Skip to content

Architecture Docs → ABIO execution

Task

Goal specification with scoring criteria.

Overview

Task defines what an agent should accomplish, including the setup, goal, scoring criteria, and constraints on available actions.

Property Type Description
name str Task identifier
task_type TaskType Predict, diagnose, or cure
world World Initial world setup
available_measurements List[Measurement] What agent can observe
available_actions List[Action] What agent can do
Method Returns Description
score(result) float Score the agent's result
is_complete(world) bool Check if goal is achieved

Discussion

Task Types

  • Predict: Forecast future concentrations
  • Diagnose: Identify disease from symptoms
  • Cure: Restore healthy homeostasis

Usage Example

from alienbio import Task, TaskType

task = Task(
    name="diagnose_diabetes",
    task_type=TaskType.DIAGNOSE,
    world=patient_world,
    available_measurements=[glucose_test, insulin_test],
    available_actions=[],  # Diagnosis only
)

result = agent.solve(task)
score = task.score(result)

Protocol

from typing import Protocol, List, Any
from enum import Enum

class TaskType(Enum):
    PREDICT = "predict"
    DIAGNOSE = "diagnose"
    CURE = "cure"

class Task(Protocol):
    """Goal specification for agents."""

    name: str
    task_type: TaskType
    world: World
    available_measurements: List[Measurement]
    available_actions: List[Action]

    def score(self, result: Any) -> float:
        """Score the agent's result."""
        ...

    def is_complete(self, world: World) -> bool:
        """Check if goal is achieved."""
        ...

See Also