Keep provider keys off the client
Applications authenticate to Daptin while provider credentials stay in server configuration.
AI model gateway
Keep the client stable while models change.
Point OpenAI-compatible clients at Daptin, configure the models and providers behind it, and keep credentials, access, and usage handling on your server.
- OpenAI-compatible
- Streaming
- Provider routing
One application endpoint
Use familiar SDKs across different model backends.
The latest release exposes model discovery plus compatible chat, completion, and embedding endpoints. Chat supports streaming and tool calls where the selected provider supports them.
Change routing without rewriting the app
Map client-visible model names to configured provider deployments behind the gateway.
Preserve streaming experiences
Stream compatible provider output through to chat and generation interfaces.
Connect usage to the product
Pair model calls with Daptin metering and quotas when AI access belongs to a plan.
Products that benefit
Keep model choice behind an application-owned endpoint.
Chat and assistants
Use compatible messages, streaming output, response formats, and tool calls through familiar SDKs.
Search and classification
Call embeddings or completion-style models without distributing provider keys to every client.
Tool-driven workflows
Let a capable chat model request application-owned functions without giving the provider direct access to those systems.
Metered AI access
Associate token or compute usage with a Daptin plan instead of adding accounting around every model request.
Compatibility without pretending equivalence
One client contract, provider-specific capability.
Provider and credential records map client-visible models to the deployment behind them. Daptin keeps the upstream credential server-side and carries compatible streaming and non-streaming responses back to the caller.
Not every provider supports every chat parameter, tool, response format, embedding model, or streaming behavior. The latest release does not expose Responses or image-generation endpoints. Test the exact route, error shape, access rule, and usage counter your product depends on.
Configure a model provider →