Spaces:
Sleeping
Sleeping
File size: 3,162 Bytes
ee117a1 | 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 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 | """
models.py — Type-safe contracts for the Medical Diagnostic Environment.
These Pydantic models define the interface between the LLM agent and the environment:
- DiagnosticAction: What the agent sends (questions, tests, diagnoses)
- PatientObservation: What the agent receives (feedback, test results, progress)
- ClinicalState: Full episode state (for debugging, not sent to agent)
"""
from typing import Optional, List, Dict
from openenv.core.env_server import Action, Observation, State
from pydantic import Field
class DiagnosticAction(Action):
"""
Actions the LLM agent can take during diagnosis.
The agent must choose one action per step:
1. ask_question: Gather patient history
2. order_test: Request diagnostic test results
3. submit_diagnosis: Make final diagnosis (ends episode)
"""
action_type: str # "ask_question", "order_test", "submit_diagnosis"
question: Optional[str] = None # Used when action_type="ask_question"
test_name: Optional[str] = None # Used when action_type="order_test"
diagnosis: Optional[str] = None # Used when action_type="submit_diagnosis"
class PatientObservation(Observation):
"""
What the agent observes after taking an action.
Inherits from Observation:
- done: bool → Is the episode over?
- reward: Optional[float] → Reward signal
Adds medical-specific fields:
- message: Human-readable feedback
- patient_response: Answer to question (if applicable)
- test_result: Test outcome with interpretation
- questions_asked: History of all questions
- tests_completed: History of all completed tests
- patient_data_revealed: What the agent has discovered so far
"""
message: str # Feedback from environment
patient_response: Optional[str] = None # Answer to a question asked
test_result: Optional[Dict] = None # {"test_name": "X", "result": "...", "interpretation": "..."}
questions_asked: List[str] = Field(default_factory=list)
tests_completed: List[str] = Field(default_factory=list)
patient_data_revealed: Dict = Field(default_factory=dict)
steps_taken: int = 0 # How many actions so far
max_steps: int = 15 # Maximum steps allowed
class ClinicalState(State):
"""
Complete internal state snapshot. Contains hidden information (diagnosis, true findings).
Use for debugging only - NEVER send to agent.
Inherits from State:
- episode_id: str → Unique episode identifier
- step_count: int → Current step number
Adds clinical fields:
- true_diagnosis: The correct diagnosis (hidden from agent)
- patient_case: Case identifier
- patient_details: Full patient information (hidden)
- difficulty: ease|medium|hard
"""
true_diagnosis: str = ""
patient_case: str = ""
patient_id: str = ""
patient_details: Dict = Field(default_factory=dict)
difficulty: str = "easy"
questions_asked: List[str] = Field(default_factory=list)
tests_completed: List[str] = Field(default_factory=list)
final_diagnosis_submitted: Optional[str] = None
final_accuracy: float = 0.0
|