Migrating from v2
Nothing changes for apps already using the v2 routes. They take the same requests and return the same response shapes, errors included.
| v2 route | In v3 |
|---|---|
POST /v1/generate-questions (/batch) |
Unchanged. |
POST /v1/detect-language (/batch) |
Same request and response. It is now served by lid-608, the more accurate model, but still answers only with one of the 25 v2 languages. model (lid-neural-25 / lid-lite-25) picks lid-neural-608 or lid-lite-608, and variant picks the mode (questions → traffic, passages → coverage). |
GET /v1/languages, GET /v1/lid/models |
Unchanged. |
POST /v1/warmup with no body |
Unchanged: warms question generation and language ID only, never the other GPUs. |
GET /health |
Unchanged, plus a version field. |
Moving to the new endpoints
Section titled “Moving to the new endpoints”To use the new features, move to:
/v1/questions/generate, which adds theusageblock./v1/language-id, which covers all 608 languages.