The base URL for every API request.
https://api.everyais.com/v1
When you enter the base URL in the OpenAI SDK, Cursor, opencode, and similar clients, always include /v1.
The client appends paths such as /chat/completions after it.
Anthropic-compatible clients (Claude Code and the like) are the only exception — the SDK appends /v1/messages itself, so leave /v1 off the base URL and enter https://api.everyais.com.
Timeouts
The limit is determined by the path, not by whether the request is streaming.
| Path | Limit |
|---|---|
/v1/chat/completions · /v1/messages · /v1/responses POST | 5 minutes |
/v1/images/generations · /v1/images/edits · /v1/videos/generations POST | 5 minutes |
Everything else — GET /v1/models · /v1/outputs · job status polling | 29 seconds |
Inference POSTs wait up to 5 minutes even with stream: false. If you set your client timeout to 30 seconds, you will cut off image and long generation requests that are processing normally, so set it generously (the OpenAI SDK default is 10 minutes).
To get the first token of a long response quickly, use stream: true; for work that takes minutes, such as video, use the async job model (submit → polling).