pub async fn generate_streaming(
client: &dyn LlmClient,
req: &StructuredRequest,
on_partial: PartialObjectCallback,
) -> Result<StructuredResult>Expand description
Generate a structured JSON object with streaming partial updates.
Calls on_partial with progressively more complete partial objects as tokens
arrive. Returns the final validated object.
A streamed first attempt may be followed by bounded non-streaming repair
calls when max_repair_attempts is non-zero. Repair calls publish only the
final corrected object, avoiding a second misleading partial stream.