video1.0<iframe src="https://www.loom.com/embed/a90bed5b12434de3a098f3594e83e9bb" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/a90bed5b12434de3a098f3594e83e9bb-6036a80e9a55bb56.gif300.139This Loom explains a prototype for building a document Q&A RAG system for the reclaim process, using public APIs and Superbase to store and retrieve content. The author describes running a local backend and front end, ingesting documents into a vector store, and exposing the knowledge base via an MCP server tool, with citation-backed responses. They demonstrate a simple evaluation pipeline using FAFSA-related information and report position five out of twelve test question results with an overall precision of about 83% for 33%. They discuss next steps focused on improving reliability and measurable quality, including hybrid retrieval and re-ranking, benchmarking, and adding guardrails and better observability.