File size: 2,383 Bytes
6f12524
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
"""
Data models for the Data Quality Agent Environment.

An ML-focused environment where an AI agent must inspect, diagnose,
and remediate data quality issues in realistic datasets - the kind of
work every ML engineer does before training a model.
"""

from typing import Dict, List, Optional
from openenv.core.env_server.types import Action, Observation
from pydantic import Field


class DataQualityAgentAction(Action):
    """Action for the Data Quality Agent environment - analyst commands."""

    message: str = Field(
        ...,
        description=(
            "Data quality command(s), semicolon-separated. "
            "Examples: DESCRIBE, INSPECT age, CHECK_MISSING, "
            "FIX_MISSING age median, VALIDATE, SUBMIT_REPORT"
        ),
    )


class DataQualityAgentObservation(Observation):
    """Observation from the Data Quality Agent environment."""

    # Main output
    dashboard: str = Field(default="", description="Human-readable data quality report")
    step_number: int = Field(default=0, description="Current step number")
    total_steps: int = Field(default=40, description="Max steps in episode")

    # Dataset overview
    dataset_name: str = Field(default="", description="Name of the dataset being audited")
    num_rows: int = Field(default=0, description="Number of rows in dataset")
    num_columns: int = Field(default=0, description="Number of columns in dataset")

    # Issue tracking
    issues_detected: int = Field(default=0, description="Number of distinct issues detected so far")
    issues_fixed: int = Field(default=0, description="Number of issues remediated so far")
    total_issues: int = Field(default=0, description="Total ground-truth issues (hidden)")

    # Quality metrics
    quality_score: float = Field(default=0.0, description="Current overall data quality 0-1")
    completeness: float = Field(default=0.0, description="Data completeness score 0-1")
    consistency: float = Field(default=0.0, description="Data consistency score 0-1")
    accuracy: float = Field(default=0.0, description="Data accuracy score 0-1")

    # Phase tracking
    phase: str = Field(default="inspection", description="Current phase: inspection / remediation / validation")

    task_name: str = Field(default="", description="Current scenario name")
    task_description: str = Field(default="", description="Scenario description")