​

Highlights

• Follow-up question UX and generation quality improvements in Coach.
• Onboarding processing runtime guardrail for safer chat behavior during onboarding jobs.
• Chat agent v2 web-search integration and prompt-behavior tuning.

​

Included updates

• Coach follow-up improvements:
  • maintained dedicated endpoint POST /api/chat/followup_questions
  • tightened follow-up output constraints and display flow
• Onboarding runtime guardrail:
  • added queued/in-progress/failed state disclaimer handling in chat runtime
  • validated across onboarding processing task states
• Web-search integration:
  • integrated factual web retrieval path in chat agents v2
  • strengthened citation handling and non-business benign scope behavior

​

Validation summary (DEV)

• Scenario-planning trigger validation suite passed for trigger/no-trigger boundary cases.
• Follow-up question flow validated for shortened output constraints and UI rendering behavior.
• Onboarding processing guardrail validated for in_progress and failed runtime states.

​

Change classification

• Added:
  • web-search execution path in chat agents v2 flow
  • onboarding-processing runtime guardrail messaging
• Changed:
  • follow-up question generation constraints and UI behavior
  • router and summary prompt behavior for scenario-planning trigger semantics
• Unchanged:
  • no public API contract break
  • no required client migration for existing endpoints

​

Notes

• No public endpoint contract break was introduced in this patch line.
• Repository baseline remains sdk-version.json = 1.5.0; publish this page as final release notes when version is bumped to 1.5.1.