Build PDF Search RAG System: Mistral OCR + Weaviate DB

This n8n workflow creates a complete Retrieval-Augmented Generation (RAG) system for PDF documents. It starts by processing uploaded PDFs with Mistral AI's OCR to accurately extract text, even from scanned or image-based files. The extracted text is then automatically chunked into optimal sizes for vectorization, preserving context for better retrieval.

Embeddings are generated using Cohere's multilingual models and stored in a Weaviate vector database under a 'KnowledgeDocuments' collection. This enables efficient semantic similarity searches, enhanced by Cohere reranking for top precision. Document metadata like filename, source, and timestamp is included for comprehensive tracking.

The system exposes the knowledge base via an MCP (Model Context Protocol) server, allowing seamless integration as an AI tool in external workflows. Benefits include rapid deployment of enterprise-grade document search, reduced manual data entry, and scalable AI-powered querying across large PDF corpora. Ideal use cases: legal document review, research archives, customer support knowledge bases, and technical manual search.