Meta-Agent Factory MVP 🚀

A Python-based system for generating agent applications from templates. This MVP provides the core functionality for automated agent generation using Claude Code patterns.

What's Been Built

✅ Core Architecture

• Python 3.12 Backend: Modern Python implementation
• Template-Based Generation: Jinja2 template engine for flexible code generation
• Pydantic Models: Type-safe data validation and serialization
• Modular Design: Clean separation of concerns with core, models, templates, utils

✅ Key Features Implemented

• Agent Specification: Structured input for agent requirements
• Template System: YAML-based template definitions with Jinja2 templating
• Code Generation: Produces executable Python source code
• Documentation Generation: Automatic README and usage guides
• Dependency Management: Automated requirements.txt generation

✅ Working Templates

• Data Analyst Agent: For data processing, analysis, and report generation
• (More templates can be easily added)

Quick Start

Installation

pip install Jinja2 PyYAML pydantic click

Generate Your First Agent

1. Create a specification file (agent_spec.yaml):

agent_type: data_analyst
name: SalesAnalyzer
description: "Analyzes sales data and generates comprehensive reports"
requirements:
  - name: data_loading
    description: "Load sales data from CSV files"
    priority: 1
  - name: sales_analysis
    description: "Analyze sales trends and patterns"
    priority: 1
dependencies:
  - pandas
  - matplotlib
  - seaborn
output_formats:
  - html
  - json

2. Generate the agent:

PYTHONPATH=src python maf_cli.py generate agent_spec.yaml

3. Use the generated agent:

cd output/SalesAnalyzer
pip install -r requirements.txt
python SalesAnalyzer_agent.py

Project Structure

src/
  meta_agent_factory/
    core/           # Core generation logic
    models/         # Data models and schemas
    templates/      # Agent templates (YAML + Jinja2)
    utils/          # Utility functions

Technical Stack

• Python 3.12: Primary implementation language
• Jinja2: Template engine for code generation
• Pydantic: Data validation and serialization
• PyYAML: Template configuration parsing
• Click: Command-line interface

Next Steps

The MVP is fully functional and ready for:

1. Adding more agent templates
2. Enhancing template capabilities
3. Adding web API interface
4. Improving validation and error handling
5. Adding testing and CI/CD pipelines

Usage Examples

See test_spec.yaml for a complete example specification and the generated output in output/SalesAnalyzer/.