Crate fabula_discovery 

Automated sifting pattern discovery.

Provides a pluggable generate-evaluate framework for discovering narratively interesting patterns from simulation trace data.

Architecture

The system works as an iterative loop:

1. A CandidateGenerator proposes candidate patterns from a TraceCorpus
2. PatternEvaluators score each candidate
3. Scored results feed back to the generator for the next round
4. A PatternFilter decides which patterns to keep

The DiscoverySession orchestrates this loop with configurable budgets.

Modules

evaluators

generators

Structs

DiscoverySession

Orchestrates the generate-evaluate loop with configurable budgets.

PairwiseHit

A co-occurrence of two edges sharing a node with a computed Allen relation.

PatternScore

Per-evaluator scores for a candidate pattern.

ScoredPattern

A candidate pattern paired with its evaluation scores.

SessionConfig

Configuration for a discovery session.

SessionHistory

The result of a completed discovery session.

ThresholdFilter

Accepts patterns whose composite score meets or exceeds a threshold.

TraceCorpus

An indexed log of edges for pattern discovery.

TraceEdge

A single edge in a trace corpus.

Enums

SharedNode

How two edges share a node.

Traits

CandidateGenerator

Proposes candidate patterns. Receives scored feedback to guide the next round.

PatternEvaluator

Scores a candidate pattern against a corpus.

PatternFilter

Decides whether a scored pattern is worth keeping.

Functions

pattern_to_dsl

Convert a Pattern<String, String> to fabula DSL text.