Retrain the AI
Rebuilds the AI assistant's knowledge index. Every page from your latest deployed docs is re-chunked and re-embedded, so answers immediately reflect current content.
Retraining happens automatically on every deploy — you normally never need this endpoint. Use it only to force a rebuild (e.g. after a partial index or unexpected answer quality issues).
Path Parameters
The project ID.
Example
curl -X POST "https://api.syntext.dev/v1/projects/prj_abc123/ai/retrain" \
-H "Authorization: Bearer stx_abc12345_..."
Response
Returns 202 Accepted — retraining runs in the background.
{
"data": {
"status": "running",
"startedAt": "2025-01-15T10:30:00.000Z"
}
}
Errors
| Status | Code | Meaning |
|---|---|---|
409 |
already_running |
A retrain is already in progress for this project. |
400 |
no_deployed_build |
No deployed build with source available — deploy your docs first. |
Poll Retrain Status
curl "https://api.syntext.dev/v1/projects/prj_abc123/ai/retrain" \
-H "Authorization: Bearer stx_abc12345_..."
Response
{
"data": {
"status": "completed",
"startedAt": "2025-01-15T10:30:00.000Z",
"finishedAt": "2025-01-15T10:30:42.000Z",
"pageCount": 38,
"chunkCount": 214
}
}
status is one of idle, running, completed, or failed (failures include an error message).