effect-uai-streaming-tool-output

Use a streaming pattern with an input schema and run function that emits intermediate events to the UI, then apply a finalize step to reduce events into a single structured final result for the model.

Long-running operations lack intermediate event streaming, preventing real-time progress updates. Implementing a streaming tool pattern surfaces live sub-agent narratives and progress updates to the UI without leaking incomplete final outputs.

A finalize function reduces the stream of intermediate events into a single model-facing structured output. This ensures the AI agent receives a clean final answer derived from the complete stream of progress updates.

Yes, the streaming tool pattern emits real-time intermediate events for user visibility while the finalize step ensures only a single structured final result is delivered as the model-facing output.

Intermediate events are needed for long-running tool operations such as downloads, searches, or data processing where users benefit from live feedback and partial results during execution without exposing incomplete final outputs.

Yes, the streaming pattern supports real-time intermediate events including sub-agent narratives, progress updates, and partial results, all surfaced to the UI during heavy computations before finalizing a structured result.