Module strategy 

Trimming strategies assign information value to tree nodes.

Each strategy implements domain-specific value scoring. The StrategyResolver maps tool names to strategies via:

1. Exact match in TOML overrides
2. Hardcoded defaults by tool name
3. Strip proxy prefix and retry 1-2
4. Fallback to Default strategy

Structs

CascadingStrategy

Cascading trim — chronological decay for comments. Newest comments are most valuable: p(i) = beta^(n-1-i).

DefaultStrategy

Default strategy — uniform value 1.0 for all nodes.

ElementCountStrategy

Element count trimming — decreasing value by position.

HeadTailStrategy

Head+Tail strategy for logs — 30% head, 70% tail with error boost.

ItemMetadata

Metadata for priority-aware value scoring.

PriorityStrategy

Priority strategy — composite scorer for issues and MRs.

RandomStrategy

Random strategy — assigns deterministic pseudo-random values via LCG.

ReversedStrategy

Reversed strategy — mirror of ElementCount: last item = 1.0, first = 0.3.

SizeProportionalStrategy

Size-proportional trimming for diffs — value weighted by file type.

StrategyResolver

Resolves tool names to trimming strategies.

ThreadLevelStrategy

Thread-level trimming for discussions. resolved=0.3, unresolved=1.0; first+last comment always get value boost.

Enums

TrimStrategyKind

Strategy type identifier.

Traits

TrimStrategy

Trait for assigning information value to tree nodes.

Functions

assign_diff_values

Assign diff values with explicit file paths.

assign_discussion_values

Assign discussion values with resolved status.

assign_priority_values

Assign priority values with explicit per-item metadata.

create_strategy

Create a strategy instance from its kind.