n8n RAG Workflow: Google Drive to Pinecone with Context Chunking

Turn Google Drive documents into high-quality Pinecone knowledge for RAG—using context chunking

This n8n workflow automatically pulls a document from Google Drive, splits it into context-preserving chunks using section boundary markers, generates contextual metadata with OpenAI via GPT-4.0-mini (OpenRouter), and stores the results in a Pinecone vector store to improve Retrieval-Augmented Generation (RAG) accuracy.

What this workflow does

• Retrieve a source document from Google Drive for ingestion into your RAG pipeline.
• Extract the document’s text content using predefined section boundary markers to identify logical boundaries.
• Create context-based chunks (Code node) by splitting the document at those boundaries so each chunk retains meaningful context within the full source.
• Loop through each chunk (Loop node) to process chunks individually while maintaining linkage to the overall document context.
• Generate contextual metadata per chunk (Agent node) using GPT-4.0-mini via OpenRouter to support more accurate retrieval.
• Prepend context to each chunk and create embeddings, preparing the data for storage in Pinecone.

Use cases

• SaaS teams building RAG over internal docs (policies, product specs, SOPs) stored in Google Drive.
• Automation engineers standardizing document ingestion to preserve context and reduce retrieval ambiguity.
• Operations and support powering chat or search experiences that cite the most relevant section-level context.

Technical details

• Workflow type: n8n automation workflow for RAG ingestion
• Trigger: Manual trigger
• Integrations / nodes used: Google Drive, Code, Loop, Agent, Set (plus split out, sticky note)
• RAG destination: Pinecone vector store
• LLM: OpenAI GPT-4.0-mini via OpenRouter for chunk contextualization