Interface TypedNetworkObserverOptions

Construction options for the observer.

interface TypedNetworkObserverOptions {
    llm: ITypedExtractionLLM;
    maxTokens?: number;
    temperature?: number;
    timeoutMs?: number;
}

Properties

llm maxTokens? temperature? timeoutMs?

Properties

llm

llm: ITypedExtractionLLM

LLM adapter implementing the 6-step extraction call.

Optional maxTokens

maxTokens?: number

Max output tokens. Default 4096 (Hindsight extractions are typically 50-200 facts × ~30 tokens each).

Optional temperature

temperature?: number

Temperature. Default 0 for deterministic extraction.

Optional timeoutMs

timeoutMs?: number

Per-attempt request timeout in milliseconds. When the underlying llm.invoke() does not resolve within this window the attempt is abandoned and the observer falls through to the retry path. Used to prevent stale TCP sockets / hung OpenAI requests from deadlocking long-running ingest pipelines. Default 30 000 ms (30 s).

Settings

Member Visibility

Theme

On This Page

llmmaxTokenstemperaturetimeoutMs