[Agentic Chat] Do not respond to NewCheckpoint message

What does this MR do and why?

Related MR: feat: do not expect response for newCheckpoint (gitlab-org/modelops/applied-ml/code-suggestions/ai-assist!2961 - merged)

The response is empty anyway and not sending it saves us time on a round-trip because the server don't have to wait until the next chunk is sent.

This is the time between each chunk in milliseconds in my local GDK:

  • The delay between most of the chunks decreased from 100+ms to <10ms. On production the number will be a bit bigger because we need to take WS message time from the server to client into account
  • There are intermittent 500ms entries due to some blocking reads on Duo Workflow Service, we can address/mitigate them separately once all clients are adapted to use this unidirectional streaming
Before After
Screenshot_2025-07-16_at_13.11.22 Screenshot_2025-07-16_at_13.09.51
Edited by Igor Drozdov

Merge request reports

Loading