AI Agent - Intelligent Automation Platform

A comprehensive AI-powered automation platform that combines natural language processing, document analysis, log analysis, calendar management, and GitHub PR review capabilities using Claude AI (Anthropic) and LangChain.

🚀 Overview

This AI-assistant is designed to facilitate efficient machine learning model training (RAG-based, storing issues and resolutions in a local vector database), automate pull request reviews with detailed comments on code quality and optimization, manage Gmail calendar, compose and send/delete mail with voice-controlled instructions, and summarize logs/documents/reports.

It uses Claude AI (Anthropic's state-of-the-art LLM) for fast, reliable, and intelligent responses. Python (as the backend) wrapped in FastAPI and React as the Frontend.

👁️ Preview

Features

🤖 AI Chat Agent

• Natural language interface powered by Claude AI (Anthropic)
• Context-aware conversations with superior reasoning
• Mathematical computation support
• Document Q&A capabilities
• Faster response times compared to local LLMs

📄 Document Analysis

• Upload and analyze PDF and text documents
• AI-powered document summarization
• Question-answering on uploaded documents

📊 RAG-based Log Analyzer

• Upload and index log files
• Semantic search across logs using FAISS vector store
• Natural language queries for log analysis
• Support for multiple log files with consolidated indexing

📅 Google Calendar Integration

• OAuth 2.0 authentication
• Create, view, and delete calendar events
• Schedule meetings with Google Meet links
• Retrieve events by date
• Support for attendees and locations

📧 Gmail Integration

• Read and summarize emails
• Mark emails as read/unread
• AI-powered email summarization

🔍 GitHub PR Review

• Automated pull request review
• Code quality analysis
• Vulnerability detection
• Best practices suggestions
• AI-generated review comments

🧠 RAG based Training & Knowledge Base

• Save issue-resolution pairs
• RAG based engine selects three most relevant solutions from the local saved resolution history, before sending to the LLM for final verdict.
• Training history management

Architecture

AI-python/
├── main_fastapi.py           # Main FastAPI backend server
├── math_ai_agent_doc.py       # LLaMA 3 agent with tool support
├── rag_log_analyzer.py        # RAG-based log analysis
├── training_store.py          # Issue resolution training store
├── pr_review.py               # GitHub PR review automation
├── gmail_auth.py              # Gmail OAuth authentication
├── ai-agent-ui/               # React frontend
├── logs/                      # Log files directory
├── embeddings/                # FAISS vector store
├── uploaded_docs/             # Uploaded documents storage
└── credentials_*.json         # OAuth credentials

Prerequisites

System Requirements

• Python 3.10+
• Node.js 16+ (for frontend)
• Claude CLI (Claude Code) - Command-line interface for Claude AI

Claude CLI Setup

This application now uses the Claude CLI instead of the Claude API for better integration and ease of use.

1. Install Claude CLI:

   # Using npm (recommended)
   npm install -g @anthropic-ai/claude-code
   
   # Verify installation
   claude --version

2. Authenticate with Claude:

   # Sign in to Claude CLI
   claude auth login
   
   # Test the connection
   claude chat -m "Hello, Claude!"

3. Important Note:

   • If running inside a Claude Code session, you'll need to unset the CLAUDECODE environment variable when starting your backend
   • The backend will automatically handle this in most cases

Why Claude CLI instead of local LLMs or API?

• 10-100x faster than local Ollama/LLaMA
• No GPU required - runs entirely in the cloud
• Superior reasoning and code understanding
• More reliable responses with better formatting
• No API key management - authentication handled by CLI
• Lower resource usage on your machine

Installation

Quick Setup

Use the automated setup script:

./setup.sh

Or follow manual installation:

Backend Setup

1. Clone the repository:

git clone https://github.com/sandeepknd/AI-agents.git
cd AI-agents

2. Install Claude CLI (if not already installed):

npm install -g @anthropic-ai/claude-code
claude auth login

3. Install Python dependencies:

pip install -r requirements.txt

4. Set up environment variables:

# Copy the example file (if it exists)
cp .env.example .env 2>/dev/null || touch .env

# Edit .env and add your tokens:
# GITHUB_TOKEN=your_github_token_here (required for PR review)
# HF_TOKEN=your_huggingface_token (optional, for better rate limits)
#
# NOTE: ANTHROPIC_API_KEY is NO LONGER NEEDED - we use Claude CLI instead

5. Configure Google OAuth:
   • Go to Google Cloud Console
   • Create a project and enable Google Calendar API and Gmail API
   • Create OAuth 2.0 credentials
   • Download credentials and save as:
     • credentials_calendar.json (for Calendar)
     • credentials_per_gmail.json (for Gmail)

Frontend Setup

cd ai-agent-ui
npm install

Running the Application

Start Backend Server

# From the root directory
# If running OUTSIDE Claude Code session:
uvicorn main_fastapi:app --reload --host 0.0.0.0 --port 8000

# If running INSIDE Claude Code session (unset CLAUDECODE variable):
unset CLAUDECODE && uvicorn main_fastapi:app --reload --host 0.0.0.0 --port 8000

The API will be available at:

• API: http://localhost:8000
• API Documentation: http://localhost:8000/docs

Note: The first request may take a few seconds as the Claude CLI initializes.

Start Frontend

# From ai-agent-ui directory
cd ai-agent-ui
npm start

The UI will be available at: http://localhost:3000

API Endpoints

Chat & AI Agent

• POST /ask - Send queries to the AI agent

  {
    "query": "What is the derivative of x^2?"
  }

Document Management

• POST /upload - Upload and analyze documents
• POST /upload-log - Upload log files for indexing
• POST /analyze-log - Query log files

Calendar Operations

• GET /authorize-calendar - Start OAuth flow
• GET /oauth2callback - OAuth callback
• GET /get-events - Get upcoming events
• POST /create-event - Create calendar event
• GET /get-events-by-date?date=YYYY-MM-DD - Get events for specific date
• POST /schedule-meeting - Schedule meeting with Google Meet

  {
    "title": "Team Standup",
    "start_time": "2024-03-15T10:00:00",
    "end_time": "2024-03-15T11:00:00",
    "description": "Daily standup meeting",
    "attendees": ["user@example.com"],
    "create_meet_link": true
  }

• DELETE /delete-event?event_id=xxx - Delete event

Training & Knowledge Base

• POST /train-model - Save issue-resolution pair

  {
    "issue": "Server not responding",
    "resolution": "Restart the service using systemctl restart app"
  }

• POST /suggest-resolution - Get AI-powered resolution suggestions
• GET /get-training-history - View training history
• DELETE /clear-training-history - Clear training data

GitHub PR Review

• POST /webhook - GitHub webhook for PR events
• POST /comment - Post comment on PR

  {
    "pr_url": "https://github.com/user/repo/pull/123",
    "comment": "LGTM! Great work on this feature."
  }

• POST /generate-comment - Generate AI review comment

  {
    "pr_url": "https://github.com/user/repo/pull/123"
  }

Usage Examples

Chat with AI Agent

curl -X POST http://localhost:8000/ask \
  -H "Content-Type: application/json" \
  -d '{"query": "What is 15 factorial?"}'

Upload and Analyze Log

curl -X POST http://localhost:8000/upload-log \
  -F "file=@application.log"

curl -X POST http://localhost:8000/analyze-log \
  -H "Content-Type: application/json" \
  -d '{"query": "Show me all error messages"}'

Schedule a Meeting

curl -X POST http://localhost:8000/schedule-meeting \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Project Review",
    "start_time": "2024-03-20T14:00:00",
    "end_time": "2024-03-20T15:00:00",
    "attendees": ["team@example.com"],
    "create_meet_link": true
  }'

Generate PR Review

curl -X POST http://localhost:8000/generate-comment \
  -H "Content-Type: application/json" \
  -d '{"pr_url": "https://github.com/owner/repo/pull/42"}'

Technology Stack

Backend

• FastAPI - Modern web framework for APIs
• LangChain - LLM orchestration framework
• Claude CLI (Anthropic) - State-of-the-art LLM via command-line interface
• FAISS - Vector similarity search
• Sentence Transformers - Text embeddings
• Google API Client - Calendar and Gmail integration

Frontend

• React 19 - UI framework
• Axios - HTTP client
• Tailwind CSS - Styling
• Framer Motion - Animations
• React Icons - Icon library

AI/ML

• Claude 3.5 Sonnet - Latest Claude model for chat and reasoning
• all-MiniLM-L6-v2 - HuggingFace embedding model for RAG
• sentence-transformers - Sentence similarity model

Configuration

Environment Variables

# .env file configuration
GITHUB_TOKEN=your_github_personal_access_token  # Required for PR review
HF_TOKEN=your_huggingface_token  # Optional, for higher rate limits

# NOTE: ANTHROPIC_API_KEY is NO LONGER needed
# Authentication is handled by Claude CLI (claude auth login)

Calendar OAuth Scopes

The application requests the following Google Calendar scopes:

• https://www.googleapis.com/auth/calendar

Gmail OAuth Scopes

For Gmail integration:

• https://www.googleapis.com/auth/gmail.modify
• https://www.googleapis.com/auth/gmail.readonly

Troubleshooting

Claude CLI Issues

Test Claude CLI connection:

# Check if Claude CLI is installed
claude --version

# Test basic functionality
claude chat -m "Hello, Claude!"

# Check authentication status
claude auth status

Common Errors

"Claude CLI command 'claude' not found"

# Install Claude CLI
npm install -g @anthropic-ai/claude-code

# Check if it's in PATH
which claude

# If not in PATH, add to your shell profile (~/.bashrc or ~/.zshrc):
export PATH="$PATH:$HOME/.npm-global/bin"

"Cannot be launched inside another Claude Code session"

# When starting the backend, unset the environment variable:
unset CLAUDECODE && uvicorn main_fastapi:app --reload --host 0.0.0.0 --port 8000

"Authentication failed" or "Not logged in"

# Re-authenticate with Claude CLI
claude auth logout
claude auth login

"Timeout" or "Slow responses"

• First request may take longer (5-10 seconds) as CLI initializes
• Subsequent requests should be faster (1-3 seconds)
• Check your internet connection

Import Errors

If you encounter LangChain import errors, ensure you have the correct packages:

pip install --upgrade langchain langchain-community langchain-text-splitters

Testing the Integration

Run the test script to verify Claude CLI integration:

# From the project root
python test_claude_cli.py

Port Already in Use

# Find process using port 8000
lsof -i :8000

# Kill the process
kill -9 <PID>

Development

Running Tests

# Frontend tests
cd ai-agent-ui
npm test

# Backend tests (if implemented)
pytest

Code Style

The project follows PEP 8 for Python and Prettier for JavaScript/React.

Security Considerations

1. Credentials: Never commit credentials_*.json, token.json, or .env files to version control
2. OAuth Tokens: Tokens are stored locally in token.json and token_mail.pickle
3. GitHub Token: Store in environment variable or .env file
4. Claude Authentication: Managed by Claude CLI - credentials stored in ~/.claude/
5. CORS: Frontend is restricted to http://localhost:3000 in production

Contributing

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

License

[Specify your license here]

Acknowledgments

• Anthropic for Claude AI - the most capable AI assistant
• LangChain for the excellent LLM framework
• FastAPI for the modern Python web framework
• React community for the frontend ecosystem

Support

For issues, questions, or contributions, please open an issue on GitHub.

---

Built with ❤️ using Claude AI, LangChain, and FastAPI