n8n PDF RAG Chatbot Workflow: OpenAI + Pinecone + Cohere

Build a PDF-powered RAG chatbot in n8n (OpenAI + Pinecone + Cohere)

If you want your AI assistant to answer questions using the content of your own PDF documents, this n8n PDF RAG Chatbot Workflow gives you a complete, ready-to-use Retrieval-Augmented Generation (RAG) template—so your chatbot can ingest PDFs, index them, and retrieve relevant passages for accurate responses.

What this workflow does

This workflow runs in two distinct stages:

• 1) Data ingestion & indexing: Upload a .pdf via the n8n Form Trigger. The Default Data Loader processes the PDF, and a Recursive Character Text Splitter divides the content into smaller chunks.
• 2) Embeddings + storage: The Embeddings OpenAI node converts each text chunk into vector embeddings. Those embeddings are then upserted into your selected `Pinecone` index using the Pinecone Vector Store node, creating a searchable knowledge base.

Note: This workflow uses paid services (OpenAI, Pinecone, Cohere), so costs will incur based on usage. Review each provider’s pricing to estimate expenses.

Use cases

• Answering customer questions from PDF policy documents, manuals, or FAQs
• Building an internal knowledge assistant for uploaded reports and documentation
• Creating a SaaS operator-friendly chatbot that can ground responses in your document library

Technical details

• Self-hosting requirement: Contains community nodes compatible only with the self-hosted version of n8n.
• Core integrations/nodes: OpenAI (Embeddings), Pinecone (Vector Store), and Cohere (re-ranking).
• Workflow components (as included): Form Trigger, Default Data Loader, Recursive Character Text Splitter, Pinecone Vector Store, plus LangChain-related nodes such as n8nn8n-nodes-langchainlm chat open ai, n8nn8n-nodes-langchainreranker cohere, and chat/agent trigger nodes.

Video tutorial (Bahasa Indonesia): N8N Tutorial: Membangun Chatbot RAG dengan Pinecone, OpenAI, & Cohere