AI-Powered RAG Chat for Technical Document Automation

Transform your document repositories into intelligent, conversational knowledge bases with this AI-powered RAG workflow that combines Mistral AI with pgVector semantic search to deliver precise, source-cited answers from your technical documentation.

What this workflow does

This automation creates a sophisticated chat interface that allows users to query document collections using natural language. The workflow leverages a Mistral Large AI agent equipped with specialized tools to intelligently decide whether to perform semantic vector searches for content or SQL queries for metadata. When users ask questions, the system searches through pgVector-stored document chunks for relevant content while simultaneously querying structured publication data including titles, authors, DOIs, and download URLs. All responses include proper source citations and maintain conversation context through persistent PostgreSQL chat memory.

Use cases

• Institutional knowledge management: Government agencies and research organizations can make vast publication libraries instantly searchable
• Standards compliance: Engineering teams can quickly find specific requirements across multiple technical standards documents
• Research acceleration: Academic teams can conversationally explore paper collections with automatic source attribution
• Technical support: Customer service teams can rapidly locate answers from product documentation and manuals
• Legal document review: Law firms can efficiently search through case files and regulatory documents

Technical details

Built with enterprise-grade n8n nodes including LangChain Agent integration, Mistral Cloud LLM, and PostgreSQL tools. The workflow exposes a public chat endpoint through the n8n Chat Trigger node and maintains conversation history for contextual responses. The AI agent utilizes two specialized tools: a Vector Store tool for semantic similarity search over pgVector tables and a SQL tool for structured metadata queries. All database operations are read-only for security, while the sticky note node provides workflow documentation.

Perfect for organizations requiring intelligent document search capabilities with proper source attribution and conversational user experience.