Bosca / AI

Models

Bring your model, any provider

A model record names an LLM backend and holds the configuration the runtime needs to call it. Bosca AI runs on OpenAI and Google — provider-agnostic, with no lock-in.

The record

A name and a config bag

A model is a thin record: a key, a name, a type, and a configuration object. The type names the backend, and the configuration holds whatever it takes to reach it — an endpoint, a key reference, default parameters. No provider list is baked in; the record is a label and a bag of settings.

  • A key and a name to reference the model by.
  • A type that names the backend it talks to.
  • A configuration object for endpoint, credentials, and defaults.
model · gpt-5
keygpt-5
nameGPT-5
typeopenai.chat.GPT5
config{ endpoint, apiKeyRef }

Provider-agnostic

OpenAI or Google

The runtime speaks to both out of the box — point a model at the provider you want and the executor handles the rest. Which backends and credentials are reachable is up to how your deployment is configured, not a fixed menu.

  • OpenAI and Google, wired and ready.
  • Your deployment decides which backends and keys are reachable.
  • No provider enum to work around — the type is yours to set.
providers
OpenAI wired
Google wired

Across the platform

The same models, everywhere

These records aren't just for chat. The models you configure power AI features throughout Bosca — the Kit assistant, image and audio generation, content descriptions, and more — all pointed at the providers you choose.

Keep exploring

More on AI