Minds

Retrain a mind

Re-run training for a mind. Failed accepted training preserves its verified original sources, settings and worker phase; when that request cannot be verified the mind is rebuilt from its current profile, and its instructions are rewritten from that profile. Ordinary rebuilds regenerate stored knowledge embeddings. The response is the training block — poll GET /v1/minds/{id}/training until readyToChat is true. The call never refuses a retrain: if one is already queued or running, that attempt is returned (200); if a Study run is answering with the mind, the retrain is recorded and starts automatically once the run has finished (202, retrainAfterRuns.studyIds).

post/api/v1/minds/{mindId}/retrain

Request

  • Base URL: https://api.getminds.ai
  • URL: https://api.getminds.ai/api/v1/minds/{mindId}/retrain
  • Auth: HTTP bearer

Path parameters

mindIdstring uuid required

Mind UUID.

Response

Retrain queued, or the retrain already in flight.

status'queued' | 'running' | 'completed' | 'failed' required

Training lifecycle state. A failed attempt is retried automatically until the mind is ready, so it reports queued; failed is reported only for a mind without a system prompt, until it is retrained or its prompt is regenerated.

readyToChatboolean required

True when the mind can answer: training has completed, or it keeps answering with its current version while an update trains (servingPreviousVersion). This is the moment a mind (and an Audience whose minds all report it) is ready: chat and Studies can use it.

beingRepairedboolean

Present and true while the mind cannot answer yet because its last training attempt failed and it is retrained automatically (status is queued). Waiting for readiness may stop waiting for it: a Study runs without it when enough others answer.

servingPreviousVersionboolean

Present and true while an update of a ready mind trains: the mind keeps answering with its current version until the update lands.

learningInBackgroundboolean

True while a mind that is already readyToChat keeps learning from its sources in the background: a wider source collection, then a system prompt rebuilt from everything collected. Its answers can still deepen until it settles; nothing needs to wait for it.

fullyTrainedboolean

True once the mind is readyToChat and its background knowledge step has settled (learningInBackground false). Automatic validation of a new Audience waits for every member to be fully trained.

messagestring nullable

Human-readable progress / status message.

startedAtstring date-time nullable

When training started (null until running).

completedAtstring date-time nullable

When training finished (completed or failed).

Changes

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