Skip to main content

Module classify

Module classify 

Source
Expand description

Stage 2 of the pipeline: classify each collected commit using a four-tier cascade.

§Tiers

  1. Exact — Aho-Corasick multi-keyword match (case-insensitive).
  2. Regex — pre-compiled regex patterns.
  3. Fuzzy — structural heuristics (merge/revert/ticket-prefix).
  4. LLM — optional async fallback via an OpenAI-compatible API.

Tiers 1–3 are synchronous and run in parallel across commits via Rayon. Tier 4 is async and serialized.

Re-exports§

pub use classifier::ClassificationEngine;
pub use classifier::ClassificationEngineConfig;
pub use errors::ClassifyError;
pub use errors::Result;
pub use pipeline::ClassificationPipeline;
pub use pipeline::ClassificationStats;
pub use rules::Rule;
pub use rules::RuleSet;
pub use taxonomy::SubcategoryDef;
pub use taxonomy::TaxonomyRegistry;
pub use taxonomy::TopLevelCategory;
pub use tiers::ClassificationResult;

Modules§

classifier
Cascade orchestrator combining the classification tiers.
errors
Error types for the classify module.
pipeline
End-to-end classification pipeline: read DB → classify → write back.
rules
Rule loading and types for the classification cascade.
sources
Multi-source classification: external ticket systems as high-confidence classification signals.
taxonomy
Two-level classification taxonomy.
tiers
Implementations of the four classification tiers.