Stream AI chat response (SSE)
Send a message and receive the AI response as a Server-Sent Events stream. Optionally bind to an incident or alert for context. For eligible unbound home chat, checkpoint_id and decision confirm or cancel approval-gated writes. Events include confirmation_required for pending writes and checkpoint_resolved before replay. If an action-item write needs required fields the model did not supply, no checkpoint is created and the turn completes with a plain reply linking to that incident’s action-item form in the Rootly web app. The stream emits done with status: complete and no confirmation_required event; only confirmation_required events are resumable. status events carry a short progress line ({status}) such as the running tool’s title or its progress message; they never include tool arguments or results. skill_loaded events carry {skill: {id, name, slug, version}} when an AI Skill is loaded. Limited to 60 requests per minute per credential. Rate-limit responses include Retry-After and X-RateLimit-* headers. Requires ai.chat:write OAuth scope or an API key.
Authorizations
Query Parameters
Response
SSE event stream