OpenEnv-compatible multi-turn data analytics environment for RL agent training.
import requests
BASE = "https://albert-einstein-09-codedark.hf.space"
# Reset for new task
obs = requests.post(f"{BASE}/reset").json()
print(f"Task: {obs['question']}")
# Run Python code
obs = requests.post(f"{BASE}/step", json={
"tool": "run_python",
"args": "<code>result = df.shape</code>"
}).json()
print(f"Result: {obs['stdout']}")
# Submit answer
obs = requests.post(f"{BASE}/step", json={
"tool": "submit_answer",
"args": "<answer>42.5</answer>"
}).json()
print(f"Reward: {obs['reward']}")
run_python - Execute Python/pandas coderead_notes - Read saved notessave_note - Save note for laterclarify - Ask clarifying questionsubmit_answer - Submit final answer