Python API¶

The Python API is small and built around evidence, beliefs, registries, policies, and context rendering.

TrajectoryEvidence¶

from bayesian_agent import TrajectoryEvidence

event = TrajectoryEvidence(
    task_id="sop_12",
    skill_id="benchmark/sop_bench",
    context="sop_bench",
    outcome="failure",
    failure_mode="xml_wrapped_answer",
    input_tokens=70123,
    output_tokens=4242,
)

total_tokens is automatically computed when omitted.

BayesianSkillRegistry¶

from bayesian_agent import BayesianSkillRegistry

registry = BayesianSkillRegistry("temp/beliefs.json")
belief = registry.record(event)

print(belief.success_probability)

Use an in-memory registry for tests:

registry = BayesianSkillRegistry.in_memory()

SkillContextBuilder¶

from bayesian_agent import SkillContextBuilder

context = SkillContextBuilder(registry).render(task_context="sop_bench", limit=5)
print(context)

The renderer orders beliefs by context match, posterior success probability, observation count, and token cost.

RewritePolicy¶

from bayesian_agent.core.policy import RewritePolicy

decision = RewritePolicy().decide(belief)
print(decision.action, decision.reason, decision.confidence)

Default actions are:

explore
compress
patch
split
retire

End-to-End Example¶

from bayesian_agent import BayesianSkillRegistry, SkillContextBuilder, TrajectoryEvidence

events = [
    TrajectoryEvidence(
        task_id="task_1",
        skill_id="skill/search_then_verify",
        context="qa",
        outcome="success",
        input_tokens=1200,
        output_tokens=300,
    ),
    TrajectoryEvidence(
        task_id="task_2",
        skill_id="skill/search_then_verify",
        context="qa",
        outcome="failure",
        failure_mode="missing_verification",
        input_tokens=1400,
        output_tokens=350,
    ),
]

registry = BayesianSkillRegistry.in_memory()
registry.record_many(events)

print(SkillContextBuilder(registry).render(task_context="qa"))

Public Import Surface¶

The package root exports:

from bayesian_agent import BayesianSkillRegistry, SkillContextBuilder, TrajectoryEvidence

Lower-level types are available from bayesian_agent.core.