Spaces:
Sleeping
Sleeping
Commit ·
512b667
1
Parent(s): 6b248b4
feat: add synchronous HTTP client (client.py)
Browse filesThin httpx-based wrapper around the DataCleaning HTTP API exposing
reset() / step() / state() / health() with typed Pydantic return values.
Supports context manager usage for automatic connection cleanup.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
client.py
ADDED
|
@@ -0,0 +1,114 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Synchronous HTTP client for the Data Cleaning OpenEnv environment.
|
| 3 |
+
|
| 4 |
+
Usage
|
| 5 |
+
-----
|
| 6 |
+
from client import DataCleaningEnvClient, DataCleaningAction
|
| 7 |
+
|
| 8 |
+
client = DataCleaningEnvClient(base_url="http://localhost:8000")
|
| 9 |
+
|
| 10 |
+
# Start a new episode (task_id 1/2/3 or omit for round-robin)
|
| 11 |
+
result = client.reset(task_id=1)
|
| 12 |
+
print(result.observation.task_description)
|
| 13 |
+
|
| 14 |
+
# Take a step
|
| 15 |
+
action = DataCleaningAction(
|
| 16 |
+
operation="fill_missing",
|
| 17 |
+
column="salary",
|
| 18 |
+
params={"strategy": "median"},
|
| 19 |
+
)
|
| 20 |
+
result = client.step(action)
|
| 21 |
+
print(result.observation.current_score, result.reward, result.done)
|
| 22 |
+
|
| 23 |
+
# Inspect state
|
| 24 |
+
state = client.state()
|
| 25 |
+
print(state.episode_id, state.errors_remaining)
|
| 26 |
+
"""
|
| 27 |
+
|
| 28 |
+
from typing import Optional
|
| 29 |
+
import httpx
|
| 30 |
+
from pydantic import BaseModel
|
| 31 |
+
|
| 32 |
+
from models import DataCleaningAction, DataCleaningObservation, DataCleaningState
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
class StepResult(BaseModel):
|
| 36 |
+
"""Returned by reset() and step()."""
|
| 37 |
+
observation: DataCleaningObservation
|
| 38 |
+
reward: float
|
| 39 |
+
done: bool
|
| 40 |
+
info: dict = {}
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
class DataCleaningEnvClient:
|
| 44 |
+
"""
|
| 45 |
+
Thin synchronous wrapper around the DataCleaning HTTP API.
|
| 46 |
+
|
| 47 |
+
All methods raise httpx.HTTPStatusError on non-2xx responses.
|
| 48 |
+
"""
|
| 49 |
+
|
| 50 |
+
def __init__(self, base_url: str = "http://localhost:8000", timeout: float = 30.0):
|
| 51 |
+
self.base_url = base_url.rstrip("/")
|
| 52 |
+
self._client = httpx.Client(base_url=self.base_url, timeout=timeout)
|
| 53 |
+
|
| 54 |
+
# ------------------------------------------------------------------
|
| 55 |
+
# Core API
|
| 56 |
+
# ------------------------------------------------------------------
|
| 57 |
+
|
| 58 |
+
def reset(self, task_id: Optional[int] = None) -> StepResult:
|
| 59 |
+
"""
|
| 60 |
+
Start a new episode.
|
| 61 |
+
|
| 62 |
+
Parameters
|
| 63 |
+
----------
|
| 64 |
+
task_id : int | None
|
| 65 |
+
1 = Easy (fill missing values)
|
| 66 |
+
2 = Medium (fix formats + duplicates)
|
| 67 |
+
3 = Hard (full pipeline)
|
| 68 |
+
None = round-robin (1 → 2 → 3 → 1 …)
|
| 69 |
+
"""
|
| 70 |
+
payload = {"task_id": task_id} if task_id is not None else {}
|
| 71 |
+
resp = self._client.post("/reset", json=payload)
|
| 72 |
+
resp.raise_for_status()
|
| 73 |
+
return StepResult(**resp.json())
|
| 74 |
+
|
| 75 |
+
def step(self, action: DataCleaningAction) -> StepResult:
|
| 76 |
+
"""
|
| 77 |
+
Apply one cleaning operation and return the updated observation.
|
| 78 |
+
|
| 79 |
+
Parameters
|
| 80 |
+
----------
|
| 81 |
+
action : DataCleaningAction
|
| 82 |
+
operation : str – one of fill_missing / drop_duplicates /
|
| 83 |
+
fix_format / replace_value / drop_outliers / fix_dtype
|
| 84 |
+
column : str – target column (optional for drop_duplicates)
|
| 85 |
+
params : dict – operation-specific parameters
|
| 86 |
+
"""
|
| 87 |
+
resp = self._client.post("/step", json=action.model_dump())
|
| 88 |
+
resp.raise_for_status()
|
| 89 |
+
return StepResult(**resp.json())
|
| 90 |
+
|
| 91 |
+
def state(self) -> DataCleaningState:
|
| 92 |
+
"""Return current episode metadata without modifying state."""
|
| 93 |
+
resp = self._client.post("/state")
|
| 94 |
+
resp.raise_for_status()
|
| 95 |
+
return DataCleaningState(**resp.json())
|
| 96 |
+
|
| 97 |
+
def health(self) -> dict:
|
| 98 |
+
"""Ping the server. Returns {"status": "ok"} if healthy."""
|
| 99 |
+
resp = self._client.get("/health")
|
| 100 |
+
resp.raise_for_status()
|
| 101 |
+
return resp.json()
|
| 102 |
+
|
| 103 |
+
# ------------------------------------------------------------------
|
| 104 |
+
# Context manager support
|
| 105 |
+
# ------------------------------------------------------------------
|
| 106 |
+
|
| 107 |
+
def __enter__(self):
|
| 108 |
+
return self
|
| 109 |
+
|
| 110 |
+
def __exit__(self, *_):
|
| 111 |
+
self.close()
|
| 112 |
+
|
| 113 |
+
def close(self):
|
| 114 |
+
self._client.close()
|