Fine-Tune API

Create fine-tuned models using training data exported from your Xilos query log. Fine-tuning produces a custom model that can be imported back into Xilos as a routing target, giving you a model specialized for your use case.

Endpoints

Method	Path	Description
`GET`	`/fine-tune/models`	List available base models for fine-tuning
`POST`	`/fine-tune/create`	Create a new fine-tune job
`GET`	`/fine-tune/jobs`	List all fine-tune jobs

List Base Models

GET /api/v1/fine-tune/models

Returns the base models available for fine-tuning.

Response

[
  {
    "id": "gpt-4o",
    "provider": "openai",
    "display_name": "GPT-4o",
    "context_window": 128000,
    "supports_finetune": true
  },
  {
    "id": "gpt-4o-mini",
    "provider": "openai",
    "display_name": "GPT-4o mini",
    "context_window": 128000,
    "supports_finetune": true
  },
  {
    "id": "llama-3-1-8b",
    "provider": "meta",
    "display_name": "Llama 3.1 8B",
    "context_window": 128000,
    "supports_finetune": true
  }
]

cURL

curl https://api.xilos.ai/api/v1/fine-tune/models \
  -H "Authorization: Bearer YOUR_X..._KEY"

Create Fine-Tune Job

POST /api/v1/fine-tune/create

Request Body

Field	Type	Required	Description
`base_model`	string	Yes	Base model ID from the available models list.
`training_data_file`	string	Yes	File ID or URL of the exported training data (from Training Data API).
`name`	string	Yes	Name for the resulting fine-tuned model.
`hyperparameters`	object	No	Optional hyperparameters: `epochs`, `batch_size`, `learning_rate`.

cURL

curl -X POST https://api.xilos.ai/api/v1/fine-tune/create \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_X..._KEY" \
  -d '{
    "base_model": "gpt-4o-mini",
    "training_data_file": "file_01HXYZ",
    "name": "My custom code model",
    "hyperparameters": {
      "epochs": 3,
      "batch_size": 16
    }
  }'

Response

{
  "id": "ftjob_01HXYZ",
  "name": "My custom code model",
  "base_model": "gpt-4o-mini",
  "status": "queued",
  "created_at": "2025-01-15T10:30:00Z"
}

Info: Fine-tune job statuses: queued → running → succeeded or failed. Poll the jobs endpoint to track progress.

List Fine-Tune Jobs

GET /api/v1/fine-tune/jobs

Returns all fine-tune jobs for your organization, including status and resulting model ID.

Response

[
  {
    "id": "ftjob_01HXYZ",
    "name": "My custom code model",
    "base_model": "gpt-4o-mini",
    "status": "succeeded",
    "result_model_id": "ft:gpt-4o-mini:custom-code:abc123",
    "created_at": "2025-01-15T10:30:00Z",
    "finished_at": "2025-01-15T12:45:00Z"
  }
]

cURL

curl https://api.xilos.ai/api/v1/fine-tune/jobs \
  -H "Authorization: Bearer YOUR_X..._KEY"

Fine-Tune Pipeline

Export Training Data

Use the Training Data API to export high-quality query-response pairs from your query log.

Choose a Base Model

List available base models and select one that matches your use case. Smaller models like gpt-4o-mini are faster and cheaper to fine-tune.

Create a Job

Submit a fine-tune job with your training data and base model. Monitor status until the job succeeds.

Import the Model

Once the job succeeds, the fine-tuned model appears in your model list. Add it as a routing target in your Routing Rules.

Evaluate

Use Eval Datasets to compare the fine-tuned model's performance against the base model.

Warning: Fine-tuning requires a sufficient volume of high-quality training data. We recommend at least 500 records with faithfulness and relevance scores of 0.8 or higher. See the Fine-Tuning Guide for detailed guidance.

Training Data Export API Suggestions API