Add WFGY 16 Problem Map LLM Debugger (Nebius-compatible RAG app)

[onestardao]

🔗 Linked Issue

N/A. This is a new RAG app and is not linked to an existing issue.

✅ Type of Change

• ✨ New Project/Feature
• 🐞 Bug Fix
• 📚 Documentation Update
• 🔨 Refactor or Other

📝 Summary

This PR adds a new RAG application under rag_apps/wfgy_llm_debugger called WFGY 16 Problem Map LLM Debugger.

The app is a Nebius compatible notebook plus minimal CLI that turns messy LLM or RAG bugs into one of 16 reproducible failure modes, each mapped to a concrete fix in the open source WFGY Problem Map.

This includes:

• rag_apps/wfgy_llm_debugger/README.md
  • Follows the project README template structure
  • Explains the 16 Problem Map classes and how the debugger works
  • Shows how to run it on Nebius hosted notebooks, Colab, or any OpenAI compatible endpoint
• rag_apps/wfgy_llm_debugger/main.py
  • Single cell style script that:
    • downloads TXTOS.txt and ProblemMap/README.md from the WFGY repo
    • builds a combined system prompt
    • asks the user for a bug description, including prompt, answer and logs
    • returns one primary Problem Map number No.1 to No.16, an optional secondary candidate, and suggested next steps

The script depends only on openai and requests and talks to any OpenAI compatible API (Nebius, OpenAI, etc).

📖 README Checklist

• I have created a README.md file for my project.
• My README.md follows the official .github/README_TEMPLATE.md.
• I have included clear installation and usage instructions in my README.md.
• I have added a GIF or screenshot to the assets folder and included it in my README.md.

✔️ Contributor Checklist

• I have read the CONTRIBUTING.md document.
• My code follows the project's coding standards.
• I have placed my project in the correct directory (e.g., rag_apps).
• I have included a requirements.txt or pyproject.toml for dependencies.
• I have added a .env.example file if environment variables are needed and ensured no secrets are committed.
• My pull request is focused on a single project or change.

💬 Additional Comments

• The app uses only openai and requests. I kept it as a single script so it can be dropped into Colab or Nebius notebooks without extra files. I can add a minimal requirements.txt if you prefer.
• I did not add a GIF or screenshot yet to keep the initial PR small. I am happy to add a simple Nebius or Colab screenshot that shows the debugger in action if that would be useful for the gallery.
• Naming and folder placement are flexible. If you prefer a different name or location for this app, I am happy to adjust.

[shivaylamba]

You are not using Nebius Token Factory end point for LLM calls. Please refer to the docs and apply the changes.

[onestardao]

Thanks a lot for the review and for pointing out the Nebius Token Factory detail.

I have updated both the README and the main.py script.

The debugger is now explicitly described as working with any OpenAI-compatible endpoint, and Nebius is given as a first class example. In the script, users can paste their Nebius Token Factory URL as the base_url and use any Nebius model id as the model name, so it should work fine in a Nebius hosted notebook as well.

I kept the code provider-agnostic on purpose, because the main goal of this app is to demonstrate the WFGY 16 Problem Map debugger pattern, and many people will want to run it on different providers or internal gateways. If you would still prefer a version that hardcodes the Nebius Token Factory endpoint and a specific Nebius model, I am happy to adjust and follow your preferred style.

Please let me know if the current wording and setup look good to you.

[onestardao]

hi, just a small ping and no rush at all.

the changes were approved on Feb 9, so I just wanted to double-check if there is anything else you would like me to adjust before this can be merged. I am happy to update the README or code if needed.

thanks again for maintaining this repo.