Telemetry destinations
Send the numbers Searchability already records about your search into your own monitoring tools — Datadog, Grafana, Honeycomb, New Relic or anything that accepts OpenTelemetry (OTLP).
- A continuous stream of metrics, traces and logs; a webhook is one event at a time.
- One auth header is the only secret. Encrypted when stored, never shown again.
- Found under Insights → Telemetry.
Stream telemetry to your stack
A telemetry destination streams your search telemetry into your own monitoring stack. Search Stack already records how searches perform — latency, result counts, zero-result rates, embedding and rerank timings, judge verdicts. A destination sends that stream, as standard OpenTelemetry (OTLP), straight to Datadog, Grafana, Honeycomb, New Relic or any OTLP backend. They live under Insights → Telemetry.
Endpoint, auth header, signals
Click Add Destination, pick your provider (or Custom), and fill in the endpoint and an auth header for your key — kept encrypted, never shown again. Choose the OTLP transport (HTTP or gRPC) and tick which signals to send: traces, metrics, logs.
Search-quality metrics, tagged
With the metrics signal on, these instruments flow to your backend. Every metric carries an account tag for per-tenant dashboards, and the search-quality metrics — eval pass rate, zero-result rate, judge verdicts — carry pin and verdict tags no generic search backend can emit. Exported metrics also hands you a ready-made starter dashboard.
Add a destination
- Insights → Telemetry → Add Destination. Pick a provider (Grafana Cloud, Honeycomb, New Relic), which fills in the endpoint and header name, or Custom for any other OTLP endpoint.
- Set the endpoint of your monitoring tool — the
/v1/<signal>path is added for you — and the OTLP transport: HTTP/protobuf (port 4318) or gRPC (port 4317). - Paste your key into the auth header.
- Tick the signals (traces, metrics, logs) and click Create. A health dot on the row shows whether data is getting through.
The three signals
| Signal | What you get |
|---|---|
| Traces | One trace per search request: a timeline of each step (embedding, retrieval, rerank, judging). |
| Metrics | The numbers below. What you chart and alert on, and the basis of the search-quality story. |
| Logs | Log lines for notable events, linked to the traces above. |
What the metrics tell you
A fixed set, listed under Exported metrics. Names all start searchstack.; each carries an account tag, so several accounts can share one destination.
| Metric | What it is |
|---|---|
searchstack.search.duration | How long search requests take (a histogram — read the median or the slowest 5%). |
searchstack.search.requests | How many searches are handled. |
searchstack.search.zero_results | Searches that returned nothing. |
searchstack.search.duplicate_hits.dropped | Results removed because another result in the same response already had that id. Should stay at zero. |
searchstack.embedding.duration / .failures | How your embedding model performs, and when a call fails. |
searchstack.rerank.duration / .failures | The same for your reranker. |
searchstack.eval.run.pass_rate | Pass rate of a completed eval run. Tags: eval_set, model, reranker, judge. |
searchstack.judge.verdicts | Judge pass/fail verdicts, tagged by provider and model. |
A pass rate is recorded once per run, so it is a histogram — read it with histogram_quantile. A zero-result rate is zero_results / requests. Series names follow the standard OTLP→Prometheus naming rules.
The search-quality story
Only the eval and judge metrics carry which eval set (and its pinned version) and which verdict each number came from.
- A dashboard can show a pass-rate line per eval set, a live zero-result trend and the judge-verdict split — an SLO for result quality, not just uptime.
- Pass rate fell — alert when the median of
searchstack.eval.run.pass_ratedrops a few points below its recent normal. The same drop theeval.run-regressedwebhook fires on. - Zero-result rate spiked — alert when
zero_results / requestsrises above your limit (say 2%).
Start from the built-in dashboard
Exported metrics → Download starter dashboard builds one from these metrics.
- Grafana, New Relic, Datadog — import the JSON directly in their UIs.
- Honeycomb — the file is a Boards API request body you POST with your own key.
- Its Search quality row gives the pass-rate line, zero-result trend, judge split and a “has it got worse” number.