ai-python-streaming-tools

Use async-generator tools when a tool should show progress while it runs.

A streaming tool yields many values, but the agent still needs one final tool result for the next model turn. The return type tells the SDK how to combine those yields.

Text Chunks

Use ai.StreamingTextTool when yielded strings should concatenate into the final tool result.

@ai.tool
async def draft_reply(topic: str) -> ai.StreamingTextTool:
    yield "Checking records for "
    yield topic