AI providers & model routing
Connect OpenAI, OpenRouter, DeepSeek or a custom OpenAI-compatible provider and choose providers independently per website.
Workspace credentials
Open Admin → AI providers to store provider credentials encrypted with your application key. A website can then reference one credential for chat and another for embeddings.
Chat vs embeddings
Use the provider that best fits conversation quality and cost for chat. Embeddings can use a different provider because not every chat provider exposes a compatible embeddings endpoint.
Custom providers
Custom endpoints must use HTTPS and should implement an OpenAI-compatible /chat/completions API. Test credentials in staging before sending production traffic.