trailblazer09/referencerover

This is a command-line PDF RAG evaluation harness. A user supplies a question or requests an evaluation run; the CLI loads PDFs, indexes their text, retrieves relevant chunks, and passes the question and context to an LLM. Results include references, grounding status, and per-call token and latency metrics. The graph separates ingestion/indexing, question answering, and external services. Some implementation is unsampled, so ingestion and CLI wiring also rely on the README’s description; uncertain internal calls are omitted.