Shareable AI Assistant conversations and a clearer Logs and Traces overview

Share an AI Assistant conversation

You can now share an AI Assistant thread as a read-only link scoped to your organization. A teammate who opens it sees the full investigation — the question, the tools the assistant ran, and the answer — without being able to change your thread.

Only the owner of a thread can create a share, and the reader must be signed in to the same organization. This is for handing over an in-flight investigation, or attaching the reasoning to an incident review, without pasting screenshots.

A clearer overview on Logs and Traces

The headline stats on the Logs and Traces overviews now carry a sparkline and a comparison against the previous day. Total logs, logs per minute, error rate, and active services each show their direction over the window, so the number tells you whether it is climbing without opening a chart first.

Quickstart, recent, and saved queries collapse into a single row above the stats. Your own queries and your team’s queries sit behind one dropdown with their counts, so picking up a saved investigation does not push the overview off the screen.

Improvements

• Dashboards: Add to Dashboard now works on chart panels across every Discover sub-page, not only the Service Overview tab. Any chart you are looking at can be promoted to a custom dashboard without leaving the page
• Discover
  • Applications (RUM) sessions can be filtered by any span attribute in the session, including app.screen.name, instead of only attributes on the Session Start and Session End spans
  • Filter value suggestions in Applications (RUM) are scoped to the selected time range, so the picker no longer offers values that cannot match
  • The Applications (RUM) session details panel is reorganized into a body and sidebar, with a clearer session status indicator, a refined timeline, and Attributes, Backend, Queries, and Assets tabs. Every trace link now goes to the same place
  • Host Details shows the mount point for filesystem capacity and usage, with the device name secondary, so /u01 and /var are readable instead of /dev/nvme*
• Control Plane: Physical indexes support contains conditions, so you can index on a substring match such as a service name containing /aws instead of enumerating exact values
• Integrations: The React Native RUM SDK ships an Expo config plugin. Under Continuous Native Generation, the Android Maven repo and the iOS Podfile pod source are applied automatically on expo prebuild instead of being wiped and re-added by hand

Fixes

• AI Assistant: Any chat turn that ran longer than 180 seconds died with an unexpected-error message and discarded the answer, because streaming routes inherited a global per-request deadline
• Dashboards: The doughnut chart did not render
• Discover
  • The Applications (RUM) sessions list ended partway through the selected time range when the underlying query hit its row limit, silently dropping the most recent sessions
  • The Databases RDS Infrastructure tab had never returned data, because it queried the wrong metric family and matched endpoints too loosely
  • The Exceptions detail chart was empty for exceptions with an empty service label, while the occurrence count above it was non-zero
• Logs: A horizontal scrollbar appeared under each log line on Windows and Linux Chrome while scrolling log cells
• Alerting: Alert Monitor and dashboard requests returned a 400 Invalid Time error for users whose browser clock ran a few seconds ahead of the server
• Agents: MCP drop-rule tools called a legacy routing endpoint instead of the current settings endpoints, so creating a drop rule failed