Skip to Content

Gateway protocols and measures

Generated from Metering.GATEWAY_API_KINDS. All listed routes use POST. Streaming means chat-style server-sent-event support; speech may return incremental audio separately. A mounted API does not imply every model or provider supports it.

APIPathRequest shapeSSE streamingMeasures
Chat Completions/v1/chat/completionsJSON model, messages; optional streamYesrequests, tokens, costMicros
Completions/v1/completionsJSON model, prompt; optional streamYesrequests, tokens, costMicros
Responses/v1/responsesJSON model, input; optional streamYesrequests, tokens
Messages/v1/messagesAnthropic JSON model, messages, max_tokens; optional streamYesrequests, tokens
Embeddings/v1/embeddingsJSON model, inputNorequests, tokens
Image generation/v1/images/generationsJSON model, prompt and model optionsNorequests, images
Image editing/v1/images/editsProvider-compatible multipart image inputNorequests, images
Speech/v1/audio/speechProvider-specific JSON text, voice and modelNorequests, characters, audioMilliseconds, audioBytes
Transcription/v1/audio/transcriptionsMultipart file, modelNorequests, audioMilliseconds, audioBytes
Translation/v1/audio/translationsMultipart file, modelNorequests, audioMilliseconds, audioBytes
System One/v1/systemoneJSON state, questionsNorequests, tokens

costMicros is an estimated provider cost in millionths of a US dollar, not a subscription credit balance. Audio duration uses milliseconds and audio size uses bytes internally; the Modem UI can present different units.

See request authentication and provider compatibility, limits and estimates, and product configuration.

Workflow guides

Last updated on