Suggestion: add WFGY (open-source LLM / RAG debugging toolkit)

[onestardao]

Hi, first of all, thank you for curating this list.
I have a suggestion for an additional tool that might be useful for people building agents.

Project: WFGY · LLM / RAG debugging and “semantic firewall” toolkit

• GitHub: https://github.com/onestardao/WFGY
• License: MIT

Very short overview:

• WFGY focuses on diagnosing and repairing broken RAG / LLM pipelines, instead of being “yet another agent framework”.
• It ships with a 16-problem “Problem Map” that covers the most common failure modes in RAG (retriever drift, index poisoning, vectorstore issues, bad chunking, prompt collisions, etc).
• The recommended workflow is simple: copy-paste your failing prompt or logs into the system, map it to a problem number, then follow the prescribed fix or checklist.
• It is provider-agnostic and works with existing stacks (LangChain, LlamaIndex, custom agents) by acting as a text-level semantic firewall in front of the model.

Given that this repo already lists a mix of agent frameworks, tools, and learning resources, WFGY might fit either:

• as a tool under an “Agent / RAG debugging” category, or
• as an advanced learning resource for people who want to go deeper into evaluation and failure analysis.

If this sounds aligned with the scope of the list, I would be happy to open a PR with a short line like:

WFGY – MIT licensed toolkit for debugging RAG and LLM agents using a 16-problem Problem Map and semantic firewall.

Thanks a lot for considering it.