title	OpenEnv Data Cleaning
emoji	🧹
colorFrom	blue
colorTo	indigo
sdk	docker
tags	openenv reinforcement-learning data-engineering

OpenEnv Data Cleaning Environment

An OpenEnv environment for training RL agents on SQL data cleaning tasks. The agent works with a SQLite database and issues queries to fix data quality issues like duplicates, inconsistent formatting, and missing values.

Tasks

Level	Task	What the agent needs to do
Easy	Deduplication	Remove duplicate rows from users table, keep lowest id
Medium	Standardization	Normalize emails in contacts — trim whitespace, lowercase
Hard	Imputation	Fill NULL salaries in employees with department averages

Setup

pip install -e .

# required
export HF_TOKEN="your-hf-token"

# optional (have defaults)
export MODEL_NAME="Qwen/Qwen2.5-72B-Instruct"
export API_BASE_URL="https://router.huggingface.co/v1"

python inference.py

Docker

docker build -t data-cleaning-env .
docker run -p 7860:7860 data-cleaning-env

Structure

├── data_cleaning_env.py    # environment (reset/step/state)
├── inference.py            # agent loop
├── server/app.py           # FastAPI server
├── openenv.yaml            # OpenEnv config
├── pyproject.toml
├── Dockerfile
└── requirements.txt