Telemetry from ai_reviews
Metrics are grouped by activity, quality, and risk to make trends easier to scan and compare.
Track how much review work is happening and how often PRs require additional runs.
Total PRs reviewed
101
Unique PRs
101
Distinct PR identity (repo + PR id)
Review runs
155
All collected runs for those PRs
Re-review rate
21.8%
PRs with 2+ runs
Avg runs per PR
1.53
Total runs / unique PRs
Median runs to non-blocking
1
First run index where final decision is non-blocking
Median first→latest time
34m 42s
Across PR lifecycles with multiple runs
Understand decision quality, approval flow, and overall evaluation index.
First-pass approval
49.5%
Run 1 code verdict = APPROVED
Eventual approval
54.5%
Latest run code verdict = APPROVED
Approval rate
67.1%
Code verdict = APPROVED / with verdict
Block rate
92.1%
final_decision_blocking
EQI
48.6
0-100 cohort
Highlight security findings and operational cost of review cycles.
Avg review duration
106s
Code + security duration (sample)
Vulnerabilities (counts sum)
8769
Dependency severity counts in sample
Secrets intercepted
1
secret_scan_secrets_found