Every org's floor
Every org's floor changing, on one stream, for the operator.
Part of the operator API: every org's floor changing, on one stream, for
whatever serves the platform as a whole. One org's key reads its own floor at GET /v1/events; this is
the same moments, every org at once, on the operator's key alone. There is no ?token=: nothing
that reads this runs in a browser.
SSE, live only, from the moment it opens: no cursor, nothing replayed. The entries are the org
feed's: agent.registered · agent.detached, call.ringing · call.dialing · call.started ·
call.ended, attention.requested · attention.answered, supervisor.took_over ·
supervisor.released. A turn is never on it. Each frame is one entry wrapped with the org whose
log it is and the environment its call runs in (production, sandbox); an agent's own entries serve
both environments, and their env is null. A reader that serves one environment, such as a notifier that
pages people for production's calls, keeps the frames of that environment and drops the rest:
id: 1
event: call.ringing
data: {"org":"org_…","env":"production","entry":{"seq":1,"ts":1727170000.1,"call":"CA_…","agent":"clinica-norte","type":"call.ringing","ephemeral":false,"data":{…}}}id: is the entry's seq in its own log, so ids from two calls interleave and a reconnect opens
from now. retry: 1000 opens the stream and : ping comes every 25 s of quiet. An entry whose log
nobody claimed yet carries no org and is left out.