Struct LlmExtractionService 

pub struct LlmExtractionService { /* private fields */ }

LLM-based structured data extraction service

Wraps one or more AIProvider instances and implements ScrapingService. On each execute() call the service:

1. Reads schema and optionally content from ServiceInput.params.
2. Iterates through the provider list until one returns Ok.
3. Returns extracted data in ServiceOutput.data (as JSON string).
4. Metadata includes which provider succeeded and elapsed time.

Provider Fallback

Providers are tried in the order they were added. The first success short-circuits the chain. Errors from skipped providers are logged as warnings, not propagated.

Implementations

impl LlmExtractionService

pub fn new( providers: Vec<Arc<dyn AIProvider>>, config: ExtractionConfig, ) -> Self

Create a new extraction service with an ordered fallback chain

Example

use stygian_graph::application::extraction::{LlmExtractionService, ExtractionConfig};
use stygian_graph::adapters::ai::ollama::OllamaProvider;
use std::sync::Arc;

let providers: Vec<Arc<dyn stygian_graph::ports::AIProvider>> = vec![
    Arc::new(OllamaProvider::new()),
];
let service = LlmExtractionService::new(providers, ExtractionConfig::default());

Trait Implementations

impl ScrapingService for LlmExtractionService

fn execute<'life0, 'async_trait>( &'life0 self, input: ServiceInput, ) -> Pin<Box<dyn Future<Output = Result<ServiceOutput>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

Execute structured extraction via the provider fallback chain

Example

use stygian_graph::application::extraction::{LlmExtractionService, ExtractionConfig};
use stygian_graph::adapters::ai::ollama::OllamaProvider;
use stygian_graph::ports::{ScrapingService, ServiceInput};
use serde_json::json;
use std::sync::Arc;

let providers: Vec<Arc<dyn stygian_graph::ports::AIProvider>> = vec![
    Arc::new(OllamaProvider::new()),
];
let service = LlmExtractionService::new(providers, ExtractionConfig::default());
let input = ServiceInput {
    url: "<h1>Hello World</h1>".to_string(),
    params: json!({
        "schema": {"type": "object", "properties": {"heading": {"type": "string"}}},
    }),
};
// let output = service.execute(input).await.unwrap();

fn name(&self) -> &'static str

Service name for identification in logs and metrics