Repository navigation
fix(codex): make auto-review free from October 6 - #1354
validatedev wants to merge 4 commits into
Conversation
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
There was a problem hiding this comment.
Important
The zero-cost rule also rewrites usage from before the announced free-service change. Its historical scope needs to be corrected or supported by an explicit retroactive policy before merging.
Reviewed changes Reviewed the Codex pricing rule, parser/cache changes, model-breakdown behavior, regression tests, documentation, and screenshot.
- Known Free Pricing: Recognizes
codex-auto-reviewas $0 before catalog lookup or paid fallbacks, including cached events with the former Luna mapping. - Parsing And Cache: Removes the dated auto-review fallback and invalidates persisted parsed events while retaining Luna Reserve pricing.
- Breakdown Visibility: Keeps free auto-review tokens in totals and a separate named row alongside paid models.
- Regression Coverage: Adds exact assertions for fallback handling, cached mappings, speed tiers, mixed models, and spend tiles; documents historical recalculation.
The full macOS SwiftPM suite could not be run in this Linux environment; GitHub's Build and Test check was still pending when inspected.
gpt-6.1-sol ｜ 𝕏
validatedev
changed the title
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes Reviewed the timestamp-aware pricing correction in 4b5ae69 against the prior review at 445e2c3.
- Restored Historical Estimates: Preserved dated paid-model estimates before the documented October 6, 2026 00:00 UTC cutoff, while keeping requests at or after it free.
- Propagated Request Time: Required timestamps throughout shared Codex pricing and passed native, pi, and OpenCode request times into that policy.
- Separated Cached Rates: Keyed OpenCode prepared rates by effective pricing model so historical and free requests cannot reuse each other's rates.
- Scoped Carried Costs: Overrode pi's carried estimates only for free-era Codex auto-review, preserving earlier positive costs and ordinary provider behavior.
- Expanded Boundary Coverage: Added exact assertions for both sides of the cutoff, UTC offsets, cached mappings, service tiers, fallback handling, and OpenCode cache order; aligned the documentation.
The macOS SwiftPM suite could not be run in this Linux environment. GitHub's Build and Test check was pending when inspected.
gpt-6.1-sol ｜ 𝕏
|
@codex review |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4b5ae69d42
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes Reviewed 492b816 against the prior Pullfrog review at 4b5ae69, with the full PR diff as context.
- Preserved Mixed-Period Visibility: Kept
codex-auto-reviewin its own breakdown row even when historical paid estimates make its aggregate cost positive but below the normal folding threshold. - Added Regression Coverage: Asserted exact Last 30 Days totals, auto-review tokens and historical cost, and continued folding of ordinary small Luna usage into
Other. - Updated Documentation And Screenshot: Documented mixed-period behavior and added a fixture-rendered screenshot showing the retained auto-review row.
The macOS SwiftPM suite could not be run in this Linux environment. GitHub's Build and Test check was still in progress when inspected.
gpt-6.1-sol ｜ 𝕏
|
@codex review |
|
Codex Review: Didn't find any major issues. Breezy! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
This branch has not been deployed
1 participant
TL;DR
Treat Codex auto-review as free from October 6, 2026 at 00:00 UTC, while preserving earlier model-rate estimates and the
codex-auto-reviewname and measured tokens. The policy follows the calendar date of Tibo’s announcement that auto-review is now free for users signed in through ChatGPT.What was happening
What this changes
Heads-up
The announcement establishes a calendar date, without an exact activation instant or a retroactive waiver. OpenUsage uses the start of October 6 in UTC as its date boundary. Local history is recalculated using each request’s recorded timestamp; synced estimates from another Mac update when that source Mac refreshes in the updated app.
Tests
swift test: 1,568 passed, 3 gated live/parity tests skipped, 0 failures.CONFIG=debug ./script/build_and_run.sh verify: rebuilt, restarted, and confirmed the app is running.git diff --checkpassed.