The ComfyUI Node Pack That Tried to Talk to Every LLM at Once

What it does ComfyUI-IF_AI_tools was a kitchen-sink collection of custom nodes that let you generate prompts, query local or remote LLMs, run OCR-RAG, and detect objects without ever leaving your ComfyUI graph. It essentially turned the image-generation UI into a universal client for language models.

The interesting bit The scope was almost aggressively broad: nanoGraphRAG, ColPali/ColQwen OCR, Florence2 vision, Omost support, and templated IF_Assistants for reasoning and reflection all lived in one repo. The author even released custom Ollama fine-tunes specifically for prompt generation, suggesting the goal was to own the entire text-to-image stack.

Key highlights

• Backends galore: Ollama, llama.cpp, LM Studio, Koboldcpp, TextGen, Transformers, plus cloud APIs from Anthropic, Groq, OpenAI, Gemini, Mistral, and xAI.
• Vision and RAG bundled in: Florence2 object detection, Supervision, nanoGraphRAG, and OCR-RAG via ColPali/ColQwen.
• IF_Assistants shipped with preset system prompts for reasoning, reflection, and reward templates.
• Custom fine-tuned models (impactframes/llama3_ifai_sd_prompt_mkr_q4km) were published to generate prompts locally.
• Archived: the author has moved active development to comfyui-llm-toolkit and ComfyUI-IF_AI_PromptImaGen.

Caveats

• The repo is explicitly archived; the README points to successor projects for continued use.
• The TODO list openly admits bugs on the latest ComfyUI and a broken Graph Visualizer node.
• The feature surface is so wide that maintenance was likely a losing battle.

Verdict ComfyUI tinkerers who want to study an ambitious integration pattern may find the code useful, but new users should install the maintained successor repos instead. As a production dependency, it is a frozen snapshot rather than a living tool.