n8n RAG Chatbot Workflow: Firecrawl + Pinecone + Gemini

Turn your website into a searchable, chat-based knowledge assistant with n8n RAG

This n8n RAG Chatbot Workflow automatically crawls your website with Firecrawl, indexes cleaned text into Pinecone using Google Gemini embeddings, and serves a public chat webhook where a Gemini-powered agent answers user questions by retrieving the most relevant chunks from your index.

What this workflow does

• Ingest (manual run): When you manually execute the workflow, it starts a full indexing run.
• Prepare Pinecone: It looks up your Pinecone index host and clears all vectors in the configured Pinecone namespace to avoid duplicate content.
• Discover pages: Using Firecrawl, it maps the target website and retrieves up to 100 discovered page URLs.
• Clean & normalize URLs: The workflow filters, deduplicates, and normalizes URLs, then processes them one at a time with a short delay to reduce request bursts.
• Extract text: For each page, it fetches HTML over HTTP, strips HTML/scripts/styles into plain text, and skips pages with too little usable content.
• Chunk + embed + index: It splits text into overlapping chunks, creates Gemini embeddings, and stores vectors in Pinecone with URL metadata.
• Chat (public webhook): When a chat message is received, the Gemini agent retrieves relevant chunks from Pinecone and uses short windowed memory to generate the final answer.

Use cases

• SaaS support teams needing instant answers grounded in their own docs (FAQ, pricing, product pages).
• Automation engineers building a RAG knowledge base for marketing or help-center websites.
• Operators who want a lightweight, website-specific chatbot without manually maintaining knowledge snippets.

Technical details

• Integrations: Firecrawl, Pinecone, Google Gemini (Gemini embeddings for ingestion and retrieval—use the same embedding model).
• Core node types (n8n): if, set, code, wait, and http request (plus sticky note documentation).

Set your Firecrawl, Pinecone, and Gemini credentials, create a Pinecone index with the correct vector dimension, and configure YOUR_PINECONE_INDEX and YOUR_PINECONE_NAMESPACE in the Pinecone-related nodes to get started.