Krishna-11Analyst/LLM-code-deployments

LLM Code Deployment System

This is an automated system for building and deploying web applications based on natural language briefs...

🚀 LLM Code Deployment System

Production-ready automated web application generator for IIT Madras TDS Project-1.

Live API: https://Krishna-11Analyst-llm-deployment-final.hf.space

✨ Features

• —✅ Correct API Endpoint: /api-endpoint (matches IIT-M specification)
• —✅ Multi-Round Support: Handles Round 1 (create) and Round 2+ (update)
• —✅ OpenAI GPT-4: Generates complete, working web applications
• —✅ Automatic GitHub Integration: Creates repositories and enables Pages
• —✅ Duplicate Detection: Uses nonce to prevent reprocessing
• —✅ File Attachments: Supports CSV, images, and other data files
• —✅ Evaluation Callback: Automatic notification with retry logic
• —✅ Background Processing: Non-blocking with immediate HTTP 200 response

📡 API Usage

Endpoint

POST /api-endpoint

Request Format

{
  "email": "student@example.com",
  "secret": "your-secret",
  "task": "unique-task-id",
  "round": 1,
  "nonce": "unique-nonce",
  "brief": "Create a Bootstrap page with weather display",
  "checks": [
    "Repo has MIT license",
    "README.md is professional",
    "Page displays weather data"
  ],
  "evaluation_url": "https://evaluation-server.com/notify",
  "attachments": [
    {
      "name": "data.csv",
      "url": "data:text/csv;base64,..."
    }
  ]
}

Response (Immediate)

{
  "status": "accepted",
  "note": "processing round 1 started"
}

Background Process

After returning HTTP 200, the system:

1. 1.Generates code using OpenAI GPT-4 (~30-60 seconds)
2. 2.Creates/updates GitHub repository (~5-10 seconds)
3. 3.Enables GitHub Pages (~10-20 seconds)
4. 4.Sends evaluation callback with repo details

🛠️ Local Development

1. Setup

# Clone or copy project
cd llm-deployment-final

# Create virtual environment
python -m venv venv

# Activate (Windows)
venv\Scripts\activate

# Activate (Mac/Linux)
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

2. Configure Environment

Create .env file with your credentials:

GITHUB_TOKEN=your_github_token
GITHUB_USERNAME=your_username
OPENAI_API_KEY=your_openai_key
USER_SECRET=your_secret

3. Run Server

uvicorn app.main:app --reload

API will be available at: http://localhost:8000

4. Test Endpoint

curl -X POST http://localhost:8000/api-endpoint \
  -H "Content-Type: application/json" \
  -d '{
    "email": "test@example.com",
    "secret": "your_secret",
    "task": "test-001",
    "round": 1,
    "nonce": "test-nonce-001",
    "brief": "Create a simple hello world page",
    "checks": ["Has hello world text"],
    "evaluation_url": "https://httpbin.org/post",
    "attachments": []
  }'

🚀 Deployment (Hugging Face Spaces)

1. Create Space

• —Go to: https://huggingface.co/new-space
• —Name: llm-deployment-final
• —SDK: Docker
• —Hardware: CPU Basic (free tier)

2. Push Code

git init
git add .
git commit -m "Initial deployment"
git remote add space https://huggingface.co/spaces/Krishna-11Analyst/llm-deployment-final
git push space main

3. Configure Secrets

In Space Settings → Repository secrets, add:

• —GITHUB_TOKEN
• —GITHUB_USERNAME
• —OPENAI_API_KEY
• —USER_SECRET

4. Wait for Build

Space will build (~5-10 minutes) and show "Running" status when ready.

📊 Architecture

FastAPI API
    ↓
┌─────────────────────┐
│ /api-endpoint       │ ← Receives JSON POST
│ - Validates secret  │
│ - Returns HTTP 200  │
│ - Starts background │
└─────────────────────┘
         ↓
┌─────────────────────┐
│ Background Process  │
│ 1. OpenAI GPT-4     │ → Generate code
│ 2. GitHub API       │ → Create repo
│ 3. Enable Pages     │ → Deploy site
│ 4. Evaluation POST  │ → Notify server
└─────────────────────┘

🔧 Technology Stack

• —FastAPI - Web framework
• —OpenAI GPT-4 - Code generation
• —PyGithub - GitHub API integration
• —HTTPX - HTTP client for callbacks
• —Docker - Containerized deployment
• —Hugging Face Spaces - Hosting platform

📋 Project Structure

llm-deployment-final/
├── .env                 # Credentials (DO NOT commit!)
├── .gitignore          # Git ignore rules
├── .dockerignore       # Docker ignore rules
├── Dockerfile          # Docker configuration
├── README.md           # This file
├── requirements.txt    # Python dependencies
├── runtime.txt         # Python version
├── test_github.py      # Local testing script
└── app/
    ├── __init__.py     # Package init
    ├── main.py         # FastAPI app & endpoints
    ├── llm_generator.py # OpenAI code generation
    ├── github_utils.py # GitHub operations
    ├── notify.py       # Evaluation callbacks
    └── signature.py    # (Empty placeholder)

✅ Submission Information

🔒 Security Notes

Post-Submission Actions

Within 24 hours of submission, regenerate exposed credentials:

1. 1.GitHub Token:
2. 2.Visit: https://github.com/settings/tokens
3. 3.Delete old token
4. 4.Create new token with repo and workflow scopes
5. 5.Update Hugging Face Space secret

1. 1.OpenAI Key:
2. 2.Visit: https://platform.openai.com/api-keys
3. 3.Revoke old key
4. 4.Create new key
5. 5.Update Hugging Face Space secret

📄 License

MIT License

👨‍💻 Author

Student ID: 23f2001849 Email: 23f2001849@ds.study.iitm.ac.in Course: IIT Madras - Tools in Data Science - Project 1 Date: October 2025

🙏 Acknowledgments

• —OpenAI for GPT-4 API
• —GitHub for repository hosting and Pages
• —Hugging Face for Spaces deployment platform
• —IIT Madras for the learning opportunity

Built with ❤️ for IIT-M TDS Project-1