Webhooks & Export

Push classification results in real-time via webhooks, or export full datasets for your data warehouse. Keep downstream systems in sync without polling — every classification triggers action.

Real-Time Event Flow

The pipeline is straightforward: data comes in (from an integration, API push, or file upload), your models classify every record, and webhooks fire with the results. Your downstream systems act on the classification — no polling, no batch delays, no stale data.

Webhook Payload

When a record is classified, Labelf POSTs the result to your endpoint. The payload includes the original text, all classification results with confidence scores, source metadata, and timestamps. Your endpoint should return 200 within 10 seconds.

{
  "event": "classification.completed",
  "timestamp": "2025-11-14T09:22:47Z",
  "dataset_id": "ds_7f3a9b2c",
  "row_id": "row_90312",
  "text": "I want to cancel, I found a better deal with Telia",
  "classifications": {
    "contact_reason": {
      "label": "Cancellation",
      "confidence": 0.94
    },
    "churn_risk": {
      "label": "High",
      "confidence": 0.91
    },
    "competitor_mention": {
      "label": "Telia",
      "confidence": 0.98
    }
  },
  "metadata": {
    "agent_id": "agent_4821",
    "customer_id": "cust_90312",
    "channel": "phone",
    "queue": "retention",
    "product": "fiber_100"
  }
}

Multiple models can classify the same record — each model's result appears as a separate key under classifications. In this example, three models ran: contact reason, churn risk, and competitor detection.

Use Cases

Webhooks turn Labelf from a classification engine into an automation layer. Here is how teams in telecom, banking, and insurance use them.

Data Export

Export classified datasets for ingestion into your data warehouse, BI tools, or custom pipelines. Exports include all metadata columns, classification results, and confidence scores.

Data Lake Integration

For teams that run Snowflake, BigQuery, or Databricks — Labelf supports direct delivery to your data lake. Configure S3 or MinIO bucket notifications for event-driven ingestion, or schedule batch exports that land in your bucket on a cadence you define. Classified data arrives in Parquet or JSON format, ready for your existing ETL pipelines.

# Scheduled export to S3 — runs nightly at 02:00 UTC
{
  "type": "scheduled_export",
  "dataset_id": "ds_7f3a9b2c",
  "destination": {
    "type": "s3",
    "bucket": "your-data-lake",
    "prefix": "labelf/classified/",
    "format": "parquet"
  },
  "schedule": "0 2 * * *",
  "incremental": true
}