Agentic Personal Assistant

A full-stack agentic RAG (Retrieval-Augmented Generation) application that allows users to upload PDF documents and chat with them using an intelligent AI agent.

🚀 Features

• PDF Document Ingestion: Upload and process PDF files into a vector database
• Agentic Chat: Intelligent agent that decides when to search the knowledge base
• Conversation Memory: Maintains context across multiple questions
• Modern UI: ChatGPT-like interface with file upload capabilities
• Observability: Integrated with LangSmith for monitoring and tracing

🏗️ Architecture

Backend (Node.js/Express)

• Server: Express.js API server with CORS and file upload support
• Agent: LangChain ReAct agent with OpenAI GPT-4o
• Vector Database: Pinecone for document storage and similarity search
• Embeddings: Pinecone-hosted llama-text-embed-v2 model
• Observability: LangSmith for tracing and monitoring

Frontend (React/Vite)

• UI Framework: React with Vite for fast development
• Styling: ChatGPT-inspired dark theme interface
• File Upload: Drag-and-drop PDF upload with progress feedback
• Chat Interface: Real-time messaging with typing indicators

📋 Prerequisites

• Node.js (v18 or higher)
• npm or yarn
• Pinecone account with index created
• OpenAI API key
• LangSmith account (optional, for observability)

🛠️ Setup

1. Clone the repository

   git clone <repository-url>
   cd agentic-personal-assistant

2. Install dependencies

   npm run install:all

3. Environment Configuration

   cp .env.example .env

   Edit .env with your API keys:

   # LLM
   OPENAI_API_KEY=your_openai_api_key
   
   # Vector DB (Pinecone)
   PINECONE_API_KEY=your_pinecone_api_key
   PINECONE_INDEX=your_pinecone_index_name
   
   # LangSmith tracing
   LANGSMITH_TRACING=true
   LANGSMITH_ENDPOINT=https://api.smith.langchain.com
   LANGSMITH_API_KEY=langsmith_key
   LANGSMITH_PROJECT="Project name"

🚀 Running the Application

Development Mode

npm run dev

This starts both the server (port 3001) and client (port 5173) concurrently.

Individual Services

# Server only
npm run dev:server

# Client only  
npm run dev:client

📁 Project Structure

agentic-personal-assistant/
├── server/                 # Backend API server
│   ├── index.js           # Express server and API routes
│   ├── agent.js           # Agent logic and memory management
│   ├── tools.js           # Knowledge base search tool
│   ├── ingest.js          # PDF ingestion pipeline
│   └── package.json       # Server dependencies
├── client/                # Frontend React app
│   ├── src/
│   │   ├── App.jsx        # Main application component
│   │   ├── App.css        # ChatGPT-like styling
│   │   └── main.jsx       # React entry point
│   └── package.json       # Client dependencies
├── .env.example           # Environment variables template
├── package.json           # Root package with scripts
└── README.md              # This file

🔄 How It Works

Document Ingestion

1. User uploads PDF via frontend
2. Server receives file and extracts text using PDFLoader
3. Text is split into chunks (1000 chars with 200 overlap)
4. Chunks are converted to embeddings using Pinecone's model
5. Embeddings are stored in Pinecone vector database

Chat Flow

1. User sends a message
2. Agent receives message with conversation history
3. Agent decides whether to search the knowledge base
4. If needed, searches Pinecone for relevant document chunks
5. Agent uses retrieved context to generate response
6. Response is sent back to user and added to conversation history

🔧 Key Components

Agent (server/agent.js)

• ReAct agent using LangChain's createAgent
• MemorySaver for conversation persistence
• Tool calling for knowledge base search

Search Tool (server/tools.js)

• Pinecone vector store integration
• Similarity search with top-k results
• Lazy initialization for environment variables

Ingestion Pipeline (server/ingest.js)

• PDF text extraction and chunking
• Batch processing (96 chunks per API call)
• Pinecone upsert operations

Frontend (client/src/App.jsx)

• React state management for chat and upload
• File upload with progress feedback
• Real-time chat interface with auto-scroll

🐛 Troubleshooting

Common Issues

1. Connection Refused Error

   • Ensure server is running on port 3001
   • Check for port conflicts: lsof -i :3001

2. Environment Variables Missing

   • Verify .env file exists in server directory
   • Check all required API keys are set

3. Pinecone API Errors

   • Verify Pinecone index exists
   • Check API key permissions
   • Ensure embedding model matches ingestion/retrieval

4. Dependency Installation Errors

   • Use --legacy-peer-deps flag for peer dependency conflicts
   • Clear node_modules and reinstall if needed

📚 Technologies Used

• Backend: Node.js, Express.js, Multer
• Frontend: React, Vite
• AI/ML: LangChain, OpenAI GPT-4o
• Vector DB: Pinecone
• Observability: LangSmith
• Development: Concurrently, ESLint

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

📄 License

This project is licensed under the MIT License.