title	Go Code Review OpenEnv
emoji	🤖
colorFrom	blue
colorTo	green
sdk	docker
sdk_version	1.0
app_file	server/app.py
pinned	false

Go Code Review OpenEnv

An OpenEnv-compatible environment where AI agents review and fix buggy Go code across multiple tasks of increasing complexity.

---

Overview

This project implements a reinforcement learning-style environment for evaluating AI agents on software engineering tasks. The agent is given buggy Go code and must:

• Identify issues in the code
• Provide a corrected version
• Improve performance across iterative steps

The system evaluates the agent using structured rewards based on:

• Bug identification accuracy
• Compilation success
• Test case correctness

---

Features

• OpenEnv-compliant API (/reset, /step, /state)
• Multi-task evaluation (3 tasks)
• Structured reward system (0.0 – 1.0)
• Robust inference pipeline with fallback handling
• Dockerized for deployment (Hugging Face Spaces compatible)

---

Tasks

Task 1: Syntax & Logic Errors

• Fix incorrect logical conditions
• Correct discount calculations

Task 2: Nil Pointer Handling

• Detect and fix nil pointer dereferences
• Ensure safe pointer usage

Task 3: Concurrency & Pagination Bugs

• Fix slice modification issues
• Handle pagination correctly
• Prevent out-of-bounds errors

---

Project Structure

fuzzy-spoon/
│
├── server/
│   ├── app.py              # FastAPI endpoints
│   ├── environment.py      # OpenEnv environment logic
│   ├── grader.py           # Evaluation logic
│   └── tasks/              # Task definitions
│
├── tasks/
│   ├── index.json
│   ├── task1_syntax/
│   ├── task2_pointer/
│   └── task3_concurrency/
│
├── inference.py            # Agent + evaluation loop
├── openenv.yaml            # OpenEnv specification
├── Dockerfile              # Deployment config
└── README.md

---

Installation

1. Clone the repository

git clone <your-repo-url>
cd fuzzy-spoon

---

2. Install dependencies

pip install fastapi uvicorn openai python-dotenv

---

3. Set environment variables

export API_KEY=<your_api_key>
export API_BASE_URL=https://api.cerebras.ai/v1
export MODEL_NAME=llama3.1-8b
export ENV_BASE_URL=http://127.0.0.1:7860

---

Running the Environment

Start FastAPI server

uvicorn server.app:app --host 0.0.0.0 --port 7860

---

Test API

Open:

http://127.0.0.1:7860/docs

---

Running Inference

python3 inference.py

---

Output Format

The system logs execution in structured format:

[START] task=... env=... model=...
[STEP] step=... action=... reward=... done=...
[END] success=... steps=... score=...

---

Reward System

Component	Range	Description
Review Score	0.0–0.4	Bug identification
Compile Score	0.0–0.2	Code compiles successfully
Test Score	0.0–0.4	Test cases passed

Total reward ∈ [0.0, 1.0]

---

Docker Deployment

Build and run:

docker build -t go-code-review .
docker run -p 7860:7860 go-code-review

---

OpenEnv Compliance

This project satisfies:

• /reset, /step, /state endpoints
• Typed observation/action schema
• Multi-step episode handling
• Reward normalization (0–1)
• Docker build compatibility

---

Environment Variables

Variable	Description
API_KEY	LLM API key
API_BASE_URL	LLM endpoint
MODEL_NAME	Model identifier
ENV_BASE_URL	OpenEnv server URL

---

Notes

• Fallback logic ensures stable execution even if LLM output is invalid
• Designed to run within 20 minutes on limited compute (2 vCPU, 8GB RAM)
• Compatible with Hugging Face Spaces evaluation pipeline

---