One of the most powerful features of the Laminar cloud platform is realtime traces. You can see spans as they are being executed, without waiting for top level span to complete. It provides immediate visibility into trace data during development and debugging processes. This feature is particularly valuable when working with long-running operations or complex agent interactions that involve multiple LLM requests.

​
Overview

Laminar uses OpenTelemetry for trace collection and processing. The default configuration uses BatchSpanProcessor, which buffers and sends traces in batches to optimize performance.

When using batch processing:

  • Traces are collected and stored temporarily on the client side
  • Data is sent periodically to the Laminar backend
  • There is a small delay between span completion and dashboard visibility
  • The delay depends on batch configuration (flush interval and batch size)

In the default configuration there might be a slight delay between when a span is finished and when it is visible in the Laminar dashboard based on the configuration of the BatchSpanProcessor (usually interval between flushes and the limit of spans in a batch).

If you want to see traces immediately, you can disable batching by setting disableBatch to true in the Laminar.initialize function. However, it’s not recommended for production environments as it may impact the performance of your application.

​
Disable batching (Optional)

You can disable batching by setting disableBatch to true in the Laminar.initialize function.

Laminar.initialize({
    // ... other options
    disableBatch: true,
});

This configuration uses SimpleSpanProcessor, which processes and sends traces immediately.

​
Performance Considerations

  • Development: Realtime processing provides immediate feedback, beneficial for debugging and development workflows
  • Production: Batch processing is recommended for better performance and resource utilization