Why we made every model swappable

It would be simpler to ship one model, one recognizer and one voice. It would also make kika.ws far less useful to most of the teams that have tried it.

Every serious deployment we have seen arrives with constraints we cannot guess, an existing contract with a provider, a region requirement, a fine-tuned model that knows their domain, or a policy that a recording must never touch a third party. A bundled stack fails all of those at once.

So the integration surface is deliberately small. You give us an endpoint, a key and a model name. If it speaks the OpenAI chat completions shape, it works, and it can be changed per organization without redeploying anything.