akullaganeshA/professional-coding-env

OpenEnv: Professional Coding Task Simulation

OpenEnv is a production-grade playground for evaluating AI software engineering agents on real-world coding tasks. It provides a session-isolated, high-performance environment with integrated graders and structured logging.

🚀 Quick Start

Prerequisites

• —Python 3.11+
• —uv (recommended) or pip

Setup

# Using uv (recommended)
uv sync

# Using pip
pip install -r requirements.txt

Run Server Locally

uv run python3 server/app.py

🛠 API Documentation

The OpenEnv server exposes both REST and WebSocket interfaces for flexible integration.

REST Endpoints

• —GET /health: Health check and system metrics.
• —POST /reset: Initialize a new session. Returns session_id and the initial observation.
• —POST /step: Execute an action in an active session. Requires session_id and action_data.
• —GET /state?session_id=<id>: Retrieve full internal state for auditing/debugging.

WebSocket Interface

Connect to ws://<host>:<port>/ws for persistent, low-latency session management.

• —Protocol: JSON-based message exchange.
• —Message Types: reset, step, state.

🧠 Core Tasks

OpenEnv includes 3 distinct real-world tasks with automated graders:

1. 1.Email Triage (Easy): Filtering urgent tickets from a support stream.
2. 2.Data Pipeline (Medium): Cleaning and imputing customer datasets using pandas.
3. 3.Security Patch (Hard): Remediating SQL injection vulnerabilities via parameterized queries.

📊 Baseline Reproduction

To run the reference baseline and verify structured logging compliance:

export API_BASE_URL="https://api.openai.com/v1"
export MODEL_NAME="gpt-3.5-turbo"
export HF_TOKEN="your-api-key"
uv run python3 inference.py

🔍 Troubleshooting

• —Connection Refused: Ensure the server is running and the port (default 7860) is exposed.
• —ModuleNotFoundError: Ensure you are using uv run or have activated your virtual environment.
• —Score is 0.0: Check the grader logic in tasks.py and ensure the agent response contains the expected patterns.

🛡 Security & Compliance

• —Zero-knowledge Reset: State is wiped using cryptographically secure nonces to prevent leakage between episodes.
• —Thread-safe: Session management utilizes asynchronous locks to ensure integrity under concurrent load.
• —Non-root Docker: Containers run as a limited user to minimize attack surfaces.