First public release
Onde Cloud v1.0.0 establishes a stable OpenAI-compatible API for applications that need hosted inference without coupling themselves to one model provider.
One API for inference
The first release exposes POST /v1/chat/completions and GET /v1/models, with streaming responses over Server-Sent Events. Clients can use the familiar OpenAI request shape while Onde Cloud owns routing and backend selection.
Application authentication
Requests authenticate with Onde application credentials. Credentials are validated against the smbCloud control plane and cached briefly, keeping the inference service stateless while preserving a clear trust boundary.
Provider-neutral routing
Public model identifiers stay neutral. Concrete upstream model and provider names remain an implementation detail, allowing the catalogue to evolve without forcing client migrations.
Usage and operations
Completion usage is reported through GresIQ with a local outbox for temporary delivery failures. The service also includes health checks, graceful shutdown, structured logging, and a browser-based chat surface for verifying deployments.