🤖 Multi-Step Agent with ChromaDB RAG

A powerful multi-step agent that implements Retrieval-Augmented Generation (RAG) using ChromaDB, Groq, and Helicone for comprehensive observability.

✨ Features

• 🧠 Multi-step reasoning process
• 📚 ChromaDB vector database integration
• 🔍 Query classification
• 🚀 Automatic knowledge base seeding
• ⏱️ Tool usage (knowledge retrieval and current time)
• 📊 Detailed Helicone session tracking
• 🌐 Simple REST API interface

🛠️ Setup

Prerequisites

• Node.js (v16+)
• npm or yarn
• Groq API key
• Helicone API key
• ChromaDB (running on port 8000)

Installation

1. Clone the repository

git clone https://github.com/rubaiat-hossain/contextual-qa-agent.git
cd contextual-qa-agent

2. Install dependencies

npm install

3. Create a .env file with your API keys

GROQ_API_KEY=your_groq_api_key
HELICONE_API_KEY=your_helicone_api_key

4. Start ChromaDB

You need a running ChromaDB instance. You can start one with Docker:

docker run -p 8000:8000 ghcr.io/chroma-core/chroma:latest

🚀 Running the Server

Start the server with:

npx tsx main.ts

The server will be running at http://localhost:3000, and will automatically seed ChromaDB with initial knowledge if the collection is empty.

📝 API Usage

Query the Agent

curl -X POST http://localhost:3000/analyze \
  -H "Content-Type: application/json" \
  -d '{"text": "what is ai?"}'

Sample Queries

• "What is AI?"
• "Tell me about Helicone"
• "What is RAG?"
• "What is the Model Context Protocol?"
• "What time is it?"

Health Check

curl http://localhost:3000/health

🧩 How It Works

The agent follows a 4-step process:

1. 🔍 Classification: Determines if the query is a question or general statement
2. 🔎 Knowledge Retrieval or Tool Use:
   • For questions: Retrieves information from ChromaDB
   • For general queries: Gets current time
3. 🤔 Reasoning: Develops a plan for answering based on retrieved information
4. 💬 Response Generation: Creates a helpful, natural response

All steps are tracked in Helicone for observability with session paths:

• /classify
• /knowledge-retrieval
• /reasoning
• /final-response

📊 Helicone Integration

This project includes detailed Helicone logging with:

• Session tracking
• Tool usage logging
• Custom properties for filtering and analysis

Visit your Helicone dashboard to view detailed analytics on your agent's performance.

💾 ChromaDB Knowledge Base

The system automatically seeds ChromaDB with initial knowledge about:

• AI fundamentals
• Helicone platform
• RAG architecture
• Observability in AI
• Model Context Protocol (MCP)
• LLMOps

You can extend the knowledge base by modifying the seedKnowledgeBase() function.

📄 License

MIT