Repository navigation
Add OpenAI Tokens/min tray metric - #925
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Merge Risk: ⚪ Minimal · up to
|
| Check name | Status | Explanation | Resolution |
|---|---|---|---|
| Provider Data Stays Siloed |  |
The pull request adds a provider-specific branch in shared code outside the allowed paths. rust/src/core/provider/id.rs:168-170 adds reports_tokens_per_minute() with `matches!(self, ProviderId::Op… |
Remove the ProviderId::OpenAIApi match from rust/src/core/provider/id.rs. Keep the default capability false in the shared Provider trait, override reports_tokens_per_minute() in rust/src/providers/openaiapi/mod.rs, and route ID-on… |
| Ui Changes Include Windows Proof |  |
The PR changes visible UI. The diff adds a Tokens/min option to the settings picker and changes tray tooltip and status-row text. The description reports Chromium proof and states that desktop CUA pro… | Run the proof on a fresh Windows build and add screenshots or a completed proof note to the PR description. Cover the settings selection and persistence, tray tooltip and status row with and without throughput, and unchanged tray icon pixel… |
-
Check name Status Explanation Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Secrets Handled Safely No changed code exposes a credential. The new tray path reads only the numeric latest_tokens_per_minuteusage count and formats it as a rate; it does not read or format API keys, cookies, OAuth cred…No Unapproved Dependencies The pull request changes 32 files, and none are Cargo.toml, package.json, npm/yarn lockfiles, or pnpm-lock.yaml. It also does not change the pinned pnpm packageManager version. No dependency addition … Description Check Check skipped - CodeRabbit’s high-level summary is enabled. Title check The title is a short imperative summary that accurately describes the main change: adding an OpenAI Tokens/min tray metric. Full details: Secrets Handled Safely
-
Explanation
No changed code exposes a credential. The new tray path reads only the numeric
latest_tokens_per_minuteusage count and formats it as a rate; it does not read or format API keys, cookies, OAuth credentials, or authorization headers. The added settings and provider fields contain only metric/capability metadata. The diff adds no logging calls, credential snapshots, or credential-like literals, and it contains no deleted, renamed, or overwritten files.
Full details: No Unapproved Dependencies
-
Explanation
The pull request changes 32 files, and none are Cargo.toml, package.json, npm/yarn lockfiles, or pnpm-lock.yaml. It also does not change the pinned pnpm packageManager version. No dependency addition requires maintainer approval.
-
Full details: Provider Data Stays Siloed
-
Explanation
The pull request adds a provider-specific branch in shared code outside the allowed paths.
rust/src/core/provider/id.rs:168-170addsreports_tokens_per_minute()withmatches!(self, ProviderId::OpenAIApi). This directly matches a specificProviderIdin shared core code. The new tray and provider-detail paths call this method, so the branch is introduced by this pull request. The changed code otherwise gates the metric to OpenAI and does not show evidence of identity or account data crossing providers.Resolution
Remove the
ProviderId::OpenAIApimatch fromrust/src/core/provider/id.rs. Keep the default capability false in the sharedProvidertrait, overridereports_tokens_per_minute()inrust/src/providers/openaiapi/mod.rs, and route ID-only capability lookups through a generic API inrust/src/core/provider_factory.rsor another allowed provider-specific mechanism. Do not add provider-specificProviderIdcases in shared tray or core code.
Full details: Ui Changes Include Windows Proof
-
Explanation
The PR changes visible UI. The diff adds a Tokens/min option to the settings picker and changes tray tooltip and status-row text. The description reports Chromium proof and states that desktop CUA proof is “to follow” and “Still to prove on a fresh Windows desktop build.” It does not include completed fresh-Windows screenshots or proof evidence.
Resolution
Run the proof on a fresh Windows build and add screenshots or a completed proof note to the PR description. Cover the settings selection and persistence, tray tooltip and status row with and without throughput, and unchanged tray icon pixels.
- Fix all pre-merge checks with AI
✨ Finishing Touches
-
📝 Generate docstrings
-
- Commit to this branch
- Create a new PR
🧪 Generate unit tests (beta)
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
…ns-menubar # Conflicts: # apps/desktop-tauri/scripts/proof-scenarios.json
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
UI proof (hybrid CDP + CUA, build 3971958)Surface: Settings > Providers > OpenAI API (proof mode Setup: fresh debug build of 3971958 (
Screenshots (cua-driver, real pixels): Not covered:
Focus: FAIL. The run started with the foreground on the terminal ( Note: with a pinned seed, the startup refresh is skipped, so the tray tooltip reads only |
1 participant
Summary
Ports the optional Tokens/min menu bar token from upstream CodexBar v0.75.0 (steipete#4495, commit 5a40ac4). It builds on the existing OpenAI Live Usage data path (#893, #897) and adds no new fetches.
MetricPreference::TokensPerMinute(tokensPerMinuteon the bridge and insettings.json).Provider::reports_tokens_per_minute()trait hook. The typed fact lives once onProviderId::reports_tokens_per_minute()and the trait default delegates to it, so the shell can check the capability from a provider id without instantiating a provider. Onlyopenaiapiqualifies.ProviderDetail.tokensPerMinuteMetricexposes it, so the Settings > Providers > OpenAI API "Menu and floating bar metric" picker offers Tokens/min without provider branching in shared code.OpenAI API 49M/min, using the latest reported minute (throughput.latestTokensPerMinute). Formatting matches upstreamtokenCountString + "/min", so the output looks like950/min,12K/min,1.5M/minor0/min.usage_metric.rsthe meter (and the floating bar) keeps the Automatic lane. The change there is a single match arm.MetricTokensPerMinuteandTrayTokensPerMinuteUnavailableare in all 10.ftlfiles andkeyGroups/shell.ts.openaiapito the proof catalog plus a browser scenario for the picker.The upstream sparkline token (
tokensSparkline) is not ported. A 34-px graph next to the icon has no Windows tray equivalent.Review fixes (second commit)
MetricPreferenceisrename_all = "lowercase", soTokensPerMinutewould have been written astokensperminutewhile the bridge usestokensPerMinute. It now serializes astokensPerMinutewith the lowercase spelling kept as a read alias, same asExtraUsage. The pre-existingMonthlyPlanhad the same mismatch and is fixed the same way (monthlyPlan, aliasmonthlyplan), so files written by either spelling keep loading. Regression tests cover both spellings for both metrics and confirm a settings file with the camelCase value loads without falling back to defaults (rust/src/settings/tests/metric_preference.rs).tokens_per_minute_status(used by both the tray status row and the tooltip) now also requiresProviderId::reports_tokens_per_minute(), so atokensPerMinutevalue saved against a provider that cannot report a rate falls back to the Automatic label instead of leaking the rate onto that provider's row. Covered bytokens_per_minute_metric_on_a_provider_without_a_rate_falls_back.settings-openaiapi-tokens-per-minute-metricnow actually selects Tokens/min in the picker and asserts the persisted value through the mocked bridge (update_settingswithproviderMetrics.openaiapi = "tokensPerMinute"), then checks the select reads the new value back.MetricPreference::display_name()("Tokens/min") has no callers in the app, the shell or the CLI; the picker labels come from the localizedMetricTokensPerMinutekey, so nothing user-facing goes unlocalized.Commands run
cargo fmt --allcargo test --manifest-path rust/Cargo.toml --lib settings::: 208 passed, 0 failed (includes the 3 new metric serde tests)cargo test --manifest-path rust/Cargo.toml --lib openai: 120 passedcargo test --manifest-path apps/desktop-tauri/src-tauri/Cargo.toml tray_presentation: 59 passed (includes the new fallback test);tray_bridge: 13 passed;provider_detail: 9 passedcargo clippy --manifest-path rust/Cargo.toml --all-targets -- -D warnings: cleancargo clippy --manifest-path apps/desktop-tauri/src-tauri/Cargo.toml --all-targets -- -D warnings: cleanpnpm test: 145 files, 1373 tests passed (one transient failure in an unrelated test on the first run; passed on both reruns)pnpm run build: OKpnpm proof:check(Chromium, mocked IPC): 13/13 passedProof
Tray-affecting: desktop CUA proof to follow. Still to prove on a fresh Windows desktop build:
get_settings_snapshotreportsproviderMetrics.openaiapi = "tokensPerMinute"and thatsettings.jsoncontains the camelCase value.OpenAI API: <n>/minand the tray-menu status row readsOpenAI API <n>/min.Reference: upstream steipete/CodexBar steipete#4495 (read-only).