Telemetry from ai_reviews
Pass-1 is the default diff review; pass-2 is an optional agentic re-review after pass-1 when enabled.
Final-pass input (code + security)
30.5k
Final-pass output (code + security)
7.3k
Pass-1 input (non-agentic)
12
Pass-1 output (non-agentic)
6.7k
Pass-2 tokens (agentic)
77.5k
Pass-1 share (of pass-1 + pass-2)
8%
Avg core IO tokens / PR
22.2k
Avg duration
1m 7s
Total duration (hours)
0h 2m 24s
Cache hit rate
79.4%
Est. cost
$0.63
Re-review rate
50%
Avg runs / PR
1.5
Activation rate
100%
Avg planner iterations
2.5
Avg context retrieved
19.1 KB
Planner token overhead
15.8k
Avg planner duration
14s
Activation rate
50%
Avg planner iterations
3
Avg context retrieved
59.3 KB
Planner token overhead
23.0k
Avg planner duration
37s
Avg cost to non-blocking: — across 0 resolved PR lifecycles.
Cache hit % is cache-read tokens divided by cache read plus cache-creation tokens, split by whether the review row is the first run on a PR or a later re-review run.
Cohort | Cache hit rate |
|---|---|
| First runsRun index 1 on each pull request | 78.9% |
| Re-runsRun index 2 and above on the same PR | 100% |
Re-runs show cache hits 21.1 percentage points above first runs in this filter.