RAG Context-Aware Chunking: Google Drive to Pinecone via OpenRouter & Gemini

This n8n workflow transforms documents from Google Drive into context-rich chunks stored in Pinecone vector database, improving RAG systems' retrieval precision. It fetches documents, extracts text using boundary markers, splits into logical sections via custom code, and loops through chunks to generate contextual metadata with an OpenRouter Agent (GPT-4o-mini). Context is prepended, embeddings created via Google Gemini, and upserted to Pinecone.

Key benefits include superior retrieval accuracy by preserving document context, reducing hallucinations in AI responses, and streamlining RAG pipeline setup. Ideal for knowledge bases, chatbots, or search apps handling large docs.

Use cases: Building enterprise RAG apps, automating content indexing for Q&A systems, enhancing semantic search in education or research tools. Based on Anthropic's contextual retrieval techniques, it saves hours of manual processing and custom scripting.