metadata

title: FallnAI Autonomous Software Developer
emoji: 🚀
colorFrom: pink
colorTo: purple
sdk: docker
app_port: 7860

FallnAI Autonomous Software Development Platform

This is a proof-of-concept autonomous software development platform. It leverages a crew of AI agents using open-source Large Language Models (LLMs) from Hugging Face to write, test, and correct its own code within a secure Docker sandbox. It automatically pushes the final, working code and documentation to a new GitHub repository.

How it works

Frontend: A simple web UI in index.html allows users to submit a software request.
Backend: The Flask application in app.py receives the request.
Agent Crew: The backend uses the CrewAI framework to orchestrate a team of specialized agents (e.g., Planner, Coder, Tester) to execute a sequential development process: plan, code, test, and document.
LLMs & Transformers: Each agent uses an LLM (accessed via the transformers library) as its core reasoning engine to perform its task.
Sandboxing: The generated code is executed inside a temporary, isolated Docker container to prevent any malicious code from affecting the host system.
Correction Loop: The LLM analyzes the output and errors from the sandbox and corrects the code until it passes.
Documentation & Deployment: A dedicated agent writes a README.md file, and a DevOps agent pushes all final files to a new, unique GitHub repository.

Give it a try! You can describe any simple Python script, such as "a program that calculates the factorial of a number."