Ollama
Local or private Ollama models for chat, completions and embeddings.
Put self-hosted Ollama models behind DeepintShield so local models get the same controls as cloud providers.
What works
- Chat Completions, Responses API and text completions
- Embeddings and list models
- Streaming and tool calling
Limits to know: No image generation, speech, transcription, files or batch. There is no default base URL, so you must set one. Authentication is optional: none for local access, or a bearer token for protected instances.
How to connect
Add your Ollama base URL (and token, if used) in Model Hub → AI Providers, create a Virtual Key, then call ollama/<model>. With a self-hosted deployment, the whole path can stay on your network.
Read the Ollama provider docs. Requests get the same Virtual Key budgets, guardrails and logs as every other provider on the AI gateway.