Module extraction 

Pattern Extractor

Extracts reusable patterns from completed episodes:

• Tool sequences that worked well
• Decision points with outcomes
• Error recovery strategies
• Context-based patterns

Example

use do_memory_core::extraction::PatternExtractor;
use do_memory_core::{Episode, TaskContext, TaskType, TaskOutcome, ExecutionStep};

let context = TaskContext::default();
let mut episode = Episode::new("Test task".to_string(), context, TaskType::Testing);

// Add some execution steps
for i in 0..3 {
    let step = ExecutionStep::new(i + 1, format!("tool_{}", i), "Action".to_string());
    episode.add_step(step);
}

episode.complete(TaskOutcome::Success {
    verdict: "Done".to_string(),
    artifacts: vec![],
});

let extractor = PatternExtractor::new();
let patterns = extractor.extract(&episode);

// Patterns may be extracted based on episode content

Structs

PatternExtractor

Pattern extractor

Constants

MAX_SEQUENCE_LENGTH

Maximum sequence length for tool sequence patterns

MIN_PATTERN_SUCCESS_RATE

Minimum success rate to extract a pattern

MIN_SEQUENCE_LENGTH

Minimum sequence length for tool sequence patterns

Functions

deduplicate_patterns

Remove duplicate patterns from a list

rank_patterns

Rank patterns by relevance/quality