Module llm_policy 

Shared policy parser/compiler for online LLM backends.

Structs

CompiledPolicy

Compiled, position-resolved policy ready for runtime evaluation.

LlmPolicy

Parsed policy specification for online LLM adaptation.

OptimizerHyperParams

Hyper-parameters attached to a training action.

PolicyRule

Interval schedule rule (start..end:action).

PolicyRuntime

Stateful cursor over a CompiledPolicy.

RepeatRule

Repeat schedule rule with cycle period and inner pattern.

RepeatSegment

One segment inside a repeating pattern.

TrainAction

Concrete training directive.

TrainScopeSet

Canonical set of train scopes (all, none, or sorted named scopes).

Enums

OptimizerKind

Optimizer family selected by a train action.

PolicyAction

Action emitted by the policy at each stream position.

PositionExpr

Position expression used in policy schedules.

ScheduleRule

Top-level schedule rule.

Functions

parse_policy_segment

Parse a policy:... string into LlmPolicy.

policy_can_train

Returns true if any schedule branch can emit a training action.

split_method_policy_segments

Split method into base method segment and optional policy:... segment.