Subscription capacity, request usage and operational diagnostics.
Subscription capacity
Status: ok. Checked: 2026-10-30 12:26 UTC.
claude
Unknown plan
5-hour94.0% remaining
Resets 2026-10-30 16:09 UTC
Weekly61.0% remaining
Resets 2026-11-01 01:59 UTC
Model quota 3: Weekly100.0% remaining
Resets 2026-11-01 02:00 UTC
codex
pro
Weekly74.0% remaining
Resets 2026-11-04 06:39 UTC
codex credits: None.
Provider quota is a snapshot and includes all clients using the account. Unknown capacity is not unused capacity. Token totals below measure Ralph usage, not subscription capacity.
Request activity
Ralph request records include review and follow-up retries and failed requests; duplicate response usage is counted once. All periods use UTC.
Today
4,880
Request records
This week
19,858
Request records
This month
19,858
Request records
Daily request records · 30 daysWeekly request records · 12 weeksMonthly request records · 12 months
Review usage
Durable Ralph ledgers include retries and failed requests; duplicate response usage is counted once. Follow-up assessments appear separately. Token totals sum reported fields; partial means some requests omitted that field. Reasoning tokens are part of output tokens, not an additional total. Failed requests include timeouts and other uncertain outcomes; 429 rate limits are a subset. Model family identifies the provider, not the account or authentication method.
Current month
Provider
Records
Input tokens
Cached tokens
Output tokens
Reasoning tokens
Failed requests
Rate limits (429)
Unknown usage records
claude
8,444
1,062,723,204 (partial)
1,012,602,018 (partial)
11,621,178 (partial)
0 (partial)
0
0
15
openai
11,388
448,793,778 (partial)
356,327,552 (partial)
3,584,247 (partial)
1,789,866 (partial)
41
0
66
Lifetime
Provider
Records
Input tokens
Cached tokens
Output tokens
Reasoning tokens
Failed requests
Rate limits (429)
Unknown usage records
claude
8,444
1,062,723,204 (partial)
1,012,602,018 (partial)
11,621,178 (partial)
0 (partial)
0
0
15
openai
11,388
448,793,778 (partial)
356,327,552 (partial)
3,584,247 (partial)
1,789,866 (partial)
41
0
66
Model usage
Provider
Model
Attempts
Input tokens
Cached tokens
Cache rate
Output tokens
Reasoning tokens
Failed requests
Rate limits (429)
claude
claude-opus-5-5
8,444
1,062,723,204 (partial)
1,012,602,018 (partial)
N/A
11,621,178 (partial)
0 (partial)
0
0
openai
gpt-6-astra
1,347
89,417,467 (partial)
75,140,096 (partial)
N/A
469,709 (partial)
200,478 (partial)
0
0
openai
gpt-6-luna
8,156
275,705,335 (partial)
216,371,584 (partial)
N/A
2,456,410 (partial)
1,336,265 (partial)
28
0
openai
gpt-6.1-sol
1,885
83,670,976 (partial)
64,815,872 (partial)
N/A
658,128 (partial)
253,123 (partial)
13
0
Routing and review distributions
Routing outcomes
Tier
Reviews
routine
75
sensitive
382
standard
78
Review distributions
Metric
Count
Median
P90
Max
Tool calls
1,665
141.0
273.0
489.0
Model seconds
1,665
474.8
1,262.3
4,456.8
Conceptual concern citations
16
2.5
4.0
6.0
Conceptual concern alternatives
16
1.0
1.0
1.0
Production trends
Daily counts, UTC. Days without recorded activity are zero.
Undated records are excluded. Addressed findings use the PR head commit date. Missing daily counts leave gaps and prevent subsequent cumulative totals for that metric.
Review days use UTC timestamps from private trace filenames and are present only when trace files can be unambiguously matched to saved result rows by PR and head.
Addressed days use the UTC new_head commit committer timestamp in the follow-up checkout; assessed_at is not used for trend bucketing.
Undated records are excluded: 798 reviews.
Accounting and limitations
1,665 saved reviews; 614 pull requests have been queued. 525 completed archaeology research. Current month review usage is 19,832 request records. Lifetime tracking has 19,832 request records for 1,799 review jobs, including 5 recorded failed attempts or retries.
Reviews without findings export
0
Reviews without stage diagnostics
1,130
Monthly trend uses request month.
Model seconds are summed model turn elapsed_seconds, not wall time.
Bot verification is not human validation.
Shared attribution is noncausal.
Token totals include reported usage only; missing token fields are marked partial.
Subscription quota includes other clients sharing the same account.