ops

Models

The model mapping: which artifact answers which endpoint, and its version.

Every other surface answers half of this. /v1/<module>/info says what one endpoint loaded; make model-versions says what is on disk but only from a shell on the box. This is the whole table, from the running service, which is what you want when a caller reports that an endpoint started answering differently and you need to know whether the artifact underneath it moved.

Per bundle: the endpoint it serves, the model values that select it, the version pinned in the owning module's constant.py, every version installed, and which of them is actually serving. status is the summary:

pinned         a pin, and that version is installed. The normal state.
unpinned       nothing pinned, so the newest on disk serves — and the
               next retrain will silently take over. Fine on a laptop.
pin-missing    pinned to a version that is not installed. The endpoint
               is failing, deliberately: see /readyz.
flat           installed with no version directory. What a bucket
               written before labs took over publishing holds; it
               serves, and the next pull replaces it with a version.
not-installed  nothing on disk at all. Run `make models-pull`.

Reports rather than raises, and stays a 200 even when something is wrong — it is the page you read to find out what is wrong. /readyz is the one that fails.

get/models

Request

  • Base URL
    https://api.intura.co (the document declares no server; this is the origin it was published from)
  • Auth
    none declared

Response

Successful Response

unknown required

Changes

No recorded changes to this operation across all 1 revision of this API.