TealfabricLibrary

MLflow Tracking Connector Guide

The MLflow Tracking connector lets Tealfabric workflows create experiments and log metrics and parameters against MLflow runs on a tracking server. Use it to persist experiment metadata from training pipelines and feed run signals into downstream governance or deployment workflows.

Document information
FieldValue
Canonical URL/docs/04_connecting-systems/connectors/m/mlflow
Version (published date)2026-05-08
Tagsconnectors, reference, mlflow
Connector IDmlflow-1.0.0

MLflow connector flow showing experiment creation, metric and parameter logging, and connection validation against an MLflow tracking server.

Configuration and connection setup

Configure the connector with your MLflow tracking server base URI. The connector does not send authentication headers; if your deployment requires auth, terminate TLS and auth at a gateway in front of the tracking server.

  • tracking_uri (required): base URI for the MLflow tracking server (for example https://mlflow.example.com).
  • timeout_seconds (optional): request timeout in seconds (default 30).

Create an experiment with create_experiment

Use create_experiment to register a new experiment by name. The connector calls POST /api/2.0/mlflow/experiments/create and returns success with data.experiment_id and data.data (full MLflow response).

const baseUrl = "https://api.example.com/api/v1";
const tenantId = "<TENANT_ID>";
const apiKey = "<API_KEY>";

async function createMlflowExperiment(integrationId: string) {
  const response = await fetch(`${baseUrl}/integrations/${encodeURIComponent(integrationId)}/execute`, {
    method: "POST",
    headers: {
      "X-API-Key": apiKey,
      "X-Tenant-ID": tenantId,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      operation: "create_experiment",
      name: "fraud-model-training",
    }),
  });
  if (!response.ok) throw new Error(`Request failed: ${response.status}`);
  return response.json();
}
curl -X POST "https://api.example.com/api/v1/integrations/<ENTITY_ID>/execute" \
  -H "X-API-Key: <API_KEY>" \
  -H "X-Tenant-ID: <TENANT_ID>" \
  -H "Content-Type: application/json" \
  -d '{
    "operation": "create_experiment",
    "name": "fraud-model-training"
  }'

{
  "success": true,
  "data": {
    "experiment_id": "42",
    "data": {
      "experiment_id": "42"
    }
  }
}

Log run metrics and parameters

Use log_metric and log_param to append tracking metadata to an existing MLflow run. Both operations require run_id, key, and value. The connector sets timestamp automatically for log_metric.

{
  "operation": "log_metric",
  "run_id": "5e8d6f7b4f6c4a0c9f9d2a8f18b0aa11",
  "key": "f1_score",
  "value": 0.94
}

{
  "success": true,
  "data": {
    "data": {}
  }
}

{
  "operation": "log_param",
  "run_id": "5e8d6f7b4f6c4a0c9f9d2a8f18b0aa11",
  "key": "model_version",
  "value": "v3.2.1"
}

Validate connectivity with test

Run test to validate required integration configuration (tracking_uri) and probe GET /api/2.0/mlflow/experiments/search. On configuration failure the connector returns Configuration validation failed. Non-test operations do not perform upfront configuration validation; they attempt the MLflow API call directly.

{
  "success": true,
  "data": {
    "message": "MLflow connection test successful",
    "details": {
      "tracking_uri": "https://mlflow.example.com"
    }
  }
}

Reliability guidance

Most MLflow integration failures are caused by incorrect tracking_uri values, network timeouts on long-running logging bursts, or upstream HTTP errors from the tracking server. Validate connectivity with test first, check every operation response for success, and increase timeout_seconds when logging large batches of metrics.

Additional resources