n8n RAG Chatbot Workflow: Groq + Pinecone + Drive Ingest

Turn your Google Drive documents into a grounded RAG chatbot—automatically

This n8n workflow delivers an AI RAG chatbot that answers incoming chat messages using a Groq Llama model grounded in a Pinecone knowledge base—and keeps that knowledge base fresh by ingesting new files from a specific Google Drive folder every minute.

What this workflow does

• Chat via public webhook trigger: Receives an incoming chat message using an n8n chat trigger exposed as a public webhook URL.
• Grounded, context-based responses: Runs a LangChain agent with a Groq Chat Model to answer using retrieved context and maintains short-term conversation memory.
• Retrieval from Pinecone: Fetches relevant knowledge base chunks from a Pinecone vector index using Google Gemini embeddings, with Cohere reranking to improve relevance.
• Returns the grounded answer: Sends the grounded response back to the chat client.
• Continuous knowledge updates: Every minute, detects new files created in a watched Google Drive folder, downloads them, splits them into chunks, generates Gemini embeddings, and inserts the chunks into the same Pinecone index for future retrieval.

Use cases

• SaaS support assistant: Answer product questions using your documentation stored in Google Drive.
• Internal knowledge Q&A: Empower teams to ask policy or procedure questions with Pinecone-backed retrieval.
• Ops-ready document ingestion: Add new PDFs or docs to the monitored folder and have them automatically available to the chatbot within minutes.

Technical details

• Integrations/credentials: Groq, Pinecone, Cohere, Google Drive (OAuth2), and Google Gemini (PaLM) API.
• Key workflow behavior: Public webhook-based chat trigger; Google Drive folder trigger running every minute; LangChain agent with Groq; Pinecone retrieval with Gemini embeddings + Cohere reranking.
• Pinecone setup: Create/select the target index (set to erhan8n in the workflow) and ensure vector dimensions match your chosen Gemini embedding model.