What you get
Three read-only endpoints:
The dashboard’s summary is two to three sentences and a follow-up task is only
created when the AI judges the call needs one. The transcript and
extracted_data are the raw material behind both, so your own rules can be as
detailed as you like.
Poll for new calls
Page through calls received since your last run.from_date and to_date
filter on created_at, the time HealOS received the call record, so keep the
latest created_at you have seen and ask for everything after it.
summary, outcome and extracted_data are filled in by the
post-call analysis, which runs within a minute of the call ending; until then
they are null. Fetch such a call again later, or use the webhook below as the
trigger.
Or react to the post-call webhook
In the dashboard, under AI Receptionist → Settings → Automations, turn on the post-call webhook and give it your URL. Within seconds of each call ending HealOS sends the call’scall_id, outcome, summary and caller number. The
webhook does not carry the transcript: use call_id to fetch it.
speaker is user for the caller and assistant for
the AI receptionist.)
Download the recording
audio/mpeg. Each download is written to your
organization’s recording access log, the same as a download from the dashboard.
From the CLI
Keep in mind
- Transcripts, summaries and recordings are protected health information. Store and transmit them as you would any patient record.
patient_idis set when the receptionist matched the caller to a chart; use it withGET /patients/{patientId}.outcomevalues:booked,rescheduled,cancelled,confirmed,could_not_book,transferred,question_answered,message_taken,hung_up,other. Filter the list withoutcome=message_taken,could_not_bookto find calls that need a human.

