[currentbits]

Adds three workflow improvements:

• CLIProxy usage: opt-in Claude session/weekly quota rows alongside native accounts. Settings normalize dashboard/server URLs, seal and redact the management key, and keep external accounts out of native routing and account switching. Uses the current v8 management API, with an explicit legacy URL or a single missing-endpoint fallback for older servers.
• GitLab review verdicts: display reviewer states and request changes only when the host advertises that action and confirms the current account's assignment and permissions. Rechecks before submission, verifies command execution and reviewer-state readback, and ignores stale UI results. Does not publish unrelated draft notes or infer overall merge approval.
• Multilingual local dictation: optional runtime-paired Nemotron 3.5 model with 32 supported locales and automatic detection. Preserves English installs/settings, checks compatibility before capture/download, shows exact selected download size, packages the shared catalog, and pins the model/language through local recording and cloud-fallback replay.

Validation: all 3,859 renderer tests, typecheck/preload sync and production build pass. Focused backend, mounted UI, protocol, secret-storage and packaging regressions pass. All 5,908 integrated Electron tests pass (4 skips); the full suite used controlled local concurrency to avoid process exhaustion. All 14 CI checks pass, including booting the four native server bundles with the shared speech catalog. All three new portable backend suites are included in Windows CI.

Limits: CLIProxy supports one configured server and Claude quotas; remote servers require HTTPS. GitLab hosts without the required GraphQL capability show an unavailable reason; merge blocking remains the host's decision. Local speech keeps the existing supported runtime platforms. Pinned native runtime/model-catalog smoke and fake-inference subprocess/WebSocket tests were run; model weights were not downloaded and acoustic accuracy was not measured. No live credentials or forge review submissions were used.

Closes #1872
Closes #1880
Closes #1888

Created with gpt-6.1-sol in Codex, using GPT-6.1 Sol workers.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-07T19:30:13.598872Z	1c70a0b	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.