Skip to main content
POST
/
models
/
{handle}
/
train
cURL
curl -X POST https://u22a8.ai/v1/models/acme.support-tone/train \
  -H "Authorization: Bearer $U22A8_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"effort": "medium"}'
{
  "state": "<string>",
  "run_id": "<string>"
}

Documentation Index

Fetch the complete documentation index at: https://u22a8-police-sweep-2026-06-01.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

Asynchronous: a first train from draft returns 202 with a run id and moves the model through busy to ready. Poll GET …/models/{handle} for the state. From a ready model, pass reoptimize_params: true to force a fresh parameter search; otherwise routine retraining on new samples happens automatically.

Authorizations

Authorization
string
header
required

API key issued in the platform console, sent as Authorization: Bearer <key>.

Path Parameters

handle
string
required

Model handle, e.g. 'u22a8.technical-writing'.

Example:

"u22a8.technical-writing"

Body

application/json

POST /v1/models/{handle}/train body. All fields optional.

From draft, starts a first training pass (the model moves to busy, then ready). From ready with reoptimize_params=True, runs a fresh parameter search. From ready without it, a no-op — routine retraining on new samples runs automatically.

effort
string | null

How much cost and time to spend tuning: low|medium|high|max. Higher levels search more thoroughly. Ignored for automatic retraining.

reoptimize_params
boolean
default:false

When True, force a fresh parameter search from a ready model. When False, a ready model is a no-op (automatic retraining handles incremental updates).

Response

Successful Response

Response from POST /v1/models/{handle}/train.

state
string
required

Current model state after the call.

Examples:

"busy"

"ready"

run_id
string | null

Newly enqueued run id, or the matching existing run id on an idempotent hit. null when state == 'ready' and no re-optimization was requested (no run was queued).