Oumi API

Oumi API

API endpoints for Oumi fine-tuning integration.

Base URL

http://localhost:3001/api/oumi

Export Dataset

Export training dataset in Oumi-compatible format.

Request

POST /api/oumi/export-dataset
Content-Type: application/json

{
  "matchIds": ["AR-2024-0142", "AR-2024-0143"],
  "format": "sft"
}

Response

{
  "datasetId": "dataset-123",
  "format": "sft",
  "records": 100,
  "downloadUrl": "https://..."
}

Trigger Fine-tuning

Submit a fine-tuning job.

Request

POST /api/oumi/fine-tune
Content-Type: application/json

{
  "datasetId": "dataset-123",
  "model": "llama-3.3-70b-versatile",
  "method": "lora",
  "config": {
    "epochs": 3,
    "learningRate": 0.0001
  }
}

Response

{
  "jobId": "job-123",
  "status": "submitted"
}

Get Fine-tuning Job Status

Get status of a fine-tuning job.

Request

GET /api/oumi/fine-tune/:jobId

Response

{
  "jobId": "job-123",
  "status": "training",
  "progress": 65,
  "startedAt": "2024-12-14T10:00:00Z",
  "estimatedCompletion": "2024-12-14T12:00:00Z"
}

Next Steps

• Oumi Integration - Integration guide
• Fine-tuning Guide - Fine-tuning workflow