Configuring AI models

Compose AI supports model configuration by modality: text, vision, speech-to-text, and text-to-speech. Open Settings > Models to review available providers and current configuration.

System models#

System models use the model credentials managed by Compose AI. Choose a system model for a modality when you want the simplest setup and do not need to provide a provider key.

Bring your own key#

Enable BYOK to configure your own provider credentials. Depending on the provider, the form can include:

• Provider and modality.
• API key.
• Model name override.
• Endpoint URL, deployment name, or region.
• Additional provider-specific configuration.

Validate a key before saving it. Credentials are stored for the workspace and can be updated or deleted from the Models page. Avoid sharing API keys in agent prompts or task attachments.