nxrmal03/loandoc-ai

LoanDoc AI is a production-ready RAG (retrieval-augmented generation) Q&A system for loan documents. It ingests PDFs/DOCX, indexes text into a local Chroma vector DB, and answers natural-language queries with cited source snippets. Who uses it: engineering teams, data scientists, or operations/legal groups needing searchable Q&A over loan paperwork. Key features: - Frontend: React + TypeScript; Backend: FastAPI + LangChain - Ingest: pypdf/python-docx → RecursiveCharacterTextSplitter (1,000 chars, 200 overlap) - Embeddings: OpenAI text-embedding-3-small; store in local Chroma - Retrieval: embed query → top-5 chunks; primary LLM: Anthropic Claude Sonnet; fallback: OpenAI GPT-4o - Configurable via backend/.env (chunk size, overlap, retrieval K, models) - Simple local dev (no Docker), tests mock LLMs and use temporary Chroma DB - Exposed API: upload, list, delete, query, health; responses include model_used and citations.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.