Response is flushed all at once instead of streaming in Copilot

[Mayank-Draup]

Hi Team,

We are using queue_text_chunk to stream responses. It works as expected in Teams, but in Copilot the entire response is returned at once instead of streaming.

We are using the same codebase for both channels, and streaming behaves correctly in Teams but not in Copilot.

Python Agent SDK version: 0.9.0

Code snippet -

sr = context.streaming_response

sr.queue_informative_update("Gathering relevant data…")
await asyncio.sleep(1.5)
sr.queue_informative_update("Compiling insights…")
await asyncio.sleep(1.5)
sr.queue_informative_update("Preparing your response…")
await asyncio.sleep(1.5)

for chunk in chunks:
    sr.queue_text_chunk(chunk)
    await asyncio.sleep(1.5)

[sayali-MSFT]

Hello @Mayank-Draup,
Streaming works in Teams because the Teams client supports rendering incremental updates sent via the Agents SDK StreamingResponse APIs, but in Copilot (M365 Copilot host) the UI often buffers bot output and shows it only after the final message is ready—so your queue_text_chunk() calls may be delivered but not progressively displayed. The recommended approach is to log/check isStreamingChannel (and channel_id) to confirm whether the Copilot channel advertises streaming support, and ensure you end the stream (endStream/end_stream) because some clients won’t render until a completion boundary is received. If Copilot still doesn’t stream, treat it as a channel limitation and implement a fallback UX (e.g., a couple of informative status updates/typing indicators and then one final message) while keeping chunk streaming enabled for Teams.
Reference Document-
1.https://community.powerplatform.com/blogs/post/?postid=82785a1b-898b-ef11-ac20-6045bda91957
2.https://microsoft.github.io/CopilotStudioSamples/ui/custom-ui/directline-js/