Working with results and items

AgentRunResult contains durable run state, messages, steps, usage and cost, warnings, finish reasons, and final output. Use text_stream() for answer text and item_stream() for provider and runtime items. Python item values are dictionary-shaped rather than the TypeScript discriminated AgentItem union.

for item in stream.item_stream():
    item_type = item.get("type")
    render_timeline_item(item_type, item)

Keep provider-specific dictionary access behind one application adapter, then persist the completed result for later reconstruction.

Streaming
Lifecycle hooks

Last modified on July 26, 2026

Streaming Dynamic turns

Start Here

Build

Observe

Reference

Working with results and items

​Related

Related