Marcel Hofer
e0985b4f7b
feat: improve video prompting and refactor QwenVL
...
Refactor the monolithic QwenVL implementation into focused core modules for model access, Hugging Face runtime and configuration, prompt state, prompt contracts, and media helpers while preserving the public compatibility facade.
Add duration-aware LTX 2.3 and MiniMax H3 prompting, including effective-duration cache identity, MiniMax frame-grid normalization, dedicated T2VA/I2VA/FL2VA/L2VA/reference contracts, and best-effort output normalization.
Improve streamed output handling by separating reasoning from final content, rejecting unfinished thinking traces, sanitizing terminal-only punctuation for legacy Windows code pages, and retaining original Unicode in responses and traces.
Expand preset recommendations, ComfyUI tooltips, system prompts, and documentation for video token budgets, thinking settings, reference roles, dialogue fidelity, and duration behavior.
Keep HF and GGUF consumers on canonical shared imports, preserve mutable export identities across reloads, and extend regression coverage for compatibility, downloads, caching, prompting, streaming, and node registration.
2026-08-10 22:04:05 +02:00
Marcel
753d66fadb
feat: add system prompt preset node ( #9 )
...
Adds a reusable system-prompt preset utility node with validated prompt selection, commercial-loader support, frontend styling, tooltip previews, and regression coverage.
2026-08-08 11:47:46 +02:00
Marcel Hofer
b0144e6ec3
fix: harden MiniMax H3 reference prompting
2026-08-08 11:16:24 +02:00
Marcel Hofer
c448e422c7
fix: harden system prompt preset node
2026-08-08 11:04:07 +02:00
craftingmod
850c96460b
feat: system prompt preset node
2026-08-08 09:10:16 +09:00
Marcel Hofer
d2a0780f58
fix: keep model rights policy commercial-only
2026-08-06 23:19:16 +02:00
Marcel Hofer
89322d6202
release: merge commercial model catalog for v2.3.21
2026-08-06 17:56:37 +02:00
Marcel Hofer
5c6e00cdf4
feat: classify model commercial usage
2026-08-06 17:39:05 +02:00
Marcel
cda5382a5e
feat: split MiniMax H3 presets and log LLM input ( #8 )
2026-08-06 17:29:17 +02:00
Marcel
52803a4ac8
fix: reuse ComfyUI text encoders ( #7 )
...
Adds native reuse of compatible Qwen3-VL and Gemma 3 checkpoints from ComfyUI/models/text_encoders and prepares release v2.3.20.
2026-08-06 17:08:26 +02:00
Marcel Hofer
33dec1631b
fix: reuse ComfyUI text encoders
2026-08-06 16:55:16 +02:00
Your Name
4d22234d46
feat: add MiniMax H3 prompt preset
2026-08-06 00:21:39 +02:00
Marcel Hofer
c38178cdb2
feat: add advanced mask preview output
2026-07-21 09:57:24 +02:00
Marcel Hofer
7daf320d24
feat: add mask-focused image analysis
2026-07-21 09:27:38 +02:00
Marcel Hofer
2c64d28d66
Release 2.3.16 repair Gemma E2B catalog
2026-07-13 12:58:46 +02:00
Marcel Hofer
a03a4a492c
Release 2.3.15 installed GGUF labels
2026-07-12 01:58:43 +02:00
Marcel Hofer
7d1b9c028b
Stabilize Gemma4 GGUF multimodal settings
2026-06-20 17:30:37 +02:00
Marcel Hofer
2b9ca76d41
Pin GGUF llama main GPU to Comfy device
2026-06-20 15:35:18 +02:00
Marcel Hofer
86ac1cee2e
Clamp unsafe Gemma GGUF CPU multimodal load
2026-06-20 15:19:20 +02:00
Marcel Hofer
b8d1c30484
Guard unsafe Gemma GGUF CPU multimodal path
2026-06-20 15:12:14 +02:00
Marcel Hofer
56c18a78ff
Filter Qwen-only handler kwargs for Gemma GGUF
2026-06-20 15:04:56 +02:00
Marcel Hofer
110fe6973a
Reuse mapped GGUF model aliases before download
2026-06-20 15:00:12 +02:00
Marcel Hofer
3b5a6752ec
Fix duplicate ThinkingLLM search entries
2026-06-20 14:47:15 +02:00
Marcel Hofer
f112bafdd7
Respect extra GGUF model paths in prompt enhancer
2026-06-20 11:19:48 +02:00
Marcel Hofer
a58079cee4
Release 2.3.13 recommendation note cleanup
2026-06-20 10:40:32 +02:00
Marcel Hofer
4f7f3905c8
Restore custom prompt image no-preset alias
2026-06-20 09:53:01 +02:00
Marcel Hofer
cdf110e3e5
Preserve ComfyUI execution node colors
2026-06-20 09:37:16 +02:00
Marcel Hofer
52503f95f8
Disable automatic GGUF finalization retries
2026-06-19 19:14:46 +02:00
Marcel Hofer
07b88246c2
Fix Advanced GGUF layout in ComfyUI 2.0
2026-06-19 18:05:15 +02:00
Marcel Hofer
ea25a8726f
Release 2.3.9 preset tooltip previews
2026-06-15 11:15:34 +02:00
Marcel Hofer
21e90d0ba4
Release 2.3.8 LTX preset support
2026-06-15 11:11:15 +02:00
Marcel Hofer
abe35d6dfb
Release 2.3.7 Whisper offline cache fix
2026-06-14 19:36:38 +02:00
Marcel Hofer
028b2df006
Harden GGUF context reset between completions
2026-06-14 07:21:52 +00:00
Marcel Hofer
de5d3e5da3
Release 2.3.5 audio model updates
2026-06-09 00:21:54 +02:00
Marcel Hofer
90df6de975
Add model recommendation guides and Hauhau GGUF support
2026-05-29 19:05:29 +02:00
Marcel Hofer
6faa6bade2
Silence stream-mode status chatter
2026-05-27 00:44:38 +02:00
Marcel Hofer
1189b6c4b3
Fix clean streaming output and guard GGUF loops
2026-05-27 00:32:08 +02:00
Marcel Hofer
c49cbd1ceb
Fix compact streaming display across ThinkingLLM nodes
2026-05-26 23:46:03 +02:00
Marcel Hofer
00f066d34f
Fix GGUF prompt enhancer context crash
2026-05-18 20:50:11 +02:00
Your Name
bcd8f4ae35
v1.0.13: harden GGUF backend diagnostics and downloads
2026-05-17 23:07:22 +02:00
Your Name
a95561e1ae
v1.0.8: Expand Linux wheel compatibility matrix (CUDA 12.8 cp310-cp314) and lock local fork
2026-05-17 21:11:40 +02:00
Your Name
2184578cfc
v1.0.7: Windows GGUF PATH overflow fix, per-node prompt persistence, return_raw_trace optional, WinError 206 runtime fix
2026-05-17 20:00:44 +02:00
Your Name
b2295ab23a
Fix Windows GGUF llama_cpp import runtime
2026-05-16 13:59:34 +02:00
Your Name
972998ef98
Fix stream toggle bug and retain versioned releases
2026-05-16 13:47:00 +02:00
Your Name
c091b30faf
Fix GGUF workflow compatibility and release v1.0.4
2026-05-16 12:46:38 +02:00
Your Name
9e18c83b5e
Fix ThinkingLLM node reload imports
2026-05-16 11:08:07 +02:00
Your Name
2250058c4b
Add legacy AILab node aliases
2026-05-16 10:55:35 +02:00
Your Name
1cae110284
Strip BOM, add test suite with collision/toggle/streaming coverage (14/14)
2026-05-16 09:42:21 +02:00
Your Name
0ca0d2448c
Initial publish of ComfyUI-ThinkingLLM
2026-05-15 11:48:10 +02:00
Marcel Hofer