Skip to content

CHANGELOG.md

Latest commit

 

History

History
121 lines (90 loc) · 5.74 KB

CHANGELOG.md

File metadata and controls

121 lines (90 loc) · 5.74 KB

Changelog

All notable changes to ComfyUI-MultiModal-Prompt-Nodes will be documented in this file.

[1.0.10] - 2026-04-02

  • Added support for Qwen3.5 local GGUF models

    • Added Qwen3.5 model detection and proper handler selection (Qwen35ChatHandler)
    • Fixed incorrect fallback to Qwen3VLChatHandler for Qwen3.5 model names
    • Updated mmproj handling for Qwen3.5 (requirement checks and auto-detection flow)

  • Improved post-run cleanup behavior for local model nodes

    • VisionLLMNode, WanVideoPromptGenerator, and QwenImageEditPromptGenerator now call cleanup() at the end of execution
    • Introduced cleanup(finalize=False/True) to separate regular unload from final teardown on process exit
    • Added safe manager re-initialization after cleanup for stable repeated runs

[1.0.9] - 2026-03-15

  • Expanded the search scope for local Qwen-family GGUF models

    • Added models/text_encoders and all subdirectories under both models/LLM and models/text_encoders to the search paths
    • Centralized model path resolution and mmproj resolution in local_gguf_utils.py to reduce duplicated logic

  • Improved mmproj selection behavior

    • The UI now shows only mmproj files from the same directory as the selected GGUF model
    • When mmproj = (Not required) is selected, the node now explicitly switches to text-only mode to avoid unnecessary vision handler usage

  • Strengthened the local prompt rewrite flow for Qwen and Wan

    • Added dedicated system prompts for qwen_image, qwen_image_edit, wan_t2v, and wan_i2v
    • Tightened prompt instructions to reduce verbose analysis-style responses and make it easier to return only the final prompt body
    • Added a second pass that preserves quoted text and normalizes the result to Simplified Chinese when Chinese output is requested but another language is returned

  • Expanded Qwen Image Edit Prompt Generator

    • Made the image input optional so Qwen-Image can also be used for text-only prompt generation
    • Treated mmproj as not required when Qwen-Image is run locally without images
    • Increased local inference max_tokens and n_ctx to better support longer prompt generation

  • Improved the robustness of Wan Video Prompt Generator

    • Added explicit validation errors when Image-to-Video is used without an input image

[1.0.8] - 2026-02-09

  • Fixed issue where Qwen2.5-VL were always loaded in text-only mode even when a valid mmproj file was specified.

    • Added vision chat handler support for Qwen2.5-VL
    • Enable vision mode automatically when supported model + mmproj are present

  • Improved mmproj auto-detection logic

    • Auto-detect now selects mmproj files based on model family prefix (qwen2, qwen3) instead of arbitrary alphabetical fallback
    • Prevents incorrect mmproj selection when multiple mmproj files exist in the same directory

[1.0.7] - 2026-01-26

Fixed

  • Fixed incorrect detection of Qwen3-VL when mmproj is set to (Not required).
    • Disabled automatic mmproj detection and prevented use of the VL handler in this case.
    • Updated GGUFModelManager.load_model and node-side mmproj interpretation to correctly respect (Not required).

[1.0.6] - 2026-01-16

Fixed

  • Fixed an issue where incorrect mmproj could remain loaded when switching between Qwen3-VL GGUF models
    • Properly unload and reload GGUF models when model or mmproj changes
    • Prevent stale vision projectors from being reused across different Qwen3-VL models
  • Improved mmproj auto-detection logic to avoid accidentally picking mmproj files from other models

Changed

  • Refined internal GGUF model lifecycle management for better stability when switching models (e.g. 8B ↔ 4B)
  • Minor internal refactors to reduce state leakage in llama-cpp-python based vision models
  • Improved README documentation for clarity and accuracy:
    • Clarified project scope as a prompt generator for QwenImageEdit and Wan2.2
    • Reorganized Credits and Dependencies to clearly separate derived works and external dependencies
    • Updated llama-cpp-python installation notes to reference the JamePeng fork documentation directly, avoiding incomplete or misleading installation instructions

Added

  • Added a backends/ directory as a structural placeholder
    • This directory does not change behavior in v1.0.6
    • Reserved for future refactoring of Local GGUF and Cloud API backends without changing node interfaces

[1.0.5] - 2026-01-13

Removed

  • General Prompt Rewriter (nodes.py) - Removed as it was unchanged from original ComfyUI-QwenPromptRewriter
    • Users should use the original ComfyUI-QwenPromptRewriter for this functionality

Changed

  • Updated documentation to reflect 3-node architecture
  • Clarified project scope: focus on multimodal (vision + text) capabilities

[1.0.4] - 2026-01-12

Author

Added

  • Initial release of ComfyUI-MultiModal-Prompt-Nodes
  • Vision LLM Node: Local GGUF vision language model support
    • Qwen2.5-VL and Qwen3-VL compatibility
    • Multi-image input (batch support)
    • Multiple style presets (raw, default, detailed, concise, creative)
  • Qwen Image Edit Prompt Generator: Image editing prompt generation
    • Dynamic model selection (local GGUF + cloud API)
    • Manual mmproj selection for Qwen3-VL
    • Multi-image support via image2/image3 inputs
  • Wan Video Prompt Generator: Video generation prompt optimization
    • Text-to-Video and Image-to-Video task support
    • Local Qwen3-VL integration
    • Wan2.2-specific prompt templates

Technical Details

  • GPL-3.0 license (due to llama-cpp-python dependency)
  • Unified category: multimodal/prompt
  • Python 3.10+ support
  • Comprehensive error handling and user feedback

[Unreleased]

Planned

  • Workflow templates
  • Extended documentation