generated: '2026-07-20' method: searched source: openapi/overshoot-openapi.yaml docs: https://docs.overshoot.ai/api-reference/introduction authentication: style: 'Bearer API key (Authorization: Bearer ovs_...)' ref: authentication/overshoot-authentication.yml unauthenticated: - GET /models - GET /billing/pricing openai_compatibility: description: >- /chat/completions is OpenAI-compatible and permissive — unknown request fields are accepted and ignored. Responses follow the OpenAI chat completions shape (id, object, created, model, choices[], usage). Streaming chunks are supported. media_addressing: scheme: ovs:// grammar: 'ovs://streams/{stream_id}?' frame_anchors: - frame_index - timestamp_ms - offset_ms segment_anchors: - start_/end_ prefixed variants of the frame anchors description: >- Reference any moment of a live stream inside a chat completion by URL — a single frame, a window, or several windows. idempotency: request_key_supported: false observed: - >- DELETE /streams/{stream_id} is idempotent on already-deleted streams within the lookup window (repeat deletes succeed). POST endpoints (streams, checkout) do not document an Idempotency-Key header. rate_limiting: ref: rate-limits/overshoot-rate-limits.yml headers: - x-ratelimit-limit - x-ratelimit-remaining - x-ratelimit-reset - retry-after status_code: 429 error_handling: ref: errors/overshoot-problem-types.yml envelopes: - 'stream/billing: { "detail": "" }' - 'chat: { "detail": { "error": { message, type, code } } }' rfc9457: false versioning: ref: lifecycle/overshoot-lifecycle.yml style: URI path prefix (/v1beta) regions: behavior: >- Requests are regionally routed; a 409 RegionError carries expected_region and requested_region so clients can retarget the correct region and skip the redirect/retry step. billing: model: prepaid credits money_unit: integer microcents (1 cent = 1,000,000 microcents) balance_fields: - balance_cents (display) - balance_microcents (exact billing unit) denied: Inference can return 402 when billing denies a request.