Repository navigation
[Review 134] [4/7] feat: explain session activity, usage and attention - #4
BIackFIame wants to merge 44 commits into
Conversation
|
@codex review Please review this PR's diff for correctness, regressions and security issues. Review only; post findings on GitHub without changing code. This is the contributor-fork review for howdeploy#134, at the identical head commit |
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. |
There was a problem hiding this comment.
💡 Codex Review
CanvasTTY/src/main/services/UsagePrices.ts
Lines 34 to 35 in 21d5c99
The duplicate key is built from the untrimmed model name, but the stored row is trimmed. A user can therefore save both model-a and model-a ; both become identical persisted entries, and priceForUsage uses find, so only the first price silently controls calculated costs while the other appears editable. Build the duplicate key from the normalized model name before accepting the table.
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".
…cklog/1-orchestration
Addressed in 3f25388. Model names are normalized once before duplicate-key validation and persistence. Save/load regressions reject whitespace-equivalent duplicates, preserve the previous valid table on failure, and verify distinct providers remain valid. Original review: #4 (review) |
|
@codex review Additional verification round 1 of 5: correctness and integration. Review current head 3f25388. Parent fixes have been integrated with observability: accepted lifecycle gating, pre-turn checkpoint barriers, restore leases, scoped review baselines, and distinct OpenCode turn IDs. Glasses attention now follows current conditions; model price keys normalize before uniqueness checks. Please examine correctness and integration regressions across lifecycle events, timelines, usage and devices. Report actionable defects; please do not apply changes. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3f253883f9
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".
|
@codex review Additional verification round 2 of 5: races and event ordering. Review current head ea765f9. Shared conversation usage now deduplicates cumulative snapshots by identity and observation order, including reload/reset cases. Glasses response attention clears on accepted input before delayed provider lifecycle updates. Please focus on races, ordering, asynchronous lifecycle/report updates, usage aggregation and device state boundaries. Report actionable defects; please do not apply changes. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ea765f97e3
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".
|
@codex review Additional verification round 3 of 5: security and provider/device boundaries. Review current head 69dad37. Response acknowledgment now uses successful structured submissions, and retained attention is filtered newest-first by current conditions. Please focus on security boundaries, provider and device integration, lifecycle correlation, data redaction and error paths. Report actionable defects; please do not apply changes. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 69dad37507
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".
|
@codex review Additional verification round 4 of 5: compatibility and error handling. Review current head 1da0f6e. Activity telemetry is restricted to the verified detector with opaque correlators; OpenCode tool outcomes follow its provider callbacks. Please focus on compatibility, error and cancellation paths, current-turn correlation and regressions across supported providers. Report actionable defects; please do not apply changes. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1da0f6eb91
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".
|
@codex review Additional verification round 5 of 5: final regressions. Review current head 7eb562b. Kimi and Hermes core overlays now register completion events independently of optional native plugins. Please perform a final regression review across observability, usage accounting, attention state, provider lifecycle and tool outcomes, and plugin/device boundaries. Report actionable defects; please do not apply changes. |
|
Codex Review: Didn't find any major issues. 🚀 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". |
|
@codex review Additional verification round 6: bounded verification after Windows CI harness repair. Review current head 5899829. The fifth additional review was clean, but Windows CI exposed incorrect command quoting in the new integration test. The test now uses Node shell mode, preserving actual generated-command and delivery assertions on Windows. Please perform a final regression review across observability, usage accounting, attention state, provider lifecycle and tool outcomes, and plugin/device boundaries. Report actionable defects; please do not apply changes. |
|
Codex Review: Didn't find any major issues. More of your lovely PRs please. 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". |
a68bd8a to
ac912f9
Compare
Review purpose
Review-only copy of the prepared CanvasTTY series in the contributor's fork. This step covers: [4/7] feat: explain session activity, usage and attention.
Related: howdeploy#134
The head is the exact upstream PR commit
21d5c99f7957ec98ea5aa893127ec6360b6084fe. The base isbacklog/2-worktreesat2b988218680596776177862ef2fb53daa5549c6b. The upstream PR is cumulative against upstream main; this review compares only this step against its predecessor, with inherited code available as context.This PR is for review, not merging. Keep feedback attached to this exact commit. Source validation, live-test limitations and series context are recorded in the upstream PR; no new tests were run when creating this review copy.