Practical guides for
AI observability.
Step-by-step walkthroughs for the workflows teams actually run in Trasys — from first instrumentation to tuning anomaly detection.
Instrumenting your first LLM call
Install the SDK and get token, cost, and latency data flowing from your first agent or chat endpoint in under ten minutes.
Debugging a broken request with a trace ID
A walkthrough of pasting a trace ID from a support ticket into the Debugger and isolating the exact span that failed.
Writing your first TQL query
Syntax basics, common aggregations, and how the same query resolves against both ClickHouse and Postgres.
Setting up alert rules and escalation policies
Define a threshold, attach it to an on-call schedule, and configure how unacknowledged alerts climb the chain.
Publishing a branded public status page
Connect your components to real incidents so your status page updates itself the moment something changes internally.
Tuning Nudges for your traffic patterns
How anomaly scoring works, and how to adjust sensitivity so Nudges surface real drift instead of normal daily variance.