Emit events from code

• Filtering and clustering runs by what happened
• Tracking outcomes and key checkpoints
• Creating high-signal analytics without parsing span inputs/outputs

​

Emitting custom events

import { Laminar, observe } from '@lmnr-ai/lmnr';

await observe({ name: 'extractProduct' }, async () => {
  // ... your logic ...

  Laminar.event({
    name: 'extraction.completed',
    attributes: {
      url: 'https://example.com/product/123',
      fields_extracted: 12,
      duration_ms: 847,
    },
  });
});

from lmnr import Laminar, observe

@observe()
def extract_product():
    # ... your logic ...

    Laminar.event(
        "extraction.completed",
        attributes={
            "url": "https://example.com/product/123",
            "fields_extracted": 12,
            "duration_ms": 847,
        },
    )

​

Best practices

• Keep event name values stable and low-cardinality (don’t include request IDs).
• Put details into attributes (and avoid sensitive data).
• Emit events inside a span context so they attach to the correct trace.

​

Viewing events

• Extract events from traces automatically with Semantic Events
• Group similar events together with Clusters
• Query custom events in the SQL Editor using the events table