Models & API keys

Configure compatible model endpoints through shared connection settings.

Create a model connection

Open the model settings in your deployment. Add a supported connection with its endpoint, credentials, and available models. Assign the appropriate text, voice, or other supported model capability through the application.

Configure compatibility through the generic connection settings rather than introducing a new environment variable for every provider or model.

Keep credentials on the server

Provider credentials belong in the encrypted secret store. Do not put API keys in a bot’s instructions, exported fixtures, client-side code, or connector capability configuration.

BYOK means you supply the key and your provider bills its usage. One credits and provider charges are distinct; the proposed commercial plans do not define a published cash value per credit.

Self-managed model endpoints

Compatible self-managed endpoints can be used where the deployment supports their protocol and capability. Check the full request path, including speech and external tools, before describing a setup as contained within a network.

Model availability and behavior vary by configured endpoint. Continued Company access through eligible Flash models is a proposed plan benefit; the exact eligible models and access terms are agreed during setup.

Next guideVoice & phone capabilities ↗
Adapted for One from the project’s current implementation and self-hosting documentation. One builds on Rakazo; the license and attribution notices are on the Licenses page. Upstream reference ↗