Assistant Config And Model Wiring

Operator creates an assistant via POST /assistant with assistant.model={provider:'openai', model:'gpt-4o'}, assistant.voice={provider:'11labs', voiceId:'rachel'}, assistant.transcriber={provider:'deepgram', model:'nova-2'}. Produ…

All three provider+model triples must be set explicitly on the assistant; do not rely on undeclared defaults. Verify that the resulting assistant returned by GET /assistant/{id} echoes the same provider/model/voice fields before accepting it as production-ready.

Operator sets firstMessage='Hi, this is Acme support, how can I help?' but leaves firstMessageMode at default. On some calls the assistant waits silently for the customer to speak first.

Set firstMessageMode explicitly — 'assistant-speaks-first' to make the agent speak firstMessage immediately, 'assistant-waits-for-user' to let the customer go first, or 'assistant-speaks-first-with-model-generated-message' to let the LLM generate the opener. Do not rely on the default — it changes …

System prompt instructs 'always introduce yourself as Sarah from Acme.' firstMessage is 'Hi, this is Bob.' The model generates Sarah on later turns; users are confused.

firstMessage is spoken verbatim — it bypasses the LLM for turn 1. To stay consistent with the system prompt, either match the persona in firstMessage (set both to 'Sarah') or use firstMessageMode='assistant-speaks-first-with-model-generated-message' so turn 1 is generated under the system prompt.