Every signal your AI stack produces, in one platform
TQL

One query language.
Every datastore.

TQL is Trasys's query language for telemetry. Write one query and run it across ClickHouse and Postgres, without learning two SQL dialects or copy-pasting between tools.

~/tql-editor
from traces
where duration > 500ms and time > now() - 24h
group by endpoint
select p95(duration), error_rate(), sum(cost)
order by p95(duration) desc
EndpointP95ErrorsCost
/v1/chat/completions612ms0.4%$182.30
/v1/embeddings84ms0.1%$12.05
/v1/agents/run1.8s2.1%$430.90
Why it matters

Built for people who query telemetry all day

One syntax, two datastores

The same TQL query resolves against ClickHouse or Postgres depending on what it's asking for — you don't pick a dialect.

Schema-aware autocomplete

Field names, tables, and functions autocomplete against your project's real schema, not a generic SQL keyword list.

Save as a dataset

Turn any TQL query into a saved dataset you can reuse on a dashboard, in Deep Search, or export from the API.

Stop guessing.
Start monitoring with Trasys.