ClickHouse's ClickStack observability stack and Hud's runtime code sensor now integrate by sharing trace IDs, pairing service-level observability with function-level runtime forensics. The combined stack lets coding agents assess risky AI-generated code changes before deployment, detect regressions right after release, and get code-level root cause context to fix issues. Hud requires no manual instrumentation and covers the application code layer that OpenTelemetry auto-instrumentation misses. Workflows include pre-deploy risk scoring, release regression monitoring with rollback, and auto-triage/auto-fix. Available via the Hud SDK connected to a ClickStack account.
Questions this post answers
How does Hud's runtime code sensor differ from OpenTelemetry auto-instrumentation?
Hud captures function-level metrics and forensic context for the actual functions a team and its coding agents write, while OpenTelemetry auto-instrumentation only hooks library and framework boundaries like incoming requests, outgoing calls, and database queries. Hud requires no manual instrumentation, runs with low overhead in production, and needs no spans added by hand. Teams weighing observability tooling for AI-generated code can track integrations like this on daily.dev.
How do ClickStack and Hud stay correlated when analyzing a production issue?
They share trace IDs. Forensic data Hud captures is enriched with the same trace IDs ClickStack stores with its traces and logs, so a failing function identified by Hud links directly to the corresponding logs in ClickHouse, and conversely an issue detected in ClickStack lets an agent query Hud for code-level context. Engineers building AI-aware incident response pipelines follow integrations like this on daily.dev.