data / Paused, local-only
Mānoa Faculty Senate RAG
Local document retrieval and citation project for Faculty Senate materials with extraction, metadata policy, LanceDB retrieval, grounded citations, and scoped evaluation evidence.
RAG pipeline
Mānoa Faculty Senate RAG
Document ingestion and retrieval-augmented question answering.
- Role
- Solo developer
- Updated
- Jul 2026
- Tech
- PythonFastAPILanceDBRAGdocument extractioncitationsevaluation
Overview
Mānoa Faculty Senate RAG is a paused, local-only retrieval project that ingests local Faculty Senate PDFs and DOCX files, extracts text and tables, chunks documents with metadata, stores embeddings in LanceDB, and returns grounded answer/citation responses through local API/UI paths.
It is useful supporting evidence for AI retrieval and citation systems. It is not presented as an institutional system, official tool, public deployment, or complete factual-answer guarantee. The project includes scoped retrieval evaluation evidence, not a guarantee of reliable factual answers across the full archive.
What I Built
The implementation includes:
- local document extraction
- table/text handling
- metadata and chunking policy
- LanceDB retrieval storage
- grounded citation paths
- local API/UI paths
- scoped retrieval evaluation
Boundaries
This page intentionally avoids claims about institutional use, formal approval, public deployment, or reliable factual answers across the full archive.
Source-safety classification and polished public demo artefacts are planned before expanding this into a fuller case study.