n8n Dual-Stage RAG Chatbot: Google Drive + Pinecone + Gemini

Turn your Google Drive PDFs into a smarter RAG chatbot—with a fast semantic cache

This n8n workflow template builds a dual-stage RAG chatbot that ingests documents from Google Drive into a Pinecone knowledge-base index, then answers questions using Gemini—while first checking a Pinecone semantic cache to return repeat answers instantly.

What this workflow does

• Ingests documents from Google Drive: triggers on a Google Drive folder update (or can run manually), lists and downloads files from a specified folder.
• Extracts text by file type: detects whether each file is a PDF or plain text and extracts the document text accordingly.
• Builds a Pinecone knowledge base: splits content into overlapping chunks, embeds them using Ollama (nomic-embed-text), and inserts vectors into a Pinecone index.
• Responds to chat queries with caching: when a chat message arrives, the workflow embeds the user query with Ollama and searches the Pinecone semantic-cache index for the closest prior answer.
• Fast cache hit behavior: if the best match score is at least 0.88, it returns the cached answer immediately.
• Cache miss fallback to Gemini Agent: on a cache miss, it uses a Google Gemini AI Agent with conversation memory and a Pinecone vector-store retrieval tool to generate a grounded answer from the knowledge base.
• Improves future responses: stores the new query-and-answer pair back into Pinecone semantic cache.

Use cases

• SaaS teams answering support questions from internal PDFs stored in Google Drive.
• Automation engineers building a retrieval-augmented assistant for policy/knowledge documents.
• Operators reducing AI compute cost by leveraging Pinecone semantic caching for repeat inquiries.

Technical details

• Pinecone indexes: two separate indexes (cosine similarity, 768 dimensions): one for the knowledge base (e.g., nomic-embed-text) and one for the semantic cache (e.g., rag-semantic-cache).
• Embeddings: Ollama embeddings using nomic-embed-text.
• AI: Google Gemini AI Agent with conversation memory + Pinecone retrieval tool.
• n8n logic & nodes: if, set, code, merge, Google Drive, plus supporting workflow nodes.