RAG AI Agent with Milvus & Cohere

This template sets up a Retrieval-Augmented Generation (RAG) AI agent that combines Cohere embeddings, Milvus vector database, and OpenAI GPT-4o for intelligent, context-aware conversations. It automatically ingests new PDF files uploaded to a Google Drive folder, indexes them into Milvus, and makes the content instantly searchable for the AI agent.

With multilingual embedding support from Cohere, scalable storage with Milvus (via Zilliz Cloud), and memory-enabled conversational context, this workflow provides a production-ready setup for enterprise-grade document-based AI assistants.

🚀 Features

• Automated Document Ingestion
  • Watches a Google Drive folder for new PDFs and processes them automatically.
• Smart File Processing
  • Extracts text, splits into chunks, and generates vector embeddings using Cohere.
• Vector Database Storage
  • Inserts embeddings into Milvus, optimized for large-scale and high-performance similarity search.
• Conversational RAG Agent
  • Uses OpenAI GPT-4o with memory and Milvus retrieval to provide accurate, context-driven answers.
• Multilingual Support
  • Cohere embed-multilingual-v3.0 model enables cross-language document understanding.
• Scalable Infrastructure
  • Leverages Zilliz Cloud for managed Milvus hosting, with cost calculator integration.
• Chat-ready Deployment
  • Trigger-based chat agent that responds in real time with retrieved knowledge.