API Models and Training

Model endpoints are primarily in Core/api_models.py and Core/api_training.py.

Model Endpoints

• GET /api/models
• GET /api/models/custom
• GET /api/models/builtin
• GET /api/models/:model_id
• DELETE /api/models/:model_id
• GET /api/models/:model_id/download
• GET /api/models/:model_id/source

Model Import Endpoints

• POST /api/models/import/kaggle
• GET /api/models/import/kaggle/info
• POST /api/models/import/kaggle/analyze
• POST /api/models/import/kaggle/stage
• GET /api/models/import/jobs/:job_id
• GET /api/models/import/jobs/:job_id/inspect
• POST /api/models/import/jobs/:job_id/apply-wrapper
• POST /api/models/import/jobs/:job_id/test

Training Endpoints

• POST /api/training/start
• GET /api/training/status/:project_id/:run_id
• POST /api/training/stop
• GET /api/training/:project_id/:task_id/logs
• POST /api/training/cost

Example: Start Training

POST /api/training/start
Content-Type: application/json

{
	"projectId": "prj_1707456789",
	"datasetId": "ds_001",
	"taskType": "object_detection",
	"config": {
		"epochs": 30,
		"batchSize": 8
	}
}

Example: Start Kaggle Import Job

POST /api/models/import/kaggle
Content-Type: application/json

{
	"source": "user/repo",
	"frameworkHint": "pytorch"
}

Common Errors

• 400 invalid training or import config
• 404 model/job not found
• 409 conflicting training job already active
• 500 runtime failure during import, wrapper generation, or test

Recommendations:

• Keep model import and wrapper validation as separate checkpoints.
• Persist import job IDs for async tracking.
• Record exact training config for reproducibility.

Related Pages

• Model Library
• Model Import