CodexScope v0.1.7
Fixes inflated peak TPM caused by repeated Codex token snapshots in overlapping JSONL files.
- Deduplicates repeated
token_countsnapshots by session, model, and cumulativetotal_token_usage. - Bumps the local cache format so affected users regenerate clean data once.
- Adds a regression test for repeated snapshot deduplication.
Observed on local data:
- Before: peak TPM
164.68M - After: peak TPM
3.92M - Records after dedupe:
44,459 - Total after dedupe:
5.09Btokens
Checks:
go test ./...npm run build:frontendnpm run release:local