Workflow Agent

A full-fledged AI agent capable of orchestrating and executing complex workflows with integrated voice capabilities. This agent uses natural language processing, intent recognition, and dynamic task planning to execute multi-step operations autonomously.

Features

• Workflow Orchestration: Plan and execute complex, multi-step workflows autonomously
• Voice Interface: Natural voice input and output capabilities
• Intent Recognition: Intelligent parsing of user intent using advanced NLP
• Dynamic Planning: Automatic task planning and replanning based on execution feedback
• Tool Integration: Extensible tool registry for integrating external services (calendar, etc.)
• Memory Management: Long-term and short-term memory for context awareness
• State Management: Robust state tracking throughout agent execution
• LLM Integration: Powered by large language models for natural interactions

Project Structure

agent/          - Core agent logic (executor, planner, intent parser, etc.)
app/            - Application setup and configuration
memory/         - Long-term and short-term memory systems
prompts/        - LLM prompt templates for various agent functions
schemas/        - Data models for execution, intents, plans, and tools
tools/          - Tool registry and integrations (e.g., calendar)
utils/          - Utilities (LLM, logging, parsing, prompt loading)
tests/          - Test suite for agent components

Installation

1. Clone the repository:

git clone <repository-url>
cd voice-agent

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up credentials:

• Place your API credentials in credentials.json
• Generate and store authentication tokens in token.json
• Configure environment variables in .env (see .env.example if available)

Usage

Start the workflow agent:

python -m app.main

Architecture

The agent operates through several core components:

• Intent Parser: Analyzes user input to extract intent and parameters
• Planner: Generates execution plans based on recognized intents
• Replanner: Adjusts plans based on execution feedback
• Executor: Carries out planned tasks using available tools
• Memory Systems: Maintains context across conversations and sessions
• Tool Registry: Manages available tools and their integrations

Configuration

Edit app/config.py to customize:

• Agent behavior and parameters
• LLM model selection and settings
• Tool availability and settings
• Memory constraints and parameters

Testing

Run tests with:

pytest tests/

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues for bugs and feature requests.

License

This project is licensed under the MIT License.

See the LICENSE file for details.