framework.md

framework

Base Python interfaces for samplers and environments.

Type Aliases

Alias	Type	Description
Context	dict[str, Any]	Generic payload used to initialize an environment.
StepResult	tuple[Any, float, bool, bool, dict[str, Any]]	Step output (observation, reward, done, truncated, info).

Classes

ContextSampler

Interface for context generation.

Methods

sample(level: int) -> Context

Samples one context payload for a requested difficulty level.

Parameters

Name	Type	Description
level	int	Difficulty or depth requested by the caller.

update(metrics: dict[str, Any]) -> None

Optional feedback hook for adaptive samplers.

Parameters

Name	Type	Description
metrics	dict[str, Any]	Metrics from training or evaluation.

Environment

Interface for interactive environments.

Methods

reset(context: Context) -> Any

Resets environment state from a context payload.

step(action: Any) -> StepResult

Applies one action and returns (obs, reward, done, truncated, info).

Examples

from framework import ContextSampler, Environment

class MySampler(ContextSampler):
    def sample(self, level: int):
        return {"level": level}

class MyEnv(Environment):
    def reset(self, context):
        return context
    def step(self, action):
        return action, 0.0, False, False, {}

sampler = MySampler()
env = MyEnv()
print(sampler.sample(2))
print(env.step({"a": 1}))

Expected output:

{'level': 2}
({'a': 1}, 0.0, False, False, {})