Dashboards

Build interactive dashboards to monitor AI application performance, visualize telemetry data, and explore LLM operations in real time

OpenLIT dashboards let you turn raw telemetry into actionable insights. You write SQL queries against your ClickHouse data, choose a visualization type, and arrange widgets on a canvas. Dashboards support filters, time ranges, and auto-refresh so you can monitor live systems without manual reloads.

How it works

1. Data collection — AI telemetry flows through OpenTelemetry instrumentation into ClickHouse.
2. Query execution — each widget runs an optimized SQL query against your telemetry data.
3. Visualization — results render as charts, tables, statistics, or rich text.
4. Interactive analysis — filters and time-range controls let you drill down and explore without editing queries.

Create a dashboard

1. Open the Dashboards section

   In OpenLIT, click Dashboards in the left navigation. The landing page lists all existing dashboards and folders.

2. Create a new dashboard

   1. Click Create new dashboard in the top-right corner.
   2. In the popup, fill in the dashboard details:
      • Name — a descriptive title (e.g. LLM Observability)
      • Description — the dashboard’s purpose (e.g. Telemetry from LLM apps and AI Agents)
      • Tags — labels for organisation and search
      • Type — select Dashboard
   3. Click Add. You are taken to an empty canvas ready for widgets.
3. Add a widget

   1. Click Create Your First Widget on the empty canvas.
   2. In the popup, choose one of two options:
      • Choose from existing widgets — select a pre-built widget from the library
      • Create new widget — click + Create New Widget to build a custom one
4. Configure the widget

   The widget editor has two main tabs:

   General tab — set the widget title, type, description, value path, prefix, and suffix.

   Query tab — write a SQL query against ClickHouse. For example:

   SELECT count(*) FROM otel_traces
   

   Click Run Query to test it before saving.

5. Save and arrange

   Click Save Changes to add the widget to your dashboard. You can then:

   • Resize widgets by dragging their corners
   • Reorder widgets by dragging them to a new position
   • Add more widgets by clicking Create Widget

Widget types

Each widget type is designed for a different kind of data and use case.

Area chart

Visualize trends over time with smooth, layered charts. Ideal for metrics like token usage, request volume, or cost that change continuously.

Bar chart

Compare categories or segments side-by-side. Useful for showing grouped metrics such as errors by model or requests by provider.

Line chart

Plot time-series data with precision. Best for tracking latency, throughput, or other performance metrics over time.

Pie chart

Show proportions and segment distribution. Good for visualizing things like traffic share by model or error categories.

Stat

Highlight a single key metric such as total requests, average latency, or estimated cost. Displays as a prominent number with optional prefix and suffix.

Table

Display structured query results in a paginated, sortable grid. Well-suited for audit logs, trace lists, and error details.

Markdown

Add notes, context, or section headers using Markdown. Useful for annotating dashboards or linking to related documentation.

Filters and time ranges

Filters make dashboards interactive — you define a filter control at the dashboard level and bind its value into widget queries using mustache-style syntax.

Time range

Select a preset time range at the top of any dashboard:

Range	Description
24H	Last 24 hours
7D	Last 7 days
1M	Last calendar month
3M	Last 3 calendar months
Custom	Any start and end date you choose

Reference the selected range inside a query with:

WHERE event_time BETWEEN {{filter.timeLimit.start}} AND {{filter.timeLimit.end}}

At runtime, OpenLIT replaces {{filter.timeLimit.start}} and {{filter.timeLimit.end}} with the values from the time range control.

Dynamic bindings

Use {{variable_name}} anywhere in a widget’s SQL query to inject a filter value. This keeps your queries flexible and reusable across different environments or segments.

SELECT
  toStartOfHour(event_time) AS time,
  count() AS value
FROM user_events
WHERE event_time BETWEEN {{filter.timeLimit.start}} AND {{filter.timeLimit.end}}
GROUP BY time
ORDER BY time

Use coalesce({{param}}, 'default') in ClickHouse to provide a fallback when a filter has no value selected.

Auto-refresh

You can configure a dashboard to re-fetch data automatically at a fixed interval:

Interval	Typical use case
Never	Static or historical dashboards
30 seconds	Live production monitoring
1 minute	Standard operational dashboards (default)
5 minutes	Low-frequency metric updates
10 minutes	Reduce ClickHouse query load

Avoid high refresh rates on queries that scan large data sets — frequent wide queries can put significant load on your ClickHouse backend.

Organise dashboards

As your dashboard library grows, use folders to keep things structured. You can nest folders inside other folders and drag-and-drop dashboards between them. Pin frequently used dashboards for quick access, or mark one as your Main Dashboard to set it as the default landing page.

---

Prompt Hub

Version, deploy, and retrieve prompts centrally without redeploying your application

Vault

Centrally store LLM API keys and retrieve them at runtime

API reference

Explore the REST API for programmatic access to prompts and secrets

Evaluations

Automatically score LLM outputs for quality, safety, and reliability