Skip to content
Documentation

Dashboard - How to Navigate in Langfuse UI

Let’s explore Langfuse UI / Dashboard, and understand how to find our way around it.

Home - high level overview

This is your starting point. The Home dashboard gives you a bird's-eye view of your AI application's health.

What to look for:

  • You will see your total traces tracked (in a given time-frame)
  • A graph of traces over time - spotting potential hot-spots and knowing when is the AI most used.
  • Your overarching model costs (e.g., how many Euros/cents you are spending on Gemini vs. Mistral).

image-20260312-233601.png

And a bit below, you get all the latency stats:

  • Trace latency Percentiles
  • Generation latency Percentiles
  • Span latency Percentiles

Do note that these include the generation time for the answers, so they depend on the answer sizes and the models that were involved in generating it.

image-20260313-162043.png

Usually the Home Dashboard is pretty good at understanding what’s the general usage of your AI app, how many requests you get, etc.

Dashboards

Dashboards section comes with 3 predefined default dashboards, but you can add your own:

image-20260316-230954.png

Some of these dashboards repeat some of the information from the Home Dashboard, but includes extra depth, i.e. you can get total count of observations made, cost per model breakdown and more.

image-20260312-233537.png

Tracing

This is the most important part in the Observability - Tracing. This is the place where you can browse all the observations made:

image-20260316-231240.png

There are different type of observations: spans, generations, events, etc. Our PHP SDK, so far, makes use of some of these, but the Langfuse has a wide support for more detailed types.

All the observations are bundled into Traces. The UI allows you to filter the Traces, find the ones you’re interested in, and then open those up and see all the chains of observations that took places as a part of that trace.

Clicking on each observation, will reveal more details: usually an Input / Output + a lot of metadata associated with that observation.

Sessions

Each user might have multiple AI interactions. Those traces are bundled into Sessions. Note that - for anonymous users the User ID will be 0.

Users

Lists all users and the associated Traces, Sessions and Scores. This helps narrow down and identify what Traces in what Sessions lead to what Scores.

image-20260316-233234.png

Other Features & The Roadmap

There are more sections and features in Langfuse - Prompt Management, Evaluations, LLM-as-a-Judge, Playground and Scores - but not all of these are implemented in the PHP SDK yet (and not bridged into Drupal either).

We’re planning to do some of these integrations in upcoming months, so stay tuned for more User Guides on how to use these.