v2.6.41

Speaking bots that run on Meeting BaaS provider keys are now billed for their voice pipeline (speech-to-text, text-to-speech and the language model) at provider cost, as tokens on the bot, after each call. Integrators running speaking bots on our keys will see a new line in the bot's token breakdown and a per-provider usage object; bots on your own provider keys are not affected.

✨New Features

• tokens.voice_pipeline on GET /v2/bots/{uuid} and GET /v2/bots — the tokens debited for a speaking bot's voice pipeline on Meeting BaaS provider keys, converted at your token price and included in tokens.total. 0 for every other bot and for speaking bots on your own keys. See Get bot details.
• provider_usage on GET /v2/bots/{uuid} — what a speaking bot consumed on each provider: stt (seconds of audio), tts (characters spoken) and llm (model, prompt, cached and completion tokens), each with the key it ran on (platform or caller) and its list cost in cents, plus billable_cents, billed_tokens, cents_per_token and the rate_card_version the costs were priced with. null for bots that are not speaking bots.

⚡Improvements

• Billing of the voice pipeline happens whichever way a speaking bot ends: a normal call end, removal through the API, a bot that never got into the meeting, and a bot that failed after reporting its usage. A report that arrives before the bot is finalized is billed as soon as it is, and a billing attempt that fails is retried and swept up automatically, never twice.
• A voice-pipeline debit that takes your balance under your auto-refill threshold triggers the refill and the balance alerts like any other consumption.
• The dashboard's bot details show the Voice Pipeline line in the consumed-tokens breakdown with the provider usage behind it, and the bots list shows a Beta badge and a version filter for bots created with a beta (pre-release) API key.

↩ All releases