One endpoint,
every model.
Tare is an OpenAI-compatible gateway in front of every model you use — ours and the ones you bring yourself. One base URL, one key, and every token accounted for.
Changelog
What this solves: What changed in Tare's caller-facing interface documentation, and when — so you can compare against the version you've already integrated.
The following records important changes to Tare's interface documentation, in reverse chronological order.
2026-09-28.1
Added /v1/images and /v1/images/generations; added the models and provider fields to the /v1/embeddings request body. No interface behavior changed; this only documents the existing endpoints and fields.
Related: Endpoint list.
2026-09-27.2
Added /v1/videos/generations and /v1/mocks/*, and corrected the documentation's statement that /v1/mocks/* uses the ApiResponse envelope.
Related: Endpoint list, Drills (Mock).
2026-09-27.1
/v1/chat/completions gained models and provider; GET /v1/generation gained four duration fields.
Related: Routing, Endpoint list.
2026-09-01.1
Added the source IP allowlist. Calls from outside the list return 403, with metadata.refusal_reason = source_ip_not_allowed.
Related: API Key.
2026-08-22.4
Added documentation for the chat endpoint and Generation.upstream_inference_cost.
Related: Endpoint list, Billing rules.