Arcane Library
Upload your technical documents, ask questions about them and get answers with evidence. Every claim cites the passage it came from, and when the documents don't cover a question, the app says so instead of inventing an answer.
- Full RAG pipeline: text extraction from PDF, Markdown and TXT, chunking, embeddings and vector search in pgvector
- Grounded answers from Claude with numbered citations; below a relevance floor the model is never called
- Multilingual E5 embeddings running locally, so Armenian and Russian documents work too
- Auth with password reset, bookmarks for saved answers, tests for chunking, extraction, search and chat
Private repository · code walkthrough on request