Struct GleaningEntityExtractor 

pub struct GleaningEntityExtractor { /* private fields */ }

Entity extractor with iterative gleaning refinement using TRUE LLM calls

This is the REAL implementation that makes actual LLM API calls for every extraction. It replaces the fake pattern-based extraction with genuine language model inference.

Implementations

impl GleaningEntityExtractor

pub fn new(ollama_client: OllamaClient, config: GleaningConfig) -> Self

Create a new gleaning entity extractor with LLM client

Arguments

• ollama_client - Ollama client for LLM inference (REQUIRED)
• config - Gleaning configuration

pub async fn extract_with_gleaning( &self, chunk: &TextChunk, ) -> Result<(Vec<Entity>, Vec<Relationship>)>

Extract entities with iterative refinement (gleaning) using TRUE LLM calls

This is the REAL implementation that makes actual LLM API calls. Expected time: 15-30 seconds per round = 60-120 seconds total for 4 rounds per chunk.

Performance

• 1 chunk, 4 rounds: ~2 minutes
• 100 chunks, 4 rounds: ~3-4 hours
• Tom Sawyer (1000 pages): ~2-4 hours

pub fn get_statistics(&self) -> GleaningStatistics

Get extraction statistics