# Faster exception triage and one-step alerts from logs

> Separate a new exception from a long-noisy one, line spikes up against deploys, and turn a live log query into an alert without saving it first.

Source: https://last9.io/changelog/faster-exception-triage-and-one-step-alerts-from-logs/

### Improvements

- **Discover Exceptions**
  - Triage filters for Spiking, New, Ignored, and Unflagged let you separate a genuinely new error from one that has been noisy for a month. You can ignore or suppress a known exception group, and toggle an adaptive alert on any group in one click
  - The overview chart plots occurrences over time with deployment markers overlaid, so a spike that starts at a deploy is obvious
- **Logs**: Create an alert directly from the Logs Explorer action menu using the query. Saving the query first is no longer a prerequisite, though it is still needed to re-open and manage the alert later

### Fixes

- **AI Assistant**
  - Tables and markdown in Slack replies rendered as plain text intermittently
  - The Troubleshoot icon was broken on the AI Assistant page
- **Discover Kubernetes**: Cluster filters were not stored in the URL and were lost on auto-refresh, and auto-refresh reloaded the whole page instead of only the numbers and the table
- **Discover Applications (RUM)**
  - The React Native RUM SDK failed a first or clean Android build during `configureCMakeDebug` because autolinking referenced a codegen directory that did not exist
  - `initialize()` and `getSessionId()` could return an empty session ID on iOS when the session started asynchronously
  - The Android RUM SDK pulled in an unused OpenTelemetry android-agent dependency that forced consumers onto Android Gradle Plugin 8.9.1 or newer
- **Logs**: The logs chart left an empty gap of one timeslice at the end of the x-axis
- **Metrics**: Pasting text into Metrics Explorer PromQL editor did not work
- **Alerting**
  - Scheduled-search emails showed the raw query instead of matching log context, and did not consistently use the configured summary and description. Resolved notifications could also lose their resolved identity when a custom summary was set
  - The Alert Monitor refresh-interval dropdown listed the acknowledgement filter options instead of 1m / 3m / 5m / 10m
  - Fetching alert rules as YAML timed out and returned a 500 for tenants with large rule counts
- **Control Plane**: The calendar in Data Usage hid the first date row when downloading a CSV