Shipping update - 2026-04-06

Highlights

• AI-assisted log investigations now clean up supported field aliases before sending a request and reject unsupported field references early, which prevents avoidable failures and makes results more predictable.
• Log fetching and log deep links now use the same query-building rules, so the data people retrieve lines up more closely with the view they open next instead of shifting because of mismatched filters or fallback paths.
• Trace investigation guidance is now easier to follow, and environment-based trace filtering uses the correct query shape, which improves the odds of getting the expected results on the first try.

What this enables

• People using AI-assisted troubleshooting can ask more natural log questions without being tripped up by minor field-name differences or malformed references that used to fail later in the workflow.
• Log searches behave more consistently across fetch flows and linked views, which makes handoffs between AI output and interactive exploration less confusing.
• Environment-specific trace checks are more dependable, making it easier to narrow an investigation to the system slice that actually matters.

Technical wins

• Added a sanitization and validation layer ahead of log-query execution so supported aliases are normalized, canonical references stay intact, and invalid dotted fields are stopped before they can trigger downstream errors.
• Refactored log-query generation into a shared path used by both fetch requests and deep links, updated filter semantics to match the intended UI behavior, and removed an automatic index fallback that could produce surprising results.
• Moved trace-tool guidance into dedicated documentation content and corrected the generated environment filter syntax, which required aligning instructions, runtime query construction, and verification coverage so the tool behavior and guidance stayed in sync.