Datasource Explore
A page for querying connected observability datasources read-only, each in its own native query language.
Features
Supported datasources and query languages
Query a connected datasource directly in the language native to its kind. Every query is read-only.
| Datasource | Query language | Example |
|---|---|---|
| Prometheus | PromQL | rate(node_cpu_seconds_total[5m]) |
| Mimir | PromQL | up |
| Loki | LogQL | {job="varlogs"} |= "error" |
| Tempo | TraceQL | { duration > 500ms } |
| ClickHouse | SQL (read-only) | SELECT count() FROM system.tables |
Picking a datasource
- Choose the datasource to query from the dropdown. Each entry shows its slug and kind.
- A datasource whose schema has been collected is flagged schema cached. This cache helps the Generate with AI feature use real names.
Time range toggle
- For range-capable kinds — Prometheus, Mimir, Loki — a time range (range) checkbox appears.
- Off runs an instant (single-point) query; on runs a time-series (range) query.
Result display
- Time-series (matrix/range) results render as a time-series area chart.
- Instant, log, trace, and SQL results render as a sortable table.
- When a result hits its cap, a result truncated warning is shown; when there is no data, an empty result notice is shown clearly.
Generate a query with AI
- Describe what you want in natural language and press Generate with AI; it turns your request into a query in the right language — using the cached schema — and fills it into the input box.
- The generated query does not auto-run. You review it and press Run yourself to execute.
How to use
- In the sidebar, click Observability > Datasource Explore
- Select the datasource to query from the top dropdown
- (Optional) For a range-capable datasource, turn on time range (range)
- Type a query in the native language, or describe it in natural language and use Generate with AI to fill the query
- Review the query, then press Run (or Enter in the input) to execute
- Read the result as a chart or a table
If no datasource is connected, the dropdown is empty and the page shows "No datasource configured — connect one in Connectors." Connect datasources and register their credentials on the Custom Agents (Connectors) page.
Press Enter in the query input to run immediately. Press Enter in the natural-language box to trigger Generate with AI.
If you see the result truncated (cap reached) warning, narrow the query's range or conditions and run again.
AI analysis tips
- "CPU utilization across all nodes over the last 5 minutes" → generates a PromQL query
- "recent logs containing error" → generates a LogQL query
- "traces that took over 500ms" → generates a TraceQL query
- "row count per table" → generates a read-only SQL query
Diagnose with AI
On the Datasources management tab, each kind's default datasource row shows a Diagnose with AI
link (supported kinds: Prometheus, ClickHouse, Loki, Mimir, Tempo). Clicking it opens the AI assistant
with a section-pinned diagnostic prompt prefilled (Prometheus/ClickHouse → /observability,
Loki/Mimir/Tempo → /monitoring). It is never sent automatically — review and send it yourself,
and it starts a fresh conversation. The agent probes the datasource using that connector's
query/schema tools.
Explore feature summary
- Curated example-query and natural-language prompt chips per kind (all 8 kinds)
- 7d/30d range presets (Prometheus/Mimir up to 30d, Loki 7d — per-kind bounds)
- Result metadata bar (rows/series · round-trip ms · query language · shape)
- Dedicated Loki log viewer (newest-first, label badges) and Tempo/Jaeger duration bars
- An "AI generated" banner on drafted queries; management-tab KPI tiles and refresh
Related pages
- Custom Agents - Connect datasources and register credentials
- AI Assistant - Conversational AI operations helper