Spaces:
Sleeping
Sleeping
Delete models.py
Browse files
models.py
DELETED
|
@@ -1,89 +0,0 @@
|
|
| 1 |
-
"""Data models for the OpenEnv email triage environment."""
|
| 2 |
-
|
| 3 |
-
from typing import Literal
|
| 4 |
-
|
| 5 |
-
from pydantic import BaseModel, field_validator
|
| 6 |
-
|
| 7 |
-
OPEN_INTERVAL_EPSILON = 1e-2
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
def _strict_open_unit_interval(raw_value: float) -> float:
|
| 11 |
-
"""Clamp numeric values to the strict open interval (0, 1)."""
|
| 12 |
-
numeric_value = float(raw_value)
|
| 13 |
-
if numeric_value <= 0.0:
|
| 14 |
-
return OPEN_INTERVAL_EPSILON
|
| 15 |
-
if numeric_value >= 1.0:
|
| 16 |
-
return 1.0 - OPEN_INTERVAL_EPSILON
|
| 17 |
-
return numeric_value
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
class EmailObservation(BaseModel):
|
| 21 |
-
"""Represents the email context visible to the agent at each step."""
|
| 22 |
-
|
| 23 |
-
email_id: str
|
| 24 |
-
subject: str
|
| 25 |
-
body: str
|
| 26 |
-
sender: str
|
| 27 |
-
timestamp: str
|
| 28 |
-
thread_history: list[str]
|
| 29 |
-
task_id: str
|
| 30 |
-
step_number: int
|
| 31 |
-
total_emails: int
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
class TriageAction(BaseModel):
|
| 35 |
-
"""Represents the action chosen by the agent for an email."""
|
| 36 |
-
|
| 37 |
-
label: Literal["urgent", "normal", "spam", "archive"]
|
| 38 |
-
summary: str
|
| 39 |
-
route_to: str
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
class RewardResult(BaseModel):
|
| 43 |
-
"""Represents deterministic grading output before reward shaping."""
|
| 44 |
-
|
| 45 |
-
score: float
|
| 46 |
-
breakdown: dict[str, float]
|
| 47 |
-
feedback: str
|
| 48 |
-
|
| 49 |
-
@field_validator("score")
|
| 50 |
-
@classmethod
|
| 51 |
-
def _validate_score(cls, value: float) -> float:
|
| 52 |
-
return _strict_open_unit_interval(value)
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
class EnvironmentState(BaseModel):
|
| 56 |
-
"""Represents full internal environment state for debugging and evaluation."""
|
| 57 |
-
|
| 58 |
-
task_id: str
|
| 59 |
-
current_step: int
|
| 60 |
-
total_steps: int
|
| 61 |
-
done: bool
|
| 62 |
-
action_history: list[TriageAction]
|
| 63 |
-
reward_history: list[float]
|
| 64 |
-
|
| 65 |
-
@field_validator("reward_history")
|
| 66 |
-
@classmethod
|
| 67 |
-
def _validate_reward_history(cls, values: list[float]) -> list[float]:
|
| 68 |
-
return [_strict_open_unit_interval(value) for value in values]
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
class StepResult(BaseModel):
|
| 72 |
-
"""Represents the standardized output of environment step calls."""
|
| 73 |
-
|
| 74 |
-
observation: EmailObservation
|
| 75 |
-
reward: float
|
| 76 |
-
done: bool
|
| 77 |
-
info: dict[str, str | int | float | bool]
|
| 78 |
-
|
| 79 |
-
@field_validator("reward")
|
| 80 |
-
@classmethod
|
| 81 |
-
def _validate_reward(cls, value: float) -> float:
|
| 82 |
-
return _strict_open_unit_interval(value)
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
class ResetResult(BaseModel):
|
| 86 |
-
"""Represents the standardized output of environment reset calls."""
|
| 87 |
-
|
| 88 |
-
observation: EmailObservation
|
| 89 |
-
info: dict[str, str | int | float | bool]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|