Codecraft AI

A multi-agent AI system that generates code through a collaborative workflow using OpenRouter API. The system uses 5 specialized agents working together to design, implement, test, review, and approve code.

Features

• 5-Agent Workflow: Architect → Coder → Tester → Reviewer → Manager
• Modern Web UI: Interactive interface to test the API directly from browser
• OpenRouter Integration: Access to multiple AI models through unified API
• RESTful API: Clean API endpoints for integration
• Full-Width Code Display: Optimized for viewing generated code

Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Set Up Environment Variables

Create a .env file in the project root:

OPENROUTER_API_KEY=sk-or-v1-your-key-here
OPENROUTER_MODEL=openai/gpt-4o-mini

Get your API key from: OpenRouter Settings

3. Run the Server

Option 1: Using the run script

python run_server.py

Option 2: Using uvicorn directly

uvicorn api.main:app --reload

4. Access the Application

• Web UI: http://127.0.0.1:8000/
• API Docs: http://127.0.0.1:8000/docs
• ReDoc: http://127.0.0.1:8000/redoc

Agent Workflow

The system uses a sequential workflow with conditional looping:

1. Architect - Designs high-level architecture and system structure
2. Coder - Implements the code based on architecture
3. Tester - Creates comprehensive test cases
4. Reviewer - Reviews code quality and test coverage
5. Manager - Makes final decision (approve/rewrite)

If the manager decides to rewrite, the process loops back to the coder for improvements.

API Endpoints

Generate Code

POST /api/v1/generate

Request:

{
  "task": "Create a Python function to calculate fibonacci numbers"
}

Response:

{
  "architecture": "...",
  "code": "...",
  "tests": "...",
  "review": "...",
  "final_decision": "approve"
}

Health Check

GET /health

Status

GET /status

Project Structure

.
├── api/
│   ├── main.py          # FastAPI application with web UI
│   └── routes.py        # API routes
├── agents/
│   ├── architect.py     # Architect agent
│   ├── coder.py         # Coder agent
│   ├── tester.py        # Tester agent
│   ├── reviewer.py      # Reviewer agent
│   └── manager.py       # Manager agent
├── orchestration/
│   ├── graph.py         # LangGraph workflow definition
│   └── state.py         # State definition
├── config.py            # Configuration and env loading
├── requirements.txt     # Python dependencies
├── run_server.py        # Server startup script
└── .env                 # Environment variables (create this)

Configuration

OpenRouter API Key

1. Sign up at OpenRouter
2. Get your API key from Settings
3. Add to .env file: OPENROUTER_API_KEY=sk-or-v1-your-key-here

Available Models

You can change the model by updating OPENROUTER_MODEL in your .env file:

• openai/gpt-4o-mini (default, cost-effective)
• openai/gpt-4o
• openai/gpt-4-turbo
• anthropic/claude-3.5-sonnet
• google/gemini-pro

See all models: https://openrouter.ai/models

Credits

Make sure you have credits in your OpenRouter account:

• Check credits: https://openrouter.ai/credits
• Add credits if needed

Usage Examples

Using the Web UI

1. Visit http://127.0.0.1:8000/
2. Enter your task description (e.g., "Create a calculator class")
3. Click "Generate Code"
4. View results: Architecture, Code, Tests, Review, and Final Decision

Using the API

Python:

import requests

response = requests.post(
    "http://127.0.0.1:8000/api/v1/generate",
    json={"task": "Create a Python function to calculate fibonacci numbers"}
)
print(response.json())

cURL:

curl -X POST "http://127.0.0.1:8000/api/v1/generate" \
  -H "Content-Type: application/json" \
  -d '{"task": "Create a Python function to calculate fibonacci numbers"}'

Troubleshooting

API Key Issues

Error: "OPENROUTER_API_KEY not found"

• Make sure .env file exists in project root
• Verify it contains: OPENROUTER_API_KEY=sk-or-v1-...
• Restart the server after updating .env

Error: "API Key Issue"

• Verify your key is active at https://openrouter.ai/settings/keys
• Check you have credits at https://openrouter.ai/credits
• Make sure you're using OPENROUTER_API_KEY (not OPENAI_API_KEY)

Quota Errors

Error: "API Quota Exceeded"

• Add credits to your OpenRouter account
• Check usage at https://openrouter.ai/activity
• Wait for quota reset or upgrade plan

Server Issues

Error: "Attribute api not found"

• Use correct command: uvicorn api.main:app --reload
• Note: Use app not api at the end

Dependencies

• fastapi - Web framework
• uvicorn - ASGI server
• langgraph - Agent orchestration
• langchain-openai - LLM integration
• langchain-core - Core LangChain components
• pydantic - Data validation
• python-dotenv - Environment variable loading

Development

Running in Development Mode

uvicorn api.main:app --reload --host 0.0.0.0 --port 8000

Project Architecture

The system uses LangGraph to orchestrate the agent workflow:

• Each agent is a node in the graph
• State is passed between agents
• Conditional edges handle the rewrite loop
• See orchestration/graph.py for workflow definition

License

This project is open source and available for use.

Support

• OpenRouter Docs: https://openrouter.ai/docs
• OpenRouter Models: https://openrouter.ai/models
• OpenRouter Credits: https://openrouter.ai/credits