Hi, thank you for maintaining this repo, it is a very useful overview for people who are learning and building RAG systems.
I would like to suggest adding one more item under the “Guides and recipes” or “Tutorials” style section.
The project is called WFGY RAG Problem Map and lives here:
https://github.com/onestardao/WFGY/tree/main/ProblemMap/README.md
In short, it is a 16 mode failure map for RAG and LLM pipelines.
Instead of showing how to build a RAG system, it focuses on the question “why did my RAG fail in this specific way” and gives a concrete slot for each kind of failure.
A very compressed overview:
It collects the most common failure patterns for real world RAG systems, for example: retrieval drift, vector store ingestion mistakes, wrong deployment order, prompt and chunking contract mismatch, etc.
Each pattern is mapped to a numbered entry with a short description and suggested fix direction, so developers can quickly say “this bug is ProblemMap No X” and follow the path.
The whole map is plain text, MIT licensed, and optimized to be copy pasted into an LLM prompt so the model can help with triage and minimal fix suggestions.
The main WFGY repo is here for context (about 1.4k stars at the moment):
https://github.com/onestardao/WFGY
If this kind of “failure oriented” resource fits the scope of rag-resources, I would be happy to adapt the wording or provide a short one line description that matches your style. If it feels out of scope, no worries at all, thanks again for putting this list together.
Hi, thank you for maintaining this repo, it is a very useful overview for people who are learning and building RAG systems.
I would like to suggest adding one more item under the “Guides and recipes” or “Tutorials” style section.
The project is called WFGY RAG Problem Map and lives here:
https://github.com/onestardao/WFGY/tree/main/ProblemMap/README.md
In short, it is a 16 mode failure map for RAG and LLM pipelines.
Instead of showing how to build a RAG system, it focuses on the question “why did my RAG fail in this specific way” and gives a concrete slot for each kind of failure.
A very compressed overview:
It collects the most common failure patterns for real world RAG systems, for example: retrieval drift, vector store ingestion mistakes, wrong deployment order, prompt and chunking contract mismatch, etc.
Each pattern is mapped to a numbered entry with a short description and suggested fix direction, so developers can quickly say “this bug is ProblemMap No X” and follow the path.
The whole map is plain text, MIT licensed, and optimized to be copy pasted into an LLM prompt so the model can help with triage and minimal fix suggestions.
The main WFGY repo is here for context (about 1.4k stars at the moment):
https://github.com/onestardao/WFGY
If this kind of “failure oriented” resource fits the scope of rag-resources, I would be happy to adapt the wording or provide a short one line description that matches your style. If it feels out of scope, no worries at all, thanks again for putting this list together.