- [x] Maintain a single alembic migration file. - [ ] Add label validation in input CSVs for fine-tuning. - [x] For invalid API key errors, return the error in the error column instead of "Failed during background job processing". - [x] Update the endpoint api/v1/fine_tuning/{job_id}/refresh: - [x] Use id instead of job_id, since job_id may be confused with provider_job_id (it should reference the fine_tuning table id). - [x] Do not return "deleted_at" in refresh fine tuning status endpoint. - [x] Generate signed URLs for S3 files in the response body - [ ] Use IDs consistently throughout the platform such converting doc id from uuid to int - [ ] Implement a mechanism to rerun evaluation in memory if something fails. - [x] Merge document upload and fine-tuning initiation into a single endpoint. (will be able to take care of label validation of CSV by this easily) - [x] Combine job refresh and evaluation start into a single endpoint, so evaluation begins automatically once the model status changes to completed.
Maintain a single alembic migration file.
Add label validation in input CSVs for fine-tuning.
For invalid API key errors, return the error in the error column instead of "Failed during background job processing".
Update the endpoint api/v1/fine_tuning/{job_id}/refresh:
Use id instead of job_id, since job_id may be confused with provider_job_id (it should reference the fine_tuning table id).
Do not return "deleted_at" in refresh fine tuning status endpoint.
Generate signed URLs for S3 files in the response body
Use IDs consistently throughout the platform such converting doc id from uuid to int
Implement a mechanism to rerun evaluation in memory if something fails.
Merge document upload and fine-tuning initiation into a single endpoint. (will be able to take care of label validation of CSV by this easily)
Combine job refresh and evaluation start into a single endpoint, so evaluation begins automatically once the model status changes to completed.