.png)
AI Summary
AI-generated release health summaries, health scores, and app version metrics
The AI Summary tab under Anomaly Detection provides an automated, natural-language executive summary of your app's release health. It compiles metrics and detects regressions between app versions to help product and engineering teams understand release quality at a glance.
1) Version-Level Health Metrics
For each active app version, the panel calculates a set of summary metrics that represent the stability of the release:
- Health Score – A circular 0-100 rating indicating overall stability. Higher is better.
- Adoption Rate – The percentage of active devices running this specific version.
- Anomalous Devices – The percentage of devices running this version that encountered at least one anomaly.
- Incidents – The total count of anomalies detected for this version.
- Critical Crashes – The count of fatal issues that caused the app to crash.
Screenshot: AI Summary metrics
2) AI-Generated Release Health Summary
Expanding the AI-Generated Summary section reveals a structured, natural-language analysis generated by LogDrop's LLM engine.
Executive Summary
A high-level diagnostic of the version's health. For example, it will warn you if a critical severity crash is affecting a new release, even if the overall adoption rate is still low.
Comparison & Regression Analysis
Compares the current release's health score and incident rate directly against previous versions (e.g., comparing v1.0.63 against v1.0.62) to detect stability improvements or regressions.
New Regressions
A bulleted breakdown of newly introduced issues in the release, such as new crash signatures or configuration issues (e.g., missing API keys or bundle identifier mismatches).
Adoption & Traffic Weight
Contextualizes the issue frequency by providing traffic weight data (e.g., 11.02% (99 registered devices out of 898 total active platform devices)).
Critical Issues & Crashes
Details critical issues, listing the exception signature, severity level, total events observed, timeline of occurrences, and a predicted root cause.
Other Key Findings
Highlights non-fatal high-severity issues (e.g., APNs push backend failures or missing parameters in notification handlers) along with device impact counts.
Screenshot: AI-Generated Summary expanded
