satyam-k/llm-code-agent

llm-code-agent

My Python agent for the LLM code project.

LLM Code Deployment Agent

This project is a FastAPI application designed to automatically receive tasks, generate code using a Large Language Model (Gemini), and deploy the resulting static site to GitHub Pages.

It handles multi-round tasks, allowing for initial creation (Round 1) and subsequent revisions (Round 2+).

Features

• —FastAPI Endpoint: A single endpoint (/api/process-task) to receive JSON task payloads.
• —Background Processing: Immediately responds 200 OK and performs all work in the background.
• —LLM Code Generation: Uses the Gemini API to generate index.html, README.md, and LICENSE files.
• —Multimodal Input: Can process image attachments as context for the LLM.
• —GitHub Integration:
• —Automatically creates a new public repository (Round 1).
• —Clones and updates existing repositories (Round 2+).
• —Uses GitPython for local git operations.
• —Uses httpx for direct GitHub API calls (repo creation, Pages activation).
• —GitHub Pages: Automatically enables or updates the GitHub Pages deployment.
• —Robust Error Handling: Includes retries with exponential backoff for all external API calls (LLM, GitHub, Notification).

Deployment (Hugging Face)

This app is designed to be deployed on Hugging Face Spaces.

1. 1.Create a new Space on Hugging Face, selecting "Docker" as the SDK.
2. 2.Push this code to the Hugging Face git repository.
3. 3.Go to the Space's Settings page.
4. 4.Add your secrets (GOOGLE_API_KEY, GITHUB_TOKEN, GITHUB_USER, STUDENT_SECRET) under "Space secrets".
5. 5.The Space will build the Dockerfile and start the app automatically.
6. 6.Your public URL will be https://[your-space-name].hf.space/api/process-task.

How to Run (For Local Testing)

1. 1.Create `.env` file:
2. 2.Copy .env.example to .env.
3. 3.Fill in your GOOGLE_API_KEY, GITHUB_TOKEN, GITHUB_USER, and STUDENT_SECRET.

1. 1.Build the Docker Container:

    docker build -t llm-agent .

1. 1.Run the Docker Container:

    # This maps port 8000 on your laptop to port 7860 inside the container
    docker run -p 8000:7860 --env-file .env llm-agent

1. 1.Test:
2. 2.Your server is now running at http://localhost:8000.