RAG-Powered WhatsApp Chatbot for Docs with GPT-4o-mini & MongoDB

This workflow consists of two main parts: Document Ingestion & Indexing, and AI-Powered Query & Response via WhatsApp. The first workflow is manually triggered to import product documentation from Google Docs, splits large documents into searchable chunks, generates vector embeddings using OpenAI, and stores them with metadata in a MongoDB Atlas vector store for efficient semantic search.

The second workflow listens for incoming WhatsApp messages, including text, audio (transcribed), images (analyzed), and documents (parsed). It converts queries to embeddings, performs similarity searches on the MongoDB store, and uses GPT-4o-mini with RAG to generate concise, context-aware answers while maintaining conversation memory across turns.

Benefits include reducing support agent time on manual searches, ensuring consistent and fast answers, and handling multimodal inputs for versatile user interactions. Ideal for tech companies' support teams, product specialists, and knowledge managers automating knowledge bases via WhatsApp.

Use cases: Internal support chatbots for product docs, customer query resolution without human intervention, and scalable AI assistance for documentation-heavy industries like software and hardware.