Chat with Google Drive Docs in n8n using Pinecone & OpenAI

Chat with your Google Drive Docs inside n8n—powered by Pinecone + OpenAI

Turn any Google Drive document into a searchable knowledge base and enable an n8n chat endpoint that answers questions with grounded, context-aware responses—using Pinecone vector search and OpenAI embeddings and chat.

What this workflow does

• Ingests a Google Drive document on demand: Run the workflow manually to start ingestion using your configured Google Drive file URL, chunking settings, and a chosen Pinecone namespace.
• Downloads and indexes content: It downloads the document as a binary file, loads it with source metadata, splits the text into overlapping chunks, generates OpenAI embeddings, and upserts vectors into your selected Pinecone index (optionally clearing the namespace first).
• Enables chat with retrieval: When an n8n chat message arrives, the workflow embeds the user’s question, retrieves the top matching chunks from Pinecone, and passes them to an OpenAI-powered agent to generate a response.
• Maintains short-term context: It uses a rolling conversation memory window so follow-up questions can stay coherent.

Use cases

• Ask questions about an internal Google Drive spec, SOP, or contract and get answers sourced from the document.
• Build a support-style chat experience for SaaS operators—backed by your own documentation stored in Google Drive.
• Speed up engineering and ops by turning long docs into instant Q&A during incident response or onboarding.

Technical details

• Triggers & flow: Manual Trigger to run ingestion; n8nn8n-nodes-langchainchat chat trigger for incoming questions.
• Document source: Google Drive (OAuth2) to download the document from the configured file link.
• Vector storage & search: Pinecone index for upserting and retrieving relevant chunks using the matching embedding dimensions (1536).
• LLM & embeddings: OpenAI chat model for responses and OpenAI embeddings for indexing and query matching.
• Agent tooling: Uses LangChain-style agent components to ground answers in retrieved Pinecone context.

Setup note: Ensure the Pinecone index selected in both ingestion and retrieval nodes matches the embedding dimensions (1536), then update the Google Drive file link, Pinecone namespace, and chunk size/overlap in Configuration.