Start a voice or text session with the expert in your app
Returns a LiveKit URL and token; the expert joins the room. Voice is billed as a web voice call, text like chat (per token); both use credits and the concurrent-call cap. A text session’s token cannot publish audio. The page’s client tools and the session’s guide come back too.
A publishable key (Authorization: Bearer rydhm_pk_…), from one of the key’s allowed websites: the Origin header must be on its list. Never a secret API key (401 secret_key_not_allowed). Rate limited per visitor and per key.
Authentication
Bearer authentication of the form Bearer <token>, where token is your auth token.
Request
voice: talk (the microphone and the expert’s voice). text: type, with no audio either way, billed like chat. One of the bubble’s modes (400 mode_not_available otherwise).
Who the page’s user is. id with hash (hex HMAC-SHA256 of id under your workspace’s user-signing secret, computed on your server) is a signed user; id alone, or anonymousId, is stored as anon_ and the id.
Your app’s JSON context (at most 4 KB).
Start (or resume) this guide.
One of the expert’s languages (400 otherwise).
Custom tools the page registered: { name (2–41 lowercase letters, digits, _), description, parameters (a JSON Schema object, at most 4 KB) }.
A preview or teach token (rydhm_pv_…): needed while the coach is in testing.