RAG Fails Silently: Debugging Retrieval, Citations, and Unsupported Claims

A practical look at debugging the evidence chain in RAG systems: retrieval, context selection, answer claims, citation, and local failure reports. RAG systems often fail in a way that is hard to see. The citations look official. The retrieved chunks look vaguely related. Then a user asks a question where the model combines one ed fact with one invented detail, and nobody notices until the answer is wrong in production. That failure mode is what I wanted to debug better. I built ContextTrace, a local-first Python SDK and CLI for tracing RAG and agent applications.