latency = (f"{_compact_value(row['latency_curr'])} / "
-                   f"{_compact_value(row['latency_base'])}")
-        throughput = (f"{_compact_value(row['throughput_curr'])} / "
-                      f"{_compact_value(row['throughput_base'])}")
-        memory = (f"{_compact_value(row['memory_curr'], 1)} / "
-                  f"{_compact_value(row['memory_base'], 1)}")
+        metric_cells = []
+        for metric, _label, precision in METRICS:
+            values = row["metrics"][metric]
+            metric_cells.append(f"{_compact_value(values['curr'], precision)} / "
+                                f"{_compact_value(values['base'], precision)}")
 
         worst_reg = ("n/a" if row["worst_regression_pct"] is None else f"{row['worst_regression_pct']:.1f}%")
         status = "FAIL" if row["failed"] else "PASS"
 
         lines.append(f"| {row['model_id']} | {row['gpu_type']} | "
                      f"{row['baseline_n']} | "
-                     f"{latency} | {throughput} | {memory} | "
+                     f"{' | '.join(metric_cells)} | "
                      f"{worst_reg} | {status} |")
 
     return "\n".join(lines) + "\n"
@@ -233,11 +283,10 @@ def _emit_markdown_summary(markdown: str, commit_sha: str) -> None:
 
     # 2. Write to Modal volume for Buildkite to pick up in post-run hook
     try:
-        perf_reports_dir = "/root/data/perf_reports"
-        os.makedirs(perf_reports_dir, exist_ok=True)
+        os.makedirs(PERF_REPORTS_DIR, exist_ok=True)
         short_sha = commit_sha[:7] if commit_sha else "unknown"
         timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        report_path = os.path.join(perf_reports_dir, f"perf_{short_sha}_{timestamp}.md")
+        report_path = os.path.join(PERF_REPORTS_DIR, f"perf_{short_sha}_{timestamp}.md")
         with open(report_path, "w", encoding="utf-8") as f:
             f.write(markdown + "\n")
         print(f"Performance report written to {report_path}")
@@ -286,6 +335,8 @@ def main() -> int:
             record["success"] = not failures
             all_failures.extend(failures)
 
+        _write_normalized_artifact(record)
+
         # Strict upload: a silent failure would freeze the rolling baseline.
         if persist_tracking:
             current_path = _write_tracking_record(record)

fastvideo/tests/performance/dashboard.py

@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 import os
+from html import escape
 from datetime import datetime
 
 import plotly.express as px
@@ -7,6 +8,15 @@ import pandas as pd
 
 from hf_store import sync_from_hf, load_as_dataframe
 
+METRICS = (
+    "latency",
+    "throughput",
+    "memory",
+    "text_encoder_time_s",
+    "dit_time_s",
+    "vae_decode_time_s",
+)
+
 # -----------------------------
 # 1. Grouping
 # -----------------------------
@@ -19,15 +29,36 @@ def group_data(df: pd.DataFrame):
 # -----------------------------
 # 2. Plot builder
 # -----------------------------
-def build_plots(df: pd.DataFrame) -> list:
+def build_plots(df: pd.DataFrame) -> tuple[list, list[dict[str, object]]]:
     figs = []
+    skipped_metrics: list[dict[str, object]] = []
 
     for (model_id, gpu_type), g in group_data(df):
         g = g.sort_values("timestamp")
 
         # One chart per metric so the y-axes aren't on wildly different scales
-        for metric in ("latency", "throughput", "memory"):
-            if g[metric].isna().all():
+        for metric in METRICS:
+            if metric not in g.columns:
+                skipped_metrics.append({
+                    "model_id": model_id,
+                    "gpu_type": gpu_type,
+                    "metric": metric,
+                    "reason": "column missing from loaded records",
+                    "records": len(g),
+                    "non_null": 0,
+                })
+                continue
+
+            non_null = int(g[metric].notna().sum())
+            if non_null == 0:
+                skipped_metrics.append({
+                    "model_id": model_id,
+                    "gpu_type": gpu_type,
+                    "metric": metric,
+                    "reason": "no non-null values in loaded records",
+                    "records": len(g),
+                    "non_null": non_null,
+                })
                 continue
 
             fig = px.line(
@@ -41,20 +72,57 @@ def build_plots(df: pd.DataFrame) -> list:
             )
             figs.append(fig)
 
-    return figs
+    return figs, skipped_metrics
+
+
+def render_skipped_metrics(skipped_metrics: list[dict[str, object]]) -> str:
+    if not skipped_metrics:
+        return ""
+
+    rows = [
+        "<h3>Skipped Metric Plots</h3>",
+        "<table>",
+        ("<thead><tr><th>Model</th><th>GPU</th><th>Metric</th>"
+         "<th>Records</th><th>Non-null</th><th>Reason</th></tr></thead>"),
+        "<tbody>",
+    ]
+    for item in skipped_metrics:
+        rows.append(
+            "<tr>"
+            f"<td>{escape(str(item['model_id']))}</td>"
+            f"<td>{escape(str(item['gpu_type']))}</td>"
+            f"<td>{escape(str(item['metric']))}</td>"
+            f"<td>{item['records']}</td>"
+            f"<td>{item['non_null']}</td>"
+            f"<td>{escape(str(item['reason']))}</td>"
+            "</tr>"
+        )
+    rows.extend(["</tbody>", "</table>"])
+    return "\n".join(rows)
 
 # -----------------------------
 # 3. Render HTML dashboard
 # -----------------------------
-def render_html(figs: list, days: int) -> str:
+def render_html(figs: list, skipped_metrics: list[dict[str, object]],
+                days: int) -> str:
     html_parts = [
         "<html>",
         "<head><meta charset='utf-8'>",
-        "<style>body { font-family: sans-serif; margin: 2rem; }</style>",
+        ("<style>"
+         "body { font-family: sans-serif; margin: 2rem; }"
+         "table { border-collapse: collapse; margin: 1rem 0 2rem; }"
+         "th, td { border: 1px solid #ddd; padding: 0.4rem 0.6rem; "
+         "text-align: left; }"
+         "th { background: #f5f5f5; }"
+         "</style>"),
         "</head><body>",
         f"<h2>Performance Dashboard (last {days} days)</h2>",
     ]
 
+    skipped_html = render_skipped_metrics(skipped_metrics)
+    if skipped_html:
+        html_parts.append(skipped_html)
+
     for fig in figs:
         html_parts.append(fig.to_html(full_html=False, include_plotlyjs="cdn"))
 
@@ -67,7 +135,7 @@ def render_html(figs: list, days: int) -> str:
 def main() -> None:
     days = int(os.environ.get("DASHBOARD_DAYS", "30"))
 
-    local_dir = sync_from_hf("/tmp/perf-tracking")
+    local_dir = sync_from_hf("/tmp/perf-tracking", reuse_existing=True)
     df = load_as_dataframe(local_dir, days=days)
 
     if df.empty:
@@ -79,8 +147,15 @@ def main() -> None:
           f"{df['gpu_type'].nunique()} GPU type(s), "
           f"date range: {df['timestamp'].min()} → {df['timestamp'].max()}")
 
-    figs = build_plots(df)
-    html = render_html(figs, days)
+    figs, skipped_metrics = build_plots(df)
+    if skipped_metrics:
+        print("Skipped metric plots:")
+        for item in skipped_metrics:
+            print(f"  - {item['model_id']} | {item['gpu_type']} | "
+                  f"{item['metric']}: {item['reason']} "
+                  f"({item['non_null']}/{item['records']} non-null)")
+    print(f"Generated {len(figs)} metric plot(s)")
+    html = render_html(figs, skipped_metrics, days)
 
     commit_sha = os.environ.get("BUILDKITE_COMMIT", "unknown")[:7]
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
@@ -97,4 +172,4 @@ def main() -> None:
     print(f"Dashboard generated: {output_file}")
 
 if __name__ == "__main__":
-    main()
+    main()

fastvideo/tests/performance/hf_store.py

@@ -25,6 +25,7 @@ from huggingface_hub import HfApi, snapshot_download
 
 HF_REPO_ID: str = os.environ.get("HF_REPO_ID", "FastVideo/performance-tracking")
 HF_TOKEN: str | None = os.environ.get("HF_API_KEY")
+SYNC_MARKER = ".hf_sync_complete"
 
 # ---------------------------------------------------------------------------
 # Low-level helpers
@@ -51,7 +52,16 @@ def safe_float(value: Any) -> float | None:
 # ---------------------------------------------------------------------------
 
 
-def sync_from_hf(local_dir: str, *, strict: bool = False) -> str:
+def _sync_marker_path(local_dir: str) -> str:
+    return os.path.join(local_dir, SYNC_MARKER)
+
+
+def sync_from_hf(
+    local_dir: str,
+    *,
+    strict: bool = False,
+    reuse_existing: bool = False,
+) -> str:
     """Download the HF dataset repo snapshot to *local_dir*.
 
     Returns *local_dir* so callers can chain: ``load_records(sync_from_hf(...))``.
@@ -61,7 +71,19 @@ def sync_from_hf(local_dir: str, *, strict: bool = False) -> str:
     unavailable. Callers that depend on the sync for correctness (e.g. the
     main-branch baseline writer) must pass ``strict=True`` so that misconfig
     or transient HF errors fail loud rather than silently reset the baseline.
+
+    When ``reuse_existing=True``, a previous successful sync in ``local_dir``
+    is reused. This avoids duplicate HF snapshot checks when compare and
+    dashboard scripts run sequentially in the same CI job.
     """
+    marker_path = _sync_marker_path(local_dir)
+    if reuse_existing and os.path.exists(marker_path):
+        print(f"hf_store: reusing existing sync at {local_dir}")
+        return local_dir
+
+    if not reuse_existing and os.path.exists(marker_path):
+        os.remove(marker_path)
+
     if not HF_REPO_ID:
         msg = "hf_store: HF_REPO_ID not set"
         if strict:
@@ -78,6 +100,12 @@ def sync_from_hf(local_dir: str, *, strict: bool = False) -> str:
             token=HF_TOKEN,
             allow_patterns="*.json",
         )
+        os.makedirs(local_dir, exist_ok=True)
+        with open(marker_path, "w", encoding="utf-8") as marker:
+            json.dump({
+                "repo_id": HF_REPO_ID,
+                "synced_at": datetime.now(timezone.utc).isoformat(),
+            }, marker)
     except Exception as exc:
         if strict:
             raise
@@ -223,14 +251,18 @@ def load_records_for_model(
 # DataFrame helpers (dashboard / analytics consumers)
 # ---------------------------------------------------------------------------
 
-_NUMERIC_COLS = ("latency", "throughput", "memory")
+_NUMERIC_COLS = (
+    "latency", "throughput", "memory",
+    "text_encoder_time_s", "dit_time_s", "vae_decode_time_s",
+)
 
 
 def normalize_dataframe(df: pd.DataFrame) -> pd.DataFrame:
     """Apply standard type coercions to a raw records DataFrame.
 
     - Parses ``timestamp`` to UTC-aware datetime.
-    - Coerces ``latency``, ``throughput``, ``memory`` to float.
+    - Coerces ``latency``, ``throughput``, ``memory``, ``text_encoder_time_s``,
+      ``dit_time_s``, ``vae_decode_time_s`` to float.
     - Adds a ``config_id`` column (first 7 chars of ``commit_sha``).
 
     Returns the mutated DataFrame (also modifies in place for efficiency).

fastvideo/tests/performance/test_inference_performance.py

@@ -10,17 +10,34 @@ import glob
 import json
 import os
 import time
+from collections.abc import Mapping
 from datetime import datetime, timezone
 
 import torch
 import pytest
 
+import fastvideo.envs as envs
 from fastvideo import VideoGenerator
 from fastvideo.logger import init_logger
 from fastvideo.worker.multiproc_executor import MultiprocExecutor
 
 logger = init_logger(__name__)
 
+STAGE_METRIC_MAP: dict[str, str] = {
+    "TextEncoderStage": "text_encoder_time_s",
+    "TextEncodingStage": "text_encoder_time_s",
+    "CLIPTextEncoderStage": "text_encoder_time_s",
+    "T5TextEncoderStage": "text_encoder_time_s",
+    "DiTStage": "dit_time_s",
+    "DenoisingStage": "dit_time_s",
+    "DmdDenoisingStage": "dit_time_s",
+    "TransformerStage": "dit_time_s",
+    "VAEDecodeStage": "vae_decode_time_s",
+    "VAEDecoderStage": "vae_decode_time_s",
+    "DecodingStage": "vae_decode_time_s",
+    "DecodeStage": "vae_decode_time_s",
+}
+
 # -- Config discovery -------------------------------------------------------
 
 _BENCHMARKS_DIR = os.path.join(
@@ -73,16 +90,53 @@ def _shutdown_executor(generator):
         generator.executor.shutdown()
 
 
+def _extract_component_times(result: dict) -> dict[str, float | None]:
+    component_times: dict[str, float | None] = {
+        "text_encoder_time_s": None,
+        "dit_time_s": None,
+        "vae_decode_time_s": None,
+    }
+    logging_info = result.get("logging_info")
+    if logging_info is None:
+        return component_times
+    if isinstance(logging_info, Mapping):
+        stages: dict = logging_info.get("stages", {}) or {}
+    else:
+        stages: dict = getattr(logging_info, "stages", {}) or {}
+    if not stages:
+        return component_times
+    logger.info("Discovered pipeline stages: %s", list(stages.keys()))
+    for stage_name, stage_data in stages.items():
+        metric_key = STAGE_METRIC_MAP.get(stage_name)
+        if metric_key is None:
+            logger.debug("Unmapped stage '%s' (%.3fs)",
+                         stage_name,
+                         stage_data.get("execution_time", 0))
+            continue
+        elapsed = stage_data.get("execution_time")
+        if elapsed is None:
+            continue
+        existing = component_times[metric_key]
+        component_times[metric_key] = (
+            elapsed if existing is None else existing + elapsed)
+    return component_times
+
+
+def _avg_component(all_component_times: list[dict], key: str) -> float | None:
+    vals = [r[key] for r in all_component_times if r.get(key) is not None]
+    return round(sum(vals) / len(vals), 3) if vals else None
+
+
 def _run_generation(generator, prompt, generation_kwargs):
-    """Run a single generation, return (elapsed_s, peak_memory_mb)."""
+    """Run a single generation, return (elapsed_s, peak_memory_mb, component_times)."""
     torch.cuda.synchronize()
     start = time.perf_counter()
     result = generator.generate_video(prompt, **generation_kwargs)
     torch.cuda.synchronize()
     elapsed = time.perf_counter() - start
     peak_memory_mb = result.get("peak_memory_mb", 0.0) or 0.0
-    return elapsed, peak_memory_mb
-
+    component_times = _extract_component_times(result)
+    return elapsed, peak_memory_mb, component_times
 
 def _write_results(results):
     """Write JSON results to the results directory."""
@@ -102,15 +156,7 @@ def _write_results(results):
 
 # -- Test -------------------------------------------------------------------
 
-
-@pytest.mark.parametrize(
-    "cfg",
-    _BENCHMARK_CONFIGS,
-    ids=[c["benchmark_id"] for c in _BENCHMARK_CONFIGS],
-)
-def test_inference_performance(cfg):
-    """Measure generation latency and peak GPU memory,
-    assert against device-aware thresholds."""
+def _run_benchmark(cfg):
     run_config = cfg.get("run_config", {})
     required_gpus = run_config.get("required_gpus", 1)
     available = torch.cuda.device_count()
@@ -152,12 +198,18 @@ def test_inference_performance(cfg):
 
         times = []
         peak_memories = []
+        all_component_times = []
         for i in range(num_measure):
             logger.info("Measurement run %d/%d", i + 1, num_measure)
-            elapsed, peak_mb = _run_generation(generator, prompt, gen_kwargs)
+            elapsed, peak_mb, component_times = _run_generation(
+                generator,
+                prompt,
+                gen_kwargs,
+            )
             logger.info("  Time: %.2fs, Peak memory: %.0fMB", elapsed, peak_mb)
             times.append(elapsed)
             peak_memories.append(peak_mb)
+            all_component_times.append(component_times)
     finally:
         _shutdown_executor(generator)
 
@@ -186,6 +238,11 @@ def test_inference_performance(cfg):
         "commit": os.environ.get("BUILDKITE_COMMIT", ""),
         "pr_number": os.environ.get("BUILDKITE_PULL_REQUEST", ""),
         "timestamp": datetime.now(timezone.utc).isoformat(),
+        "text_encoder_time_s": _avg_component(all_component_times,
+                                              "text_encoder_time_s"),
+        "dit_time_s": _avg_component(all_component_times, "dit_time_s"),
+        "vae_decode_time_s": _avg_component(all_component_times,
+                                            "vae_decode_time_s"),
     }
 
     logger.info(
@@ -203,3 +260,41 @@ def test_inference_performance(cfg):
     assert max_peak_memory <= max_mem, (
         f"Peak memory {max_peak_memory:.0f}MB exceeds "
         f"threshold {max_mem:.0f}MB for {device_name}")
+
+    component_thresholds = {
+        "text_encoder_time_s": thresholds.get("max_text_encoder_time_s"),
+        "dit_time_s": thresholds.get("max_dit_time_s"),
+        "vae_decode_time_s": thresholds.get("max_vae_decode_time_s"),
+    }
+    for metric, max_val in component_thresholds.items():
+        if max_val is None:
+            continue
+        actual = results[metric]
+        if actual is not None:
+            assert actual <= max_val, (
+                f"{metric} {actual:.3f}s exceeds threshold {max_val:.3f}s "
+                f"for {device_name}")
+
+
+@pytest.mark.parametrize(
+    "cfg",
+    _BENCHMARK_CONFIGS,
+    ids=[c["benchmark_id"] for c in _BENCHMARK_CONFIGS],
+)
+def test_inference_performance(cfg):
+    """Measure generation latency, peak GPU memory, and component-level timings
+    (text encoder, DiT, VAE decode). Assert each against device-aware thresholds.
+    """
+
+    original_env = os.environ.get("FASTVIDEO_STAGE_LOGGING")
+    original_getter = envs.environment_variables["FASTVIDEO_STAGE_LOGGING"]
+    os.environ["FASTVIDEO_STAGE_LOGGING"] = "1"
+    envs.environment_variables["FASTVIDEO_STAGE_LOGGING"] = lambda: True
+    try:
+        _run_benchmark(cfg)
+    finally:
+        if original_env is None:
+            os.environ.pop("FASTVIDEO_STAGE_LOGGING", None)
+        else:
+            os.environ["FASTVIDEO_STAGE_LOGGING"] = original_env
+        envs.environment_variables["FASTVIDEO_STAGE_LOGGING"] = original_getter

fastvideo/tests/ssim/inference_similarity_utils.py

@@ -132,7 +132,7 @@ def _assert_similarity(
     )
 
 
-def _build_init_kwargs(
+def build_init_kwargs(
     base_params: dict[str, object],
 ) -> dict[str, object]:
     init_kwargs: dict[str, object] = {
@@ -167,7 +167,7 @@ def _build_init_kwargs(
     return init_kwargs
 
 
-def _build_generation_kwargs(
+def build_generation_kwargs(
     base_params: dict[str, object],
     num_inference_steps: int,
     output_dir: str,
@@ -222,11 +222,11 @@ def run_text_to_video_similarity_test(
         base_params = params_map[model_id]
         num_inference_steps = int(base_params["num_inference_steps"])
 
-        init_kwargs = _build_init_kwargs(base_params)
+        init_kwargs = build_init_kwargs(base_params)
         if init_kwargs_override:
             init_kwargs.update(init_kwargs_override)
 
-        generation_kwargs = _build_generation_kwargs(
+        generation_kwargs = build_generation_kwargs(
             base_params,
             num_inference_steps,
             output_dir,
@@ -297,11 +297,11 @@ def run_image_to_video_similarity_test(
         base_params = params_map[model_id]
         num_inference_steps = int(base_params["num_inference_steps"])
 
-        init_kwargs = _build_init_kwargs(base_params)
+        init_kwargs = build_init_kwargs(base_params)
         if init_kwargs_override:
             init_kwargs.update(init_kwargs_override)
 
-        generation_kwargs = _build_generation_kwargs(
+        generation_kwargs = build_generation_kwargs(
             base_params,
             num_inference_steps,
             output_dir,

fastvideo/tests/ssim/latent_similarity_utils.py

@@ -0,0 +1,494 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Latent-space regression helpers for numerically fragile SSIM tests.
+
+Motivation
+----------
+Pixel-space SSIM is a poor regression signal for distilled / few-step
+models (e.g. LTX-2 distilled): a single mis-rounded bf16 accumulator in
+the VAE decoder can drive mean SSIM from ~0.95 to ~0.50 without any real
+quality regression.
+
+Inspired by diffusers' "small signature slice + bounded full-tensor
+distance" testing philosophy, applied here to the **pre-VAE latent**
+rather than the decoded pixel/audio output:
+
+* `tests/pipelines/ltx2/test_ltx2.py` (diffusers) compares
+  ``output_type='pt'`` (pixel) slices via
+  ``torch.allclose(generated_slice, expected_slice, atol=1e-4)``;
+* `tests/pipelines/stable_audio/test_stable_audio.py` (diffusers)
+  compares decoded audio samples via
+  ``np.abs(expected - actual).max() < 1.5e-3``;
+* `tests/pipelines/cogvideo/test_cogvideox.py` (diffusers) compares
+  full pixel video tensors via
+  ``numpy_cosine_similarity_distance(...) < 1e-3``.
+
+Diffusers does *not* assert on latents directly — that is a FastVideo
+adaptation. Distilled few-step pipelines amplify per-step bf16 noise
+enough that VAE-decoded comparisons are unreliable across our
+heterogeneous CI pool, so we move the assertion upstream of the VAE.
+
+Design
+------
+* Inference is run with ``output_type='latent'`` so ``DecodingStage``
+  hands back the un-decoded latent on ``result["samples"]``.
+* The reference artefact is a ``.pt`` bundle (tensor + metadata) hosted
+  on the same HF dataset as the mp4 references, selected by
+  ``<GPU>_reference_videos/<model_id>/<backend>/<prompt>.pt``.
+* Two assertions are performed:
+    1. A small signature slice (default video: ``latent[0, :, 0, :3, :3]``;
+       audio: ``latent[0, :, :8]``) is compared via cosine distance with
+       a loose tolerance. Primary pass/fail gate.
+    2. The full latent is compared via cosine distance with a slightly
+       tighter tolerance, guarding against shape-correct but globally
+       drifted outputs.
+* Tolerances default to 5e-3 (slice) and 1e-2 (full). diffusers uses
+  ``1e-3`` against deterministic CPU dummy components; we relax for
+  cross-GPU-arch bf16 differences on the rented CI pool
+  (A40/L40S/H100/B200).
+
+The helper intentionally reuses ``build_init_kwargs`` /
+``build_generation_kwargs`` from :mod:`inference_similarity_utils` so
+model params (vae tiling, sp_size, flow shift, …) flow through a single
+source of truth.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+from logging import Logger
+from typing import Any
+
+import torch
+from torch.nn.functional import cosine_similarity
+
+from fastvideo import VideoGenerator
+from fastvideo.tests.ssim.inference_similarity_utils import (
+    attention_backend,
+    build_generation_kwargs,
+    build_init_kwargs,
+    shutdown_executor,
+)
+from fastvideo.tests.ssim.reference_utils import (
+    build_generated_output_dir,
+    build_reference_folder_path,
+    select_ssim_params,
+)
+
+LATENT_REFERENCE_EXTENSION = ".pt"
+LATENT_REFERENCE_FORMAT_VERSION = 1
+
+DEFAULT_SLICE_SPEC: dict[str, Any] = {
+    "kind": "corner_3x3_first_frame",
+    "version": 1,
+}
+
+AUDIO_FIRST_8_TIMESTEPS_SPEC: dict[str, Any] = {
+    "kind": "audio_first_8_timesteps",
+    "version": 1,
+}
+
+
+def _extract_expected_slice(
+    latent: torch.Tensor,
+    spec: dict[str, Any],
+) -> torch.Tensor:
+    """Return a 1-D fp32 signature slice from ``latent`` per ``spec``.
+
+    Dispatch by ``spec["kind"]``:
+
+    * ``"corner_3x3_first_frame"`` — 5-D video latent ``[B, C, T, H, W]``;
+      returns ``latent[0, :, 0, :3, :3]`` flattened (length = C * 9).
+    * ``"audio_first_8_timesteps"`` — 3-D audio latent ``[B, C, T]``;
+      returns ``latent[0, :, :8]`` flattened (length = C * 8).
+    """
+    kind = spec.get("kind", "corner_3x3_first_frame")
+    if kind == "corner_3x3_first_frame":
+        if latent.dim() != 5:
+            raise ValueError(
+                f"corner_3x3_first_frame requires 5-D latent [B,C,T,H,W]; "
+                f"got shape {tuple(latent.shape)}")
+        _, _, t, h, w = latent.shape
+        if t < 1 or h < 3 or w < 3:
+            raise ValueError(
+                "corner_3x3_first_frame requires T>=1, H>=3, W>=3; got "
+                f"shape {tuple(latent.shape)}")
+        return (latent[0, :, 0, :3, :3]
+                .detach()
+                .to(torch.float32)
+                .reshape(-1)
+                .contiguous())
+    if kind == "audio_first_8_timesteps":
+        if latent.dim() != 3:
+            raise ValueError(
+                f"audio_first_8_timesteps requires 3-D latent [B,C,T]; "
+                f"got shape {tuple(latent.shape)}")
+        _, _, t = latent.shape
+        if t < 8:
+            raise ValueError(
+                "audio_first_8_timesteps requires T>=8; got "
+                f"shape {tuple(latent.shape)}")
+        return (latent[0, :, :8]
+                .detach()
+                .to(torch.float32)
+                .reshape(-1)
+                .contiguous())
+    raise ValueError(f"Unknown slice kind: {kind!r}")
+
+
+def _cosine_distance(a: torch.Tensor, b: torch.Tensor) -> float:
+    """Return ``1 - cos(a, b)`` as a Python float, operating in fp32."""
+    a32 = a.detach().to(torch.float32).reshape(-1)
+    b32 = b.detach().to(torch.float32).reshape(-1)
+    if a32.shape != b32.shape:
+        raise ValueError(
+            f"Cosine shape mismatch: {tuple(a32.shape)} vs {tuple(b32.shape)}"
+        )
+    sim = cosine_similarity(a32.unsqueeze(0), b32.unsqueeze(0), dim=1).item()
+    return 1.0 - float(sim)
+
+
+def save_latent_reference(
+    path: str,
+    latent: torch.Tensor,
+    *,
+    metadata: dict[str, Any],
+    slice_spec: dict[str, Any] | None = None,
+) -> None:
+    """Persist a latent bundle to ``path``.
+
+    Storage format (dict pickled via ``torch.save``):
+
+    * ``latent``: full latent as fp16 on cpu
+    * ``shape``: original shape (list)
+    * ``dtype_original``: str
+    * ``expected_slice``: fp32 1-D signature slice
+    * ``slice_spec``: dict describing how the slice was built
+    * ``metadata``: caller-provided context (prompt, backend, steps, …)
+    * ``format_version``: int
+
+    fp16 is lossy but bounded; it keeps ref artefacts small (~a few MB
+    per prompt) while preserving enough dynamic range for cosine-based
+    regression. Slice values stay fp32 because the primary assertion is
+    computed against them.
+    """
+    spec = slice_spec if slice_spec is not None else DEFAULT_SLICE_SPEC
+    parent = os.path.dirname(path)
+    if parent:
+        os.makedirs(parent, exist_ok=True)
+
+    latent_cpu = latent.detach().to("cpu")
+    payload: dict[str, Any] = {
+        "latent": latent_cpu.to(torch.float16),
+        "shape": list(latent_cpu.shape),
+        "dtype_original": str(latent_cpu.dtype),
+        "expected_slice": _extract_expected_slice(latent_cpu, spec),
+        "slice_spec": spec,
+        "metadata": metadata,
+        "format_version": LATENT_REFERENCE_FORMAT_VERSION,
+    }
+    torch.save(payload, path)
+
+
+def load_latent_reference(path: str) -> dict[str, Any]:
+    """Inverse of :func:`save_latent_reference` — always loads to cpu.
+
+    Enforces ``format_version == LATENT_REFERENCE_FORMAT_VERSION`` so a
+    schema change forces a deliberate reseed instead of silently
+    misinterpreting old artefacts.
+    """
+    # ``weights_only=False`` is required because the payload is a dict of
+    # tensors + plain-Python metadata (slice_spec, prompt, ...). The trust
+    # boundary is the controlled HF dataset configured via
+    # FASTVIDEO_SSIM_REFERENCE_HF_REPO (default
+    # FastVideo/ssim-reference-videos), which is org-write-gated.
+    payload = torch.load(path, map_location="cpu", weights_only=False)
+    fmt = payload.get("format_version") if isinstance(payload, dict) else None
+    if fmt != LATENT_REFERENCE_FORMAT_VERSION:
+        raise ValueError(
+            f"Latent reference at {path!r} has format_version={fmt!r}; "
+            f"expected {LATENT_REFERENCE_FORMAT_VERSION}. Re-seed via the "
+            "test that produces this artefact, then re-upload through "
+            "fastvideo/tests/ssim/reference_videos_cli.py upload.")
+    return payload
+
+
+def write_latent_similarity_results(
+    output_dir: str,
+    metrics: dict[str, float],
+    *,
+    reference_path: str,
+    generated_path: str,
+    num_inference_steps: int,
+    prompt: str,
+    model_id: str,
+    attention_backend_name: str,
+    slice_spec: dict[str, Any],
+    slice_cosine_threshold: float,
+    full_cosine_threshold: float,
+    passed: bool,
+) -> bool:
+    """Persist latent regression metrics next to the generated artefact.
+
+    Mirrors :func:`fastvideo.tests.utils.write_ssim_results` so downstream
+    CI tooling can scrape one schema for both pixel and latent runs.
+    The filename is ``steps{N}_{prompt[:100]}_latent.json``.
+    """
+    try:
+        os.makedirs(output_dir, exist_ok=True)
+        prompt_prefix = prompt[:100].strip()
+        filename = f"steps{num_inference_steps}_{prompt_prefix}_latent.json"
+        target = os.path.join(output_dir, filename)
+        payload = {
+            "metrics": metrics,
+            "reference_latent": reference_path,
+            "generated_latent": generated_path,
+            "model_id": model_id,
+            "attention_backend": attention_backend_name,
+            "slice_spec": slice_spec,
+            "thresholds": {
+                "slice_cosine": slice_cosine_threshold,
+                "full_cosine": full_cosine_threshold,
+            },
+            "passed": passed,
+            "parameters": {
+                "num_inference_steps": num_inference_steps,
+                "prompt": prompt,
+            },
+        }
+        with open(target, "w", encoding="utf-8") as handle:
+            json.dump(payload, handle, indent=2, sort_keys=True)
+        return True
+    except OSError:
+        return False
+
+
+def _assert_latent_similarity(
+    *,
+    logger: Logger,
+    gen_latent: torch.Tensor,
+    reference_path: str,
+    slice_cosine_threshold: float,
+    full_cosine_threshold: float,
+    model_id: str,
+    attention_backend_name: str,
+    output_dir: str,
+    generated_path: str,
+    num_inference_steps: int,
+    prompt: str,
+) -> dict[str, float]:
+    ref = load_latent_reference(reference_path)
+    expected_slice = ref["expected_slice"]
+    ref_full = ref["latent"].to(torch.float32)
+    ref_shape = tuple(ref.get("shape", ref_full.shape))
+    slice_spec = ref.get("slice_spec", DEFAULT_SLICE_SPEC)
+
+    if tuple(gen_latent.shape) != ref_shape:
+        raise AssertionError(
+            f"Generated latent shape {tuple(gen_latent.shape)} does not "
+            f"match reference shape {ref_shape} for {model_id} with "
+            f"backend {attention_backend_name}")
+
+    gen_slice = _extract_expected_slice(gen_latent, slice_spec)
+    slice_cos = _cosine_distance(gen_slice, expected_slice)
+    # The reference full tensor was fp16-quantized at seed time. Round-trip
+    # the generated tensor through fp16 so both sides share the same
+    # quantization floor and the cosine distance is symmetric. This does
+    # NOT affect ``slice_cos`` because the reference slice is persisted in
+    # fp32 (see :func:`save_latent_reference`).
+    gen_full_matched = gen_latent.to(torch.float16).to(torch.float32)
+    full_cos = _cosine_distance(gen_full_matched, ref_full)
+    max_abs_diff = float(
+        (gen_full_matched - ref_full).abs().max().item())
+
+    metrics: dict[str, float] = {
+        "slice_cosine_distance": slice_cos,
+        "full_cosine_distance": full_cos,
+        "max_abs_diff": max_abs_diff,
+    }
+    logger.info(
+        "Latent regression metrics for %s/%s: %s",
+        model_id,
+        attention_backend_name,
+        metrics,
+    )
+
+    failures: list[str] = []
+    if slice_cos > slice_cosine_threshold:
+        failures.append(
+            f"slice cosine {slice_cos:.6e} > threshold "
+            f"{slice_cosine_threshold:.6e}")
+    if full_cos > full_cosine_threshold:
+        failures.append(
+            f"full cosine {full_cos:.6e} > threshold "
+            f"{full_cosine_threshold:.6e}")
+
+    passed = not failures
+    write_latent_similarity_results(
+        output_dir,
+        metrics,
+        reference_path=reference_path,
+        generated_path=generated_path,
+        num_inference_steps=num_inference_steps,
+        prompt=prompt,
+        model_id=model_id,
+        attention_backend_name=attention_backend_name,
+        slice_spec=slice_spec,
+        slice_cosine_threshold=slice_cosine_threshold,
+        full_cosine_threshold=full_cosine_threshold,
+        passed=passed,
+    )
+
+    if failures:
+        raise AssertionError(
+            f"Latent regression exceeded tolerance for {model_id} with "
+            f"backend {attention_backend_name}: {'; '.join(failures)}. "
+            f"Full metrics: {metrics}")
+
+    return metrics
+
+
+def _extract_latent_from_result(result: Any) -> torch.Tensor:
+    """Pull a fp32 cpu latent tensor (5-D video or 3-D audio) from
+    ``generate_video`` output.
+    """
+    if not isinstance(result, dict):
+        raise RuntimeError(
+            "VideoGenerator.generate_video returned unexpected payload "
+            f"(type={type(result)!r}); expected dict with 'samples'.")
+    samples = result.get("samples")
+    if samples is None:
+        raise RuntimeError(
+            "VideoGenerator did not return latent samples. Ensure "
+            "output_type='latent' and return_frames=True for this call.")
+    if not isinstance(samples, torch.Tensor):
+        raise RuntimeError(
+            f"Expected torch.Tensor samples; got type={type(samples)!r}.")
+    gen_latent = samples.detach().to(torch.float32).cpu()
+    if gen_latent.dim() not in (3, 5):
+        raise RuntimeError(
+            "Expected 5-D video latent (B,C,T,H,W) or 3-D audio latent "
+            f"(B,C,T); got shape {tuple(gen_latent.shape)}")
+    return gen_latent
+
+
+def run_text_to_latent_similarity_test(
+    *,
+    logger: Logger,
+    script_dir: str,
+    device_reference_folder: str,
+    prompt: str,
+    attention_backend_name: str,
+    model_id: str,
+    default_params_map: dict[str, dict[str, object]],
+    full_quality_params_map: dict[str, dict[str, object]],
+    slice_cosine_threshold: float = 5e-3,
+    full_cosine_threshold: float = 1e-2,
+    init_kwargs_override: dict[str, object] | None = None,
+    generation_kwargs_override: dict[str, object] | None = None,
+    slice_spec: dict[str, Any] | None = None,
+) -> dict[str, float]:
+    """Run T2V (or T2A) inference with ``output_type='latent'`` and
+    compare to a reference latent.
+
+    Returns the computed metrics dict on success. Raises
+    ``AssertionError`` if any cosine tolerance is exceeded and
+    ``FileNotFoundError`` if the reference artefact is missing.
+    """
+    spec = slice_spec if slice_spec is not None else DEFAULT_SLICE_SPEC
+    with attention_backend(attention_backend_name):
+        output_dir = build_generated_output_dir(
+            script_dir,
+            device_reference_folder,
+            model_id,
+            attention_backend_name,
+        )
+        prompt_prefix = prompt[:100].strip()
+        output_latent_name = f"{prompt_prefix}{LATENT_REFERENCE_EXTENSION}"
+        os.makedirs(output_dir, exist_ok=True)
+
+        params_map = select_ssim_params(
+            default_params_map,
+            full_quality_params_map,
+        )
+        base_params = params_map[model_id]
+        num_inference_steps = int(base_params["num_inference_steps"])
+
+        init_kwargs = build_init_kwargs(base_params)
+        if init_kwargs_override:
+            init_kwargs.update(init_kwargs_override)
+        # Always wins: the helper exists specifically to compare on latents,
+        # so an override can never silently turn it back into a pixel run.
+        init_kwargs["output_type"] = "latent"
+
+        generation_kwargs = build_generation_kwargs(
+            base_params,
+            num_inference_steps,
+            output_dir,
+        )
+        # We serialize latents ourselves; skip the RGB encoder path.
+        generation_kwargs["save_video"] = False
+        generation_kwargs["return_frames"] = True
+        if generation_kwargs_override:
+            generation_kwargs.update(generation_kwargs_override)
+
+        generator: VideoGenerator | None = None
+        try:
+            generator = VideoGenerator.from_pretrained(
+                model_path=base_params["model_path"],
+                **init_kwargs,
+            )
+            result = generator.generate_video(prompt, **generation_kwargs)
+        finally:
+            shutdown_executor(generator)
+
+    gen_latent = _extract_latent_from_result(result)
+
+    generated_latent_path = os.path.join(output_dir, output_latent_name)
+    save_latent_reference(
+        generated_latent_path,
+        gen_latent,
+        metadata={
+            "prompt": prompt,
+            "model_id": model_id,
+            "attention_backend": attention_backend_name,
+            "num_inference_steps": num_inference_steps,
+        },
+        slice_spec=spec,
+    )
+    logger.info("Saved generated latent to %s", generated_latent_path)
+
+    reference_folder = build_reference_folder_path(
+        script_dir,
+        device_reference_folder,
+        model_id,
+        attention_backend_name,
+    )
+    if not os.path.exists(reference_folder):
+        raise FileNotFoundError(
+            f"Reference folder does not exist: {reference_folder}\n"
+            f"To download references, run:\n"
+            f"  python fastvideo/tests/ssim/reference_videos_cli.py download")
+
+    reference_latent_path = os.path.join(
+        reference_folder,
+        output_latent_name,
+    )
+    if not os.path.exists(reference_latent_path):
+        raise FileNotFoundError(
+            "Reference latent missing for prompt/backend: "
+            f"{reference_latent_path}")
+
+    return _assert_latent_similarity(
+        logger=logger,
+        gen_latent=gen_latent,
+        reference_path=reference_latent_path,
+        slice_cosine_threshold=slice_cosine_threshold,
+        full_cosine_threshold=full_cosine_threshold,
+        model_id=model_id,
+        attention_backend_name=attention_backend_name,
+        output_dir=output_dir,
+        generated_path=generated_latent_path,
+        num_inference_steps=num_inference_steps,
+        prompt=prompt,
+    )

fastvideo/tests/ssim/reference_videos_cli.py

@@ -13,6 +13,12 @@ from pathlib import Path
 
 
 VIDEO_EXTENSIONS = (".mp4", ".avi", ".mov", ".mkv", ".webm", ".flv")
+# Additional artefact types stored under the same reference folders. Latent
+# tensors (`.pt`) back the latent-slice regression tests used by flaky
+# pixel-space models (e.g. LTX-2 distilled). Keeping them in the same upload
+# flow means seeding a new test only requires one HF round-trip.
+LATENT_EXTENSIONS = (".pt",)
+REFERENCE_EXTENSIONS = VIDEO_EXTENSIONS + LATENT_EXTENSIONS
 HF_TOKEN_ENV_KEYS = ("HF_API_KEY", "HUGGINGFACE_HUB_TOKEN", "HF_TOKEN")
 HF_REPO_ENV_KEY = "FASTVIDEO_SSIM_REFERENCE_HF_REPO"
 HF_REPO_TYPE_ENV_KEY = "FASTVIDEO_SSIM_REFERENCE_HF_REPO_TYPE"
@@ -42,9 +48,14 @@ def _default_repo_type() -> str:
     return os.environ.get(HF_REPO_TYPE_ENV_KEY, DEFAULT_REPO_TYPE)
 
 
-def _iter_video_files(root: Path) -> Iterable[Path]:
+def _iter_reference_files(root: Path) -> Iterable[Path]:
+    """Yield video and latent (.pt) references under `root`.
+
+    Used by copy-local and the "has local references" marker probe so that
+    latent-only tests (no mp4) still satisfy readiness checks.
+    """
     for path in root.rglob("*"):
-        if path.is_file() and path.suffix.lower() in VIDEO_EXTENSIONS:
+        if path.is_file() and path.suffix.lower() in REFERENCE_EXTENSIONS:
             yield path
 
 
@@ -139,7 +150,7 @@ def _has_local_reference_videos(base_dir: Path, quality_tier: str) -> bool:
         return False
     tier_root = _reference_tier_root(base_dir, quality_tier)
     for ref_dir in _discover_reference_dirs(tier_root):
-        for _ in _iter_video_files(ref_dir):
+        for _ in _iter_reference_files(ref_dir):
             return True
     return False
 
@@ -173,14 +184,14 @@ def copy_generated_to_reference(
         raise FileNotFoundError(f"Generated directory not found: {generated_dir}")
 
     copied = 0
-    for video_file in _iter_video_files(generated_dir):
-        rel = video_file.relative_to(generated_dir)
+    for ref_file in _iter_reference_files(generated_dir):
+        rel = ref_file.relative_to(generated_dir)
         dst = reference_dir / rel
         if dry_run:
-            print(f"[dry-run] {video_file} -> {dst}")
+            print(f"[dry-run] {ref_file} -> {dst}")
         else:
             dst.parent.mkdir(parents=True, exist_ok=True)
-            shutil.copy2(video_file, dst)
+            shutil.copy2(ref_file, dst)
             print(f"Copied: {rel}")
         copied += 1
     return copied

fastvideo/tests/ssim/test_ltx2_similarity.py

@@ -1,11 +1,25 @@
 # SPDX-License-Identifier: Apache-2.0
-"""SSIM-based similarity test for LTX-2 distilled text-to-video.
+"""Latent-slice regression test for LTX-2 distilled text-to-video.
 
-Parameters derived from examples/inference/basic/basic_ltx2_distilled.py,
-with resolution + num_inference_steps reduced to keep GPU CI runtime
-bounded. Full-quality variant (via ``--ssim-full-quality``) falls back
-to the ``ltx2_distilled`` preset defaults.
+Pixel-space SSIM is not a useful signal for this model: 4 distilled
+steps + bf16 attention + tiled VAE decode produce outputs that pass
+visual QA but occupy a very wide region in pixel space.
+
+Inspired by diffusers' slice-vs-full regression philosophy — see
+``diffusers/tests/pipelines/ltx2/test_ltx2.py`` (compares pixel slices
+via ``torch.allclose(..., atol=1e-4)``) and
+``diffusers/tests/pipelines/cogvideo/test_cogvideox.py`` (full pixel
+tensors via ``numpy_cosine_similarity_distance(...) < 1e-3``).
+Diffusers itself does NOT compare latents; we apply the same "small
+signature slice + bounded full-tensor distance" idea to the **pre-VAE
+latent** because distilled few-step pipelines amplify per-step bf16
+noise enough that VAE-decoded comparisons are unreliable.
+
+Parameters are kept identical to the original SSIM run so that
+reference artefacts generated on Modal L40S remain bit-compatible with
+production inference.
 """
+
 import os
 
 import pytest
@@ -14,7 +28,9 @@ from fastvideo.api.sampling_param import SamplingParam
 from fastvideo.logger import init_logger
 from fastvideo.tests.ssim.inference_similarity_utils import (
     resolve_inference_device_reference_folder,
-    run_text_to_video_similarity_test,
+)
+from fastvideo.tests.ssim.latent_similarity_utils import (
+    run_text_to_latent_similarity_test,
 )
 
 logger = init_logger(__name__)
@@ -71,6 +87,13 @@ LTX2_DISTILLED_TEST_PROMPTS = [
     "deadpan, absurd, and quietly tragic.",
 ]
 
+# Tolerances chosen on top of diffusers' ``1e-3`` defaults. LTX-2 distilled
+# amplifies per-step numerical noise, and FastVideo's CI pool spans L40S /
+# A40 / H100 so cross-architecture bf16 drift must be absorbed. Values can
+# be tightened after an initial stable window of reference refreshes.
+SLICE_COSINE_DISTANCE_THRESHOLD = 5e-3
+FULL_COSINE_DISTANCE_THRESHOLD = 1e-2
+
 
 @pytest.mark.parametrize("prompt", LTX2_DISTILLED_TEST_PROMPTS)
 @pytest.mark.parametrize("attention_backend_name", ["FLASH_ATTN"])
@@ -80,7 +103,7 @@ def test_ltx2_distilled_inference_similarity(
     attention_backend_name: str,
     model_id: str,
 ) -> None:
-    run_text_to_video_similarity_test(
+    run_text_to_latent_similarity_test(
         logger=logger,
         script_dir=os.path.dirname(os.path.abspath(__file__)),
         device_reference_folder=device_reference_folder,
@@ -89,5 +112,6 @@ def test_ltx2_distilled_inference_similarity(
         model_id=model_id,
         default_params_map=LTX2_DISTILLED_MODEL_TO_PARAMS,
         full_quality_params_map=FULL_QUALITY_LTX2_DISTILLED_MODEL_TO_PARAMS,
-        min_acceptable_ssim=0.60,
+        slice_cosine_threshold=SLICE_COSINE_DISTANCE_THRESHOLD,
+        full_cosine_threshold=FULL_COSINE_DISTANCE_THRESHOLD,
     )

fastvideo/tests/ssim/test_stable_audio_similarity.py

@@ -0,0 +1,111 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Latent-slice regression test for Stable Audio Open 1.0 text-to-audio.
+
+Companion to ``test_ltx2_similarity.py`` — applies the same latent
+cosine-distance philosophy to 3-D audio latents ``[B, 64, T_latent]``.
+
+Why latent-space and not waveform-space SSIM:
+- ``dpmpp-3m-sde`` (k-diffusion) accumulates per-step bf16 noise; the
+  Oobleck VAE then magnifies any residual drift into the time-domain
+  waveform. A few mis-rounded accumulators drive sample-wise diff well
+  past audible thresholds without indicating a real regression.
+- Diffusers' own
+  ``tests/pipelines/stable_audio/test_stable_audio.py`` compares
+  decoded audio samples via
+  ``np.abs(expected - actual).max() < 1.5e-3``; that bound holds for
+  CPU dummy components but does not survive cross-architecture bf16
+  on our CI pool (L40S/A40/H100/B200).
+- Comparing the **pre-VAE latent** moves the assertion upstream of
+  the dominant noise source.
+
+Slice spec: ``audio_first_8_timesteps`` returns
+``latent[0, :, :8]`` (= 64 channels × 8 latent timesteps = 512
+elements). The full latent ``[1, 64, 1024]`` for SA-1.0 is also
+compared via cosine distance.
+"""
+
+import os
+
+import pytest
+
+from fastvideo.api.sampling_param import SamplingParam
+from fastvideo.logger import init_logger
+from fastvideo.tests.ssim.inference_similarity_utils import (
+    resolve_inference_device_reference_folder,
+)
+from fastvideo.tests.ssim.latent_similarity_utils import (
+    AUDIO_FIRST_8_TIMESTEPS_SPEC,
+    run_text_to_latent_similarity_test,
+)
+
+logger = init_logger(__name__)
+
+REQUIRED_GPUS = 1
+
+device_reference_folder = resolve_inference_device_reference_folder(logger)
+
+STABLE_AUDIO_PARAMS = {
+    "num_gpus": 1,
+    "model_path": "FastVideo/stable-audio-open-1.0-Diffusers",
+    # 8/8/1 are the SA-1.0 SamplingParam defaults; the values are
+    # placeholders unused by the audio pipeline but must satisfy the
+    # shared InputValidationStage divisible-by-8 check.
+    "height": 8,
+    "width": 8,
+    "num_frames": 1,
+    "num_inference_steps": 25,
+    "guidance_scale": 7.0,
+    "seed": 1024,
+    "sp_size": 1,
+    "tp_size": 1,
+    "fps": 24,
+}
+
+_SA_FULL_DEFAULTS = SamplingParam.from_pretrained(STABLE_AUDIO_PARAMS["model_path"])
+STABLE_AUDIO_FULL_QUALITY_PARAMS = {
+    **STABLE_AUDIO_PARAMS,
+    "num_inference_steps": _SA_FULL_DEFAULTS.num_inference_steps,
+    "guidance_scale": _SA_FULL_DEFAULTS.guidance_scale,
+    "seed": _SA_FULL_DEFAULTS.seed,
+}
+
+STABLE_AUDIO_MODEL_TO_PARAMS = {
+    "stable-audio-open-1.0-Diffusers": STABLE_AUDIO_PARAMS,
+}
+FULL_QUALITY_STABLE_AUDIO_MODEL_TO_PARAMS = {
+    "stable-audio-open-1.0-Diffusers": STABLE_AUDIO_FULL_QUALITY_PARAMS,
+}
+
+STABLE_AUDIO_TEST_PROMPTS = [
+    "Lo-fi hip hop instrumental with vinyl crackle and gentle piano.",
+]
+
+SLICE_COSINE_DISTANCE_THRESHOLD = 5e-3
+FULL_COSINE_DISTANCE_THRESHOLD = 1e-2
+
+
+@pytest.mark.parametrize("prompt", STABLE_AUDIO_TEST_PROMPTS)
+@pytest.mark.parametrize("attention_backend_name", ["TORCH_SDPA"])
+@pytest.mark.parametrize("model_id", list(STABLE_AUDIO_MODEL_TO_PARAMS.keys()))
+def test_stable_audio_inference_similarity(
+    prompt: str,
+    attention_backend_name: str,
+    model_id: str,
+) -> None:
+    run_text_to_latent_similarity_test(
+        logger=logger,
+        script_dir=os.path.dirname(os.path.abspath(__file__)),
+        device_reference_folder=device_reference_folder,
+        prompt=prompt,
+        attention_backend_name=attention_backend_name,
+        model_id=model_id,
+        default_params_map=STABLE_AUDIO_MODEL_TO_PARAMS,
+        full_quality_params_map=FULL_QUALITY_STABLE_AUDIO_MODEL_TO_PARAMS,
+        slice_cosine_threshold=SLICE_COSINE_DISTANCE_THRESHOLD,
+        full_cosine_threshold=FULL_COSINE_DISTANCE_THRESHOLD,
+        slice_spec=AUDIO_FIRST_8_TIMESTEPS_SPEC,
+        # FSDP + @torch.inference_mode in StableAudioDenoisingStage hits
+        # "Inference tensors do not track version counter" on single-GPU
+        # unshard. SA-1.0 fits on one B200 anyway.
+        init_kwargs_override={"use_fsdp_inference": False},
+    )

pyproject.toml

@@ -134,6 +134,20 @@ test = [
 
 dev = [ "fastvideo[lint]", "fastvideo[test]", ]
 
+prompt-safety = [
+    "fasttext",
+]
+
+prompt-enhancer = [
+    "httpx",
+]
+
+streaming = [
+    "fastvideo[prompt-enhancer]",
+    "fastvideo[prompt-safety]",
+    "websockets",
+]
+
 rocm = [
     "amdsmi",
 ]