AI Chat with PDFs: Q&A and Source Citations

This n8n workflow transforms static PDF documents into interactive knowledge bases. It fetches PDFs (e.g., from Google Drive), splits them into semantic chunks, embeds them using AI models, and stores them in a Pinecone vector database for efficient retrieval. When a user asks a question via n8n's built-in chat or custom triggers like Slack/Teams, the workflow retrieves relevant chunks, feeds them to an LLM (e.g., GPT-3.5 or custom), and generates responses with inline citations pointing to exact source locations, such as '[Document.pdf, lines 1-35]'.

Benefits include enhanced accuracy in AI responses by grounding them in document content (RAG architecture), reducing hallucinations, and building trust through verifiable sources. It's ideal for sales teams analyzing business proposals, legal reviews of contracts, or researchers querying technical papers. Time-saving potential is high: no manual searching, instant insights from large docs.

Use cases: In CRM/Sales (Business Proposals subcategory), query proposal details like pricing, terms, or client specifics. Customize for multi-document folders, swap LLMs (self-hosted options), or integrate with enterprise chat tools. Requires Pinecone (free tier sufficient) and an LLM API. Setup once, query repeatedly—scalable for teams handling document-heavy workflows.