Quality Assurance
RAG.md · 8 documents
RAG Deep Dive Part 7: Evaluation and Debugging RAG Systems
Teaches you to measure and debug a RAG pipeline with retrieval metrics, generation metrics, and production monitoring, all implemented from scratch in Python.
RAG Evaluation
Explains how to measure RAG system quality using four RAGAS metrics, faithfulness, answer relevancy, context precision, and context recall.
embedding-first-chunking-second-smarter-rag-retrieval-with-max-min-semantic-chunking
Explains Max-Min Semantic Chunking, an embed-first chunk-second strategy that uses semantic similarity to set chunk boundaries for better RAG retrieval.
Fix Summary: Matroska Adaptive Chunking for 128D Embeddings
Fixes a bug that truncated 128D embeddings to 4D, restoring non-zero cohesion metrics via adaptive chunking into 3D GPU operations.
SKILL-FORGE AUDIT: sop-dogfooding-pattern-retrieval
Audits a skill file for pattern retrieval via vector search, scoring it 7.2/10 and flagging missing scripts, wrong API calls, and incomplete implementation as critical blockers.
Week 2 Requirements - Completion Status
Tracks completion status of 13 project requirements for a chatbot with LLM and RAG, marking 8 as done and 5 as partial or missing.
MCP Fact-Check Design Document (Enhanced)
Designs an MCP server that validates claims against official MCP specs using embeddings and semantic search.
JudgeIt (From SuperKnowa)- Automatic Eval Framework for Gen AI Pipelines
Automates GenAI pipeline evaluation using an LLM-as-a-judge, replacing slow human evaluation for RAG, multi-turn query rewrite, and agentic workflows.