Keystroke Agent

A Python project implementing an AI assistant with tool integration, AWS service support, and asynchronous event-driven architecture.

Features

1. Advanced AI Integration:

   • Uses AWS Bedrock (Claude 3.5 Sonnet) as the default LLM
   • Asynchronous API calls for better performance
   • Configurable model selection
   • Event-driven architecture using asyncio.Queue for handling LLM responses and tool calls

2. Built-in Tools:

   • Calculator: Mathematical operations and functions
   • File Management: File operations (create, read, update, delete)
   • Presigned URL: AWS S3 presigned URL generation and file transfer

3. Interactive CLI:

   • Rich text formatting for better readability
   • Command history management
   • Dynamic conversation summarization
   • Dot commands for runtime configuration
   • Asynchronous implementation for improved responsiveness

4. Customizable Settings:

   • Model selection
   • System message
   • Assistant name
   • History limit
   • Tool enablement

Installation

pip install keystroke-agent

Configuration

1. Create a .env file with your AWS credentials:

AWS_ACCESS_KEY_ID=your_key_id
AWS_SECRET_ACCESS_KEY=your_secret_key
AWS_SESSION_TOKEN=your_session_token
AWS_DEFAULT_REGION=your_region

2. Adjust settings in keystroke/settings.py as needed:

AGENT_NAME = "AI Assistant"
DEFAULT_LLM_MODEL = "bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0"
HISTORY_LIMIT = 10
ENABLE_TOOLS = True

Usage

Start the agent:

agent

Available Dot Commands

• .help - Show available commands
• .clear - Clear conversation history
• .model <model_name> - Change the LLM model
• .view - View current settings
• .view history - View conversation history
• .system <message> - Change the system message
• .name <new_name> - Change the assistant's name

Available Tools

1. Calculator:

   • Basic arithmetic operations
   • Trigonometric functions
   • Logarithms and exponentials
   • Factorial calculation

2. File Management:

   • Create, read, update, and delete files
   • List files in directories
   • Copy and rename files
   • Recursive operations support

3. AWS S3 Integration:

   • Generate presigned URLs for upload/download
   • Upload files using presigned URLs
   • Download files using presigned URLs

Development

Built with:

• litellm - LLM integration
• rich - Terminal formatting
• boto3 - AWS SDK
• python-dotenv - Environment management
• asyncio - Asynchronous I/O

License

MIT License

Future Enhancements

• Streaming Support
• Asynchronous event-driven architecture
• Additional tool integrations
• Web interface option
• Enhanced error handling
• Conversation export functionality
• Tool usage analytics
• Custom tool loading system

WhatsApp Integration

• WhatsApp Business API integration
  • Message sending and receiving
  • Media file handling (images, documents, voice notes)
  • Group chat support
  • Interactive buttons and list messages
• Automated responses through AI
• Contact management and user session handling
• Message templates and broadcasting
• Multi-language support for WhatsApp messages
• Message status tracking and analytics