> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hipocap.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Emit events using LLMs

## Semantic Events

Semantic events is a way to track something or extract structured data that can't be explicitly logged but can be inferred from trace data.
You define a prompt that describes what to look for (for example, "Agent completes checkout process"), and our analysis LLM will produce a typed event with the fields you specify using the entire trace data as context.

## When to Use Semantic Events

* **Business outcomes** — You can describe criteria for a successful agent outcome, for example "User responded with 'Thank you'".
* **Locagical errors** — Describe a case when agent fails to complete a task, for example "Agent fails to complete checkout process".
* **Behavioral patterns** — For example, you can categorize user requests to an agent by specifiying "request category" in the structured output schema.

If you're already logging structured events in your application code, you don't need semantic events for those.

## Creating a semantic event (UI)

To define a new semantic event from trace data:

<Frame caption="Creating a new semantic event in the Laminar UI">
  <img src="https://mintcdn.com/hipocap/jGau1UNMEqA4k6Af/images/create-event.png?fit=max&auto=format&n=jGau1UNMEqA4k6Af&q=85&s=f3df60b349e72fd0e743313012ad59a6" alt="Create new semantic event form showing name, prompt, structured output schema, and trigger spans" width="3024" height="1726" data-path="images/create-event.png" />
</Frame>

1. Open **Events** → **Event definitions**.
2. Select the **Semantic** tab and click **+ Event Definition**.
3. In the "Create new semantic event" panel:
   * (Optional) **Start from a template** to prefill the prompt, schema, and trigger spans.
   * **Name**: choose a stable event name (for example, `checkout.completed` or `navigation.failed`).
   * **Prompt**: describe what the event should detect/extract from the trace/span data.
   * **Structured Output**: define a JSON schema for the event payload (keep it small and stable).
   * **Trigger Spans**: click **Add Span** and enter the span name(s) that should trigger this event. When Laminar recieves this span, we will initiate trace analysis and extract the event.
     Pick spans that is finished when the trace has enough context to infer the event, for example a span that is created when agent is done with the task.
4. Click **Create**.

## Viewing Semantic Events

In the Laminar dashboard:

1. Go to **Events** to see the raw event stream
2. Filter by event type, time range, or custom attributes
3. Click any event to see the trace it was extracted from
4. Export events to a dataset for evaluation

Once you have semantic events running, you can [cluster them](/tracing/events/clusters) to discover patterns and group similar events together.