chore(fork): sync stable v0.0.45 while preserving fork features

[YJJosh]

Brings upstream's latest stable release, v0.0.45 (66 commits since v0.0.44), into the fork. Where upstream and the fork changed the same thing, the fork's version wins.

What changed upstream

Threads without a project, restarting agent sessions from ⌘K to pick up new skills and plugins, one-click provider updates, Expo SDK 58 / React Native 0.88, vite-plus 1.0, a question-first preview for answered user-input rows, and CI split into parallel lint, typecheck and desktop-build jobs.

How the 7 conflicts were resolved

• ci.yml: took upstream's new parallel job layout and reapplied the fork's only change: GitHub-hosted runners instead of Blacksmith, and dropping the two Blacksmith-only mirror lines.
• release.yml: kept the fork's copy in full. It's hard-disabled (no triggers, every job if: false) because Dulli releases run through fork-desktop-release.yml, and keeping it whole avoids new upstream jobs slipping in without the disable.
• apps/mobile/app.config.ts: upstream's version bump plus the fork's Dulli scheme and version override.
• remoteRegistration.ts: the new Live Activity title uses the Dulli brand name, like the rest of the fork's mobile titles.
• MessagesTimeline.tsx: upstream's question-first preview for user-input rows, plus the fork's tool headings and success indicator. They apply to different row types.
• third-party-licenses.config.json: union of both lists, so the Pi package notices stay.
• pnpm-lock.yaml: upstream's lockfile, then reinstalled so the fork's dependencies (Pi SDK, Workler) are back.

The three workflows that merged without conflicts (release-desktop.yml, desktop-macos-preview-publish.yml, mobile-eas-preview.yml) stay disabled. No README changes this time.

Pi needs nothing new for upstream's ⌘K session restart: it goes through the generic registry refresh and Pi resumes its conversation on restart.

Testing

• Typecheck of every package, knip:check, and lint and format on the 249 files the fork changes relative to v0.0.45 are clean.
• Tests for the conflicted areas plus the Pi, usage, migration and workflow suites pass. One ProviderRegistry test ("re-probes when settings change the codex binaryPath") fails locally on macOS, and also fails on current main: upstream's new provider-update check finds /opt/homebrew/bin/brew, which the Linux CI runner doesn't have.
• build:desktop (including the strict license check) and release smoke pass. verify-preload-bundle fails locally with window is not defined, the same macOS-only quirk as last sync; the file matches upstream and it passes on Linux CI.
• Dev server through Portler on a VACUUM INTO copy of the real Dulli database: migrations 44–55 ran cleanly, the app paired and loaded with Dulli branding, the Fork features settings and Pi (v1.0.0, authenticated) are intact, and the Agents and Shared terminals surfaces are in the surface menu. No console errors.

Release pipeline fix

The Dulli release run from today (37049262590) failed at Build WSL CLI runtime: the single-executable check rejected the server bundle's import("workler"), because a Node single-executable can't import() a file-backed package. Every desktop build waits on that job, so no artifacts were built, and nothing was tagged or published.

• WorklerWorkspaceService now loads Workler through createRequire, the same way node-pty is loaded.
• The packaged-bundle test now fails if any external package is imported with ESM syntax. It reproduces the CI error without the fix.
• Verified locally on macOS with Node 26.8.2: the single executable builds, the CLI archive builds and passes smoke-cli-archive, and the macOS Dulli desktop build produces the .dmg and .zip. The Linux and Windows builds can only be confirmed by a real release run.

Workler 0.2.2

• Bumped workler from 0.2.1 to 0.2.2. The release only adds an optional copyLinks workspace setting; the APIs Dulli calls are unchanged, so no code changes were needed. The lockfile diff is limited to Workler's own entries.
• Server typecheck, the Workler tests (including the packaged-bundle test), the CLI packaging tests and release smoke pass.

Heads-up

v0.0.45 now requires Codex CLI >=0.159.0.

---

Made with Claude Opus 5.5 in the pi coding-agent harness

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	−40 B (−0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−3 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−37 B (−0.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.2 KiB	−44 B (−0.1%)	66.4 KiB	✅
Codex	Live turn messages	10	9	−1 (−10.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+7 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+9 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	−2 B (−0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 6ef69b6 · PR result: b20e0fd · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.