Square1 splits its surfaces across several origins. For the API in this
documentation there is exactly one base URL:
All six documented endpoints (/v1/chat/completions, /v1/embeddings,
/v1/models, /v1/models/{id}, /v1/me, /v1/usage) live on that host
and nowhere else.
The full origin map
You’ll encounter other square1.dev hosts as a user of the service. None
of them serve the API in this reference:
Don’t point an OpenAI SDK at square1.dev or dash.square1.dev — the
/v1 chat surface only exists on inference.square1.dev. If you get an
HTML page or a 404 where you expected JSON, check the host first.
Practical notes
- SDK base URL — include the
/v1 prefix:
https://inference.square1.dev/v1.
- Request body size — chat and embeddings accept request bodies up to
20 MB. Other paths on the host have a smaller limit; there’s no reason
to send large bodies to them.
- Browser clients — this API is designed for server-side use. Custom
response headers (the
x-session-* family and friends) are not exposed
to cross-origin browser JavaScript; see
Quota feedback.
