Meta x PyTorch Hackathon OpenEnv Compatible

AI-Powered Code Debugging Arena

A real-world OpenEnv environment where AI agents learn to fix broken Python code. Submit buggy code, get instant fixes, and watch tests pass in real time.

Live
Environment Status
-
Debug Tasks
0.0 – 1.0
Reward Range
3
Difficulty Levels

Live Debug Arena

Pick a task, review the buggy code, write your fix, and submit. Watch tests execute in real time.

Buggy Code
Select a task and click "Load Task" to begin.
Your Fix

How It Works

1

Reset Environment

The agent receives buggy Python code and descriptions of the tests it must pass.

2

Submit a Fix

The agent analyzes the code, identifies the bug, and submits corrected code via step().

3

Earn Reward

Tests run in a sandboxed subprocess. Reward = fraction passing (0.0–1.0). Iterate until all pass.

Task Explorer

6 real-world debugging challenges across 3 difficulty levels.

Architecture

🤖
AI Agent
LLM via OpenAI Client
step(code)
OpenEnv API
🛠
Executor
Subprocess + Timeout
Tests
Pass / Fail
🏆
Reward
0.0 – 1.0
reset() → Observation  |  step(action) → StepResult  |  state() → State

Performance

6/6
Tasks Solved by Baseline
1.000
Average Score
< 2s
Avg Response Time
100%
OpenEnv Spec Compliant