LLMs

Studio → Settings → LLMs is where you connect a model provider — the credentials Elizon uses to actually call a model. Every Agent's model, every Knowledge Base's embedding model (see Creating a Knowledge Base), and every Team's lead all resolve to one of these connections — there's no separate, hidden Elizon-managed default.

Supported providers

• Frontier labs — OpenAI, Anthropic, Gemini, xAI, DeepSeek, Meta Llama API, Mistral AI, Jina AI, Kimi (Moonshot AI), Cohere, Voyage AI, Alibaba DashScope, Z.ai, and more — API-key auth.
• Hyperscalers — AWS Bedrock, Google Vertex AI — IAM/service-account auth.
• Aggregators — OpenRouter, CometAPI — API-key auth, many underlying models through one connection.
• Self-hosted / local — Ollama, vLLM, Hugging Face TEI — just a base URL, no API key required.

Connecting one

Pick a provider, supply its credentials (an API key for most; IAM credentials or a service account for the hyperscalers; just a base URL for local providers), then select which of its models this configuration exposes to Agents and Knowledge Bases. Test Connection verifies the credentials before you save — a bad key comes back as a clear "disconnected" status, not a generic error.

Scope

An LLM Configuration is created at the organization level or scoped to one project, the same org-vs-project choice most Settings resources offer.

Rotating a credential

Update just the credential (e.g. after rotating an API key on the provider's side) without recreating the whole configuration — every Agent, Team, or Knowledge Base already pointing at it picks up the new credential immediately, nothing needs re-pointing.