README.md

metadata

title: AI Code Review OpenEnv
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: docker
app_file: inference.py
pinned: false

🔍 AI Code Review — OpenEnv RL Environment

Train and evaluate AI agents on real-world code review using structured RL environments.

A real-world reinforcement learning environment where AI agents learn structured code review: identify bugs → reason about them → fix them correctly.

Built for the Meta × Hugging Face × OpenEnv × Scaler Hackathon 2026 (Round 1 Submission).

---

🚀 Why This Matters

Most AI coding tools are built for generation, not review. But in real-world software engineering, that's the wrong priority:

• Code review is the last line of defence against production bugs
• Bugs often slip through because reviewers miss the why, not the what
• Understanding why code is wrong matters more than just patching it

👉 This environment turns code review into a measurable, learnable RL problem — with structure, reward signal, and reproducibility built in from the ground up.

---

🧠 What Makes This Environment Strong

Property	Detail
✅ Real-world task	Code review, not a toy problem
✅ Two-phase reasoning	Agent must identify before it can fix
✅ Deterministic grading	No LLM judge, no randomness
✅ Reward shaping	Partial progress is rewarded at every step
✅ Reproducible evaluation	Same input always produces same score
✅ Spec compliant	Fully compatible with OpenEnv + hackathon requirements

---

🏗️ Environment Design

Implements the full OpenEnv interface — three methods, no more:

obs      = env.reset()       # Start a new episode
result   = env.step(action)  # Execute one action → {observation, reward, done, info}
metadata = env.state         # Inspect episode state

---

📥 Observation Space

Field	Description
code	Buggy JavaScript code to review
task	Task description + current phase
history	Log of all prior steps this episode
task_id	Task number (1, 2, or 3)
language	javascript
difficulty	easy / medium / hard
category	syntax / logic / performance

---

🎮 Action Space

Field	Values
action_type	"identify" or "fix"
content	Free-text explanation or code

---

🔁 Episode Flow

identify → fix → identify → fix → identify → fix
   task 1            task 2            task 3

One episode covers all 3 tasks in order.

---

🧪 Tasks

🟢 Task 1 — Syntax Error (Easy)

// buggy
function add(a, b {
  return a + b;
}

// fixed
function add(a, b) {
  return a + b;
}

---

🟡 Task 2 — Logic Bug (Medium)

// buggy
function isEven(n) {
  return n % 2 === 1;
}

// fixed
function isEven(n) {
  return n % 2 === 0;
}

---

🔴 Task 3 — Performance Issue (Hard)

// buggy
for (let i = 0; i < arr.length; i++) {
  console.log(arr[i]);
}

// fixed
arr.forEach(item => {
  console.log(item);
});

---

🎯 Reward System

✔ Deterministic & Reproducible

No randomness. Same input → same score.

---

🧩 Identify Reward

score = matched_keywords / total_keywords

---

🛠️ Fix Reward

Condition	Score
Exact match	1.0
Whitespace-normalised	0.9
≥ 80% token overlap	0.6
≥ 50% token overlap	0.3
Below 50%	0.0

---

🎁 Identify Bonus

+0.1 bonus if identify ≥ 0.4 (final reward clamped to 1.0)

---

⚠️ Edge Case Handling

Situation	Effect
Skipped identify phase	fix score × 0.5
Unknown action_type	0.0
Repeated identify	0.0

---

📊 Max Reward Per Episode

Scope	Max Reward
Per step	1.0
Per task	~2.0
Full episode	~6.0

---

🤖 Inference Pipeline

1. Read HF_TOKEN, API_BASE_URL, MODEL_NAME
2. Initialise OpenAI client + environment
3. Reset environment
4. Identify → Fix for each task
5. Log results in required format

---

🧾 Log Format (Strict)

[START]
[STEP]
[STEP]
...
[END]

---

📊 Example Output

[START] task=code-review env=ai-code-review-env model=gpt-4.1-mini
[STEP] step=1 action=identify:task1 reward=0.71 done=false error=null
[STEP] step=2 action=fix:task1 reward=1.00 done=false error=null
[STEP] step=3 action=identify:task2 reward=0.60 done=false error=null
[STEP] step=4 action=fix:task2 reward=1.00 done=false error=null
[STEP] step=5 action=identify:task3 reward=0.45 done=false error=null
[STEP] step=6 action=fix:task3 reward=1.00 done=true error=null
[END] success=true steps=6 rewards=0.71,1.00,0.60,1.00,0.45,1.00

---

📌 Why This Stands Out

• Real-world problem (not synthetic)
• Deterministic evaluation (fair + reproducible)
• Reward shaping (better learning signal)
• Two-phase reasoning (prevents shortcutting)

---

⚙️ Setup

pip install -r requirements.txt
export HF_TOKEN=your_token
python inference.py

---

📁 Project Structure

code-review-env/
├── inference.py
├── openenv.yaml
├── Dockerfile
├── README.md
└── env/

---

🏆 Evaluation Alignment

Criteria	Covered
Real-world utility	✅
Task design	✅
Grader quality	✅
Environment design	✅
Spec compliance	✅

---

📄 License

MIT License Built for Meta × Hugging Face × OpenEnv × Scaler Hackathon 2026