Hugging Face's logo

Spaces:
arnavk
/
openenv-code-debugger
Sleeping

App Files Community
openenv-code-debugger
File size: 3,246 Bytes
b95e073
 
 
 
 
 
 
 
 
 
c14504c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
---
title: OpenEnv Code Debugger
emoji: 🐛
colorFrom: blue
colorTo: green
sdk: docker
pinned: false
app_port: 7860
---

# Code Debug OpenEnv

An **OpenEnv-compatible environment** for the Meta x PyTorch Hackathon where an AI agent debugs broken Python code.

## Overview

The agent receives buggy Python code and test descriptions, submits fixes, and is rewarded by the fraction of tests passing (0.0–1.0). The episode ends when all tests pass or the step limit is reached.

## Tasks

| Task | Difficulty | Bug Type |
|------|-----------|----------|
| task_001_off_by_one | Easy | Fibonacci returns wrong variable |
| task_002_wrong_operator | Easy | `<` instead of `>` in find_max |
| task_003_mutable_default | Medium | Mutable default argument in list builder |
| task_004_scope_bug | Medium | Closure captures loop variable by reference |
| task_005_binary_search | Hard | Binary search boundary bugs |
| task_006_graph_cycle | Hard | DFS cycle detection missing recursion stack |

## API Endpoints

| Method | Path | Description |
|--------|------|-------------|
| GET | `/health` | Health check |
| GET | `/tasks` | List all available tasks |
| POST | `/reset` | Start a new episode |
| POST | `/step/{episode_id}` | Submit fixed code |
| GET | `/state/{episode_id}` | Get episode metadata |

## Reward

```
reward = tests_passed / total_tests  # range: 0.0 – 1.0
done   = reward == 1.0 OR step_count >= max_steps
```

## Setup & Run

### Local (development)

```bash
pip install fastapi uvicorn pydantic httpx openai
cd Desktop/Meta
uvicorn code_debug_env.server.app:app --host 0.0.0.0 --port 7860 --reload
```

### Docker

```bash
cd Desktop/Meta/code_debug_env
docker build -t code-debug-env -f server/Dockerfile ..
docker run -p 7860:7860 code-debug-env
```

## Inference Script

```bash
export API_BASE_URL="https://router.huggingface.co/v1"
export MODEL_NAME="Qwen/Qwen2.5-72B-Instruct"
export HF_TOKEN="your_token"
export ENV_URL="http://localhost:7860"  # or HF Space URL

python inference.py
```

### Expected output format

```
[START] task=task_001_off_by_one env=http://localhost:7860 model=Qwen/Qwen2.5-72B-Instruct
[STEP] step=1 action='def fib...' reward=1.00 done=true error=null
[END] success=true steps=1 score=1.000 rewards=1.00
```

## Environment Variables

| Variable | Required | Description |
|----------|----------|-------------|
| `API_BASE_URL` | Yes | LLM API endpoint |
| `MODEL_NAME` | Yes | Model identifier |
| `HF_TOKEN` | Yes | Hugging Face / API key |
| `ENV_URL` | No | OpenEnv server URL (default: http://localhost:7860) |

## Project Structure

```
code_debug_env/
├── models.py          # Pydantic models (DebugAction, DebugObservation, DebugState)
├── client.py          # HTTP client wrapper
├── openenv.yaml       # Environment manifest
├── pyproject.toml     # Package metadata
├── tasks/             # Task definitions (JSON)
│   ├── easy/
│   ├── medium/
│   └── hard/
└── server/
    ├── environment.py # Core logic (reset/step/state)
    ├── executor.py    # Safe subprocess code runner
    ├── app.py         # FastAPI server
    └── Dockerfile
inference.py           # Root-level inference script
```