Module lexicon 

Lexicon-based importance scoring — LexiconScorer trait, ConfigLexiconScorer, DefaultEnglishScorer, CompositeLexiconScorer, LexiconAppender, and LexiconProposal. Lexicon-based importance scoring.

Provides LexiconScorer — a trait for domain-specific importance weighting injected at crate::ContextForge construction time via the builder API — and ConfigLexiconScorer, the default TOML-driven implementation.

The scorer runs inside crate::engine::ContextEngine::assemble after BM25 + recency decay, before the token-budget cut. Importance weighting must happen at this point to influence which entries survive the budget cut, not re-rank survivors after.

Two-layer design

Lexicon scoring is opt-in on the builder. DefaultEnglishScorer recognizes plain-English importance signals (“confirmed”, “never mind”, etc.) and is enabled via with_default_english_scorer. A persona scorer (ConfigLexiconScorer loaded from a TOML file) is added via with_persona_scorer. When both are set they stack via CompositeLexiconScorer — additive, with the engine applying the -1.0 floor clamp after fusion.

Structs

CompositeLexiconScorer

Applies multiple LexiconScorers in sequence and sums their boosts.

ConfigLexiconScorer

LexiconScorer backed by a TOML config file.

DefaultEnglishScorer

Always-on baseline scorer for plain-English importance signals.

LexiconAppender

Atomically appends or removes entries from a TOML lexicon file.

LexiconConfig

TOML schema for ConfigLexiconScorer.

LexiconPatterns

A list of case-insensitive substring patterns.

LexiconProposal

A candidate term proposed for addition to the lexicon.

Traits

LexiconScorer

Domain-specific importance scorer, injected at construction time.

Functions

bootstrap_prompt

Generate a calibration prompt for bootstrapping a persona lexicon via an LLM.