Document Library (AI Q&A)

The Document Library agent lets you attach documents, ask questions of their content, and receive AI-powered answers with traceable evidence — a full RAG (Retrieval-Augmented Generation) pipeline integrated into Nodlin.

The agent defines seven node types:

• Library 📚 — Attach documents and ask questions
• Document 📄 — Attach files (PDF), summarise to Markdown, record in vector database
• Question ❓ — Ask questions using the local vector database (RAG/LangChain)
• Answer 💬 — AI response with confidence level
• Evidence 🔍 — Supporting detail from the vector database with similarity score
• Chunk 📎 — The actual document content (Markdown) recorded in the vector database
• Entity 🏷️ — Person, place, metric, or other entity type extracted from evidence to surface a knowledge graph

How It Works

• Add documents to a Library by attaching files (e.g. corporate earnings reports, policy documents)
• Documents are automatically converted to Markdown and indexed in a vector database
• Ask questions of the library (e.g. “What are the key risks to growth in 2026?”)
• Review answers alongside the underlying evidence and source reference points

Benefits

• Traceable Answers 🔎 — Every answer links to the evidence and source document chunks that support it
• Knowledge Graph 🌐 — Entities (people, places, metrics) are extracted to surface connections
• Confidence Scoring 📊 — Answers include a confidence level so you know how reliable they are
• Collaborative 🤝 — Teams share libraries and build institutional knowledge together