Agentic Knowledge System

This project demonstrates a robust architecture for deploying autonomous agents that can:

1. Retrieve Information: Efficiently query vector databases for semantic search.
2. Reason: Use chain-of-thought prompting to synthesize answers from multiple sources.
3. Act: Execute tools to fetch real-time data or perform calculations.

Key Features

• Multi-Agent Orchestration: Uses a supervisor agent to delegate sub-tasks to specialized worker agents.
• Dynamic Tool Selection: Agents can choose from a suite of tools based on the user query.
• Self-Correction: Implements a feedback loop where agents can critique and refine their own outputs.

Technology Stack

• Framework: LangChain & LangGraph
• Vector DB: Pinecone
• LLM: GPT-4-turbo / Claude 3 Opus
• Backend: FastAPI