LangGraph Agent HR Assistant

A sophisticated HR assistant powered by LangChain, LangGraph, and multiple LLM models. This project demonstrates the implementation of a conversational AI system that can help with HR-related queries and team building recommendations.

Project Overview

This project implements an intelligent HR assistant that can:

• Process natural language queries about employees and team building
• Access and search through employee records
• Provide recommendations for team composition
• Maintain conversation context across multiple interactions

Technology Stack

Core Technologies

• TypeScript/Node.js - Main development language and runtime
• Express.js - Web server framework
• MongoDB Atlas - Database with vector search capabilities
• LangChain - LLM framework for building AI applications
• LangGraph - Framework for building stateful AI agent workflows

LLM Models Used

1. Claude 3 Sonnet (Anthropic) - Primary model for agent reasoning and responses
2. Mistral 7B Instruct - Used via OpenRouter for synthetic data generation
3. OpenAI - Used for embeddings in vector search

Key Dependencies

• @langchain/langgraph - For building stateful agent workflows
• @langchain/anthropic - Claude integration
• @langchain/openai - OpenAI integration
• @langchain/mongodb - MongoDB vector store integration
• @langchain/langgraph-checkpoint-mongodb - State persistence for conversations
• mongodb - MongoDB client
• zod - Runtime type validation
• express - Web server
• dotenv - Environment variable management

Project Structure

langraph-agent/
├── index.ts           # Main Express server setup
├── agent.ts           # LangGraph agent implementation
├── seed-database.ts   # Database seeding script
├── .env              # Environment variables
└── package.json      # Project dependencies

Component Details

• index.ts: Entry point that sets up the Express server and API endpoints for chat interactions
• agent.ts: Implements the LangGraph agent with tool definitions and conversation workflow
• seed-database.ts: Generates synthetic employee data using Mistral AI and seeds the MongoDB database

Key Features

1. Stateful Conversations: Uses LangGraph and MongoDB to maintain conversation context
2. Vector Search: Implements semantic search over employee records using MongoDB Atlas Vector Search
3. Rate Limiting: Implements intelligent rate limiting with exponential backoff
4. Tool-based Architecture: Modular design with specialized tools for different operations
5. Type Safety: Comprehensive TypeScript types and Zod schemas for runtime validation

Purpose and Use Cases

This system is designed to assist HR professionals and team leaders with:

1. Team building and composition
2. Employee information retrieval
3. Talent gap analysis
4. Resource allocation
5. Employee data management

The combination of multiple LLM models allows for:

• Cost-effective synthetic data generation (Mistral)
• High-quality reasoning and responses (Claude)
• Efficient semantic search capabilities (OpenAI embeddings)

Environment Setup

The project requires several API keys and configuration values:

• OPENAI_API_KEY - For embeddings

• OPENROUTER_API_KEY - For accessing Mistral AI

• ANTHROPIC_API_KEY - For Claude access

• MONGODB_ATLAS_URI - MongoDB connection string

• LANGSMITH_API_KEY - For LangSmith tracing (optional)

• Start a new conversation:

curl -X POST -H "Content-Type: application/json" -d '{"message": "Your message here"}' http://localhost:3000/chat

• Continue Conversation:

curl -X POST -H "Content-Type: application/json" -d '{"message": "Your follow-up message"}' http://localhost:3000/chat/{threadId}

API Endpoints

• GET / - Health check endpoint
• POST /chat - Start a new conversation
• POST /chat/:threadId - Continue an existing conversation

Getting Started

1. Clone the repository
2. Install dependencies: npm install
3. Set up environment variables in .env
4. Seed the database: ts-node seed-database.ts
5. Start the server: ts-node index.ts

The server will start on port 3000 by default.