rag-system-design
Skill0 starsChanged 5 months agoMITView on GitHub
Design or review a production RAG (retrieval-augmented generation) system — ingestion, chunking, embeddings, vector store, retrieval, reranking, prompt assembly, generation, and evaluation / hallucination control. Use when building or debugging RAG, semantic search, or Q&A-over-documents, or when retrieval quality is poor. NOT for generic ML systems (use ml-system-design).
In this repository12 other files it gives its agents.
- adr-writerSkill
- architecture-reviewSkill
- gcp-architectureSkill
- incident-postmortemSkill
- interview-modeSkill
- leadership-lensSkill
- learning-coachSkill
- llm-app-architectureSkill
- mlops-pipeline-reviewSkill
- ml-system-designSkill
- observability-reviewSkill
- principal-code-reviewSkill
The file
Files
RAG System Design
Walk every stage. Most RAG failures are retrieval failures, not model failures — diagnose retrieval before blaming the LLM.
Ingestion & chunking
- Source parsing (PDF/HTML/code) and cleaning
- Chunk size & overlap; semantic vs fixed-size splitting
- Metadata captured per chunk (source, section, timestamp, ACL)
Embeddings & index
- Embedding model choice and dimensionality
- Vector store & index type (HNSW/IVF); recall vs latency
- Hybrid search (dense + lexical/BM25)
Retrieval & reranking
- top-k selection and filters (metadata, freshness, permissions)
- Reranker (cross-encoder) when precision matters
- Query rewriting / expansion / multi-query
Generation & grounding
- Prompt assembly and context-window budgeting
- Citation / source attribution
- Refusal when context is insufficient (anti-hallucination)
Evaluation
- Retrieval: recall@k, MRR, context precision
- Generation: faithfulness/groundedness, answer relevance
- Golden set + automated eval in CI
Ops & cost
- Index freshness / re-embedding strategy
- Latency budget per stage
- Cost per query (embedding + retrieval + generation tokens)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.