Commit graph

21 commits

Author SHA1 Message Date
kami
0bfb4803e7
feat(daemon): add Phase 2C CLI wrappers (#2179)
* feat(daemon): add phase 2c cli wrappers

Co-authored-by: multica-agent <github@multica.ai>

* fix: handle desktop-gated CLI imports

Co-authored-by: multica-agent <github@multica.ai>

* fix: pass sidecar ipc path to agent wrappers

Co-authored-by: multica-agent <github@multica.ai>

* fix: make agent wrapper env explicit

Co-authored-by: multica-agent <github@multica.ai>

* fix(daemon): preserve CLI import and diff edge cases

Co-authored-by: multica-agent <github@multica.ai>

---------

Co-authored-by: multica-agent <github@multica.ai>
2026-05-28 09:08:31 +00:00
lefarcen
df8a0faff6
feat(runtimes): register AMR (vela) as an ACP stdio agent (#2355)
* feat(runtimes): register AMR (vela) as an ACP stdio agent

AMR is the vela CLI's ACP runtime mode. `vela agent run --runtime opencode`
speaks ACP JSON-RPC over stdio (see vela's
`specs/current/runtime/manual-agent-run-openrouter.md`); per
`docs/new-agent-runtime-acp.md` we expose it through the same `streamFormat:
'acp-json-rpc'` transport that already powers Hermes, Devin, Kimi, etc.

The new `defs/amr.ts` is the entire wiring — `buildArgs` returns
`['agent', 'run', '--runtime', 'opencode']`, `fetchModels` reuses
`detectAcpModels`, and the fallback list seeds the OpenRouter ids vela's
e2e baseline uses. `executables.ts`/`app-config.ts`/`metadata.ts` get the
matching `VELA_BIN`/`VELA_LINK_URL`/`VELA_RUNTIME_KEY`/`VELA_OPENCODE_BIN`
allowlist + install/docs URLs, so users can configure the per-agent env in
Settings without leaking into other adapters.

Coverage: `tests/fixtures/fake-vela.mjs` is a minimal ACP stub that returns
the documented `initialize` / `session/new` / `session/set_model` /
`session/prompt` shapes; `tests/amr-acp-integration.test.ts` spawns it via
`child_process.spawn` and drives a full turn through `attachAcpSession` and
`detectAcpModels`, so the ACP transport contract for AMR is end-to-end
verified locally even before a real `vela` binary is installed.

Validated:
- pnpm guard
- pnpm typecheck (all workspace projects)
- pnpm --filter @open-design/daemon test (2881/2881)

Deferred: real OpenRouter-backed turn through a built `vela` binary —
the runtime def needs no changes for that path, only `VELA_RUNTIME_KEY`
and `VELA_LINK_URL` in env (or Settings).

* fix(runtimes/amr): pin a concrete default model and bare openai ids

End-to-end validation against a freshly-built `vela` (nexu-io/vela@main)
+ OpenRouter surfaced two contract details the first AMR runtime def
got wrong:

1. vela rejects `session/prompt` with `session/set_model must be called
   before session/prompt`. attachAcpSession in apps/daemon/src/acp.ts
   skips set_model whenever the picked model is the synthetic 'default'
   id, so AMR's fallback list must NOT include DEFAULT_MODEL_OPTION. The
   def now ships a concrete `gpt-5.4-mini` as both `fetchModels`'
   default option and `fallbackModels[0]`, which makes attachAcpSession
   always send a real `session/set_model` for AMR turns.

2. `vela --runtime opencode` auto-prepends `openai/` to whatever modelId
   it forwards to opencode's openai provider. With OpenRouter-style ids
   like `openai/gpt-5.4-mini`, opencode receives the double-prefixed
   `openai/openai/gpt-5.4-mini` and replies `ProviderModelNotFoundError`.
   The new fallback list ships the bare ids opencode's openai registry
   actually knows about (gpt-5.4, gpt-5.4-mini, gpt-5.4-fast, etc.).

Stub + tests:
- tests/fixtures/fake-vela.mjs now enforces the set_model gate the same
  way real vela does, so a regression that silently goes back to
  model: 'default' would surface as a fatal error in tests instead of a
  hidden production failure.
- tests/amr-acp-integration.test.ts pins both contracts: no 'default' /
  no 'openai/' prefix in fallbackModels, and a negative case that
  asserts session/prompt fails when no model is set.

Adds `apps/daemon/scripts/verify-amr-real-vela.mjs` — a small dev-time
runner that drives `attachAcpSession` against a real `vela` binary and
prints the daemon's chat events, so future protocol drift can be checked
against an actual OpenRouter call.

Verified locally: `vela agent run --runtime opencode` + OpenRouter
returns the prompted string ("AMR-E2E-PASS") through the full daemon
pipeline; daemon test suite stays 2883/2883.

* fix(runtimes/amr): substitute concrete model when chat run sends 'default'

A plugin-driven AMR run from the UI surfaced a real-world hole in the
prior commit:

  json-rpc id 3: session/set_model must be called before session/prompt

The Default-design-router plugin (and any caller that doesn't pin a
real model) sends `model: 'default'` straight through, which the AMR
runtime def cannot accept — vela rejects `session/prompt` without
`session/set_model` and attachAcpSession skips set_model whenever
model === 'default'. Just leaving DEFAULT_MODEL_OPTION out of the
adapter's `fallbackModels` is not enough: the chat-run handler in
server.ts still forwarded 'default' verbatim.

This adds `resolveModelForAgent(def, resolved, env?)` as the
single source of truth for the substitution:

  1. If the caller picked a real id, pass it through.
  2. Else, if `def.defaultModelEnvVar` is set and the daemon process
     env has a non-empty value for it, return that (operator escape
     hatch — see below).
  3. Else, if the def's `fallbackModels` does NOT contain a 'default'
     id, return `fallbackModels[0].id`.
  4. Else, return the original value (the historic shape — defs that
     list 'default' themselves are untouched).

AMR sets `defaultModelEnvVar: 'VELA_DEFAULT_MODEL'`, so when
opencode's openai-provider registry deprecates `gpt-5.4-mini`
upstream, an operator can swap the fallback id without a code change
by exporting `VELA_DEFAULT_MODEL=gpt-5.5` before launching tools-dev
/ od. Worth noting the env var must live in the daemon's `process.env`
(Settings-UI per-agent env values only reach the spawned child, not
the daemon's resolver) — the new field's docblock spells this out.

Coverage:
- `tests/runtimes/resolve-model.test.ts` — 8 unit tests covering all
  four resolver branches plus the env-override happy path / fallback /
  ignore-when-user-picked-a-real-id case.
- `pnpm --filter @open-design/daemon typecheck` clean.

* chore(runtimes/amr): move AMR to the top of the base agent list

So `AMR (vela)` shows up first in the agent picker / status views,
ahead of claude / codex. Pure ordering change; no behavior delta.

* feat(amr): Sign-in / Sign-out button on the AMR Settings card

The first half of the AMR work assumed the operator would set
VELA_RUNTIME_KEY / VELA_LINK_URL on the daemon process and never
surfaced login state to users. This adds the missing UX so a fresh
install can drive the full path from Settings:

  - GET  /api/integrations/vela/status   reads ~/.vela/config.json
    for the active profile and returns { loggedIn, profile, user }
    (without leaking the runtime/control keys themselves).
  - POST /api/integrations/vela/login    spawns `vela login` once
    (409 if one is already in flight). The vela CLI opens the user's
    browser to the device-authorization page itself — Open Design
    only needs to kick the subprocess off.
  - POST /api/integrations/vela/logout   removes ~/.vela/config.json
    so the next status read returns logged-out.

`AmrAgentCard` is a dedicated agent-card component for AMR because
the existing `<button>` row can't host an interactive sub-control
(nested interactive elements). It polls /status after a login click
until the daemon reports loggedIn=true (or 5 minutes elapse), and
exposes a Sign-out action on hover. Other adapters (claude, codex,
hermes, …) keep their existing `<button>` card.

i18n: 8 new keys (settings.amrLogin / Logout / LoggingIn / etc.)
added to en + zh-CN. Other locales spread `en` and inherit the
English copy until translations land.

Coverage:
- `tests/integrations/vela.test.ts` pins the config.json reader
  against a tmp HOME — including the negative case where a profile
  has user info but no runtimeKey (still logged-out), and the
  secret-leak guard ("rt-secret-*" must not appear in the projection
  payload).
- `tests/components/AmrAgentCard.test.tsx` covers all four UI
  states (logged-out, logging-in, logged-in, logging-out) plus the
  click-propagation invariant the divergent card was built to keep.

`pnpm --filter @open-design/daemon test` 2901 / 2901 passing.
`pnpm --filter @open-design/web test` 1719 / 1719 passing.
`pnpm typecheck` + `pnpm guard` clean.

Dev script side-effects: `apps/daemon/scripts/verify-amr-real-vela.mjs`
no longer requires both VELA_RUNTIME_KEY and VELA_LINK_URL — if
VELA_PROFILE is set, the vela CLI is allowed to resolve credentials
from `~/.vela/config.json`. Added the two AMR `.mjs` fixtures to
`scripts/guard.ts` allowlist with the executable-fixture / dev-runner
rationale.

* fix(connection-test): substitute model for AMR before attachAcpSession

The chat-run path in server.ts already routes the requested model through
`resolveModelForAgent` so AMR / vela (whose CLI demands an explicit
`session/set_model` before `session/prompt`) gets the def's first
concrete fallback id when the chat run ships `model: 'default'`.
`connectionTest.ts` was wiring `attachAcpSession({ ..., model: model ?? null })`
directly, which made the Test Connection button on the AMR Settings
card deadlock with the same `session/set_model must be called before
session/prompt` error the chat-run path already handles — surfaced as a
permanent "Testing connection…" spinner in the UI.

Reuse the same helper here so Test Connection mirrors chat-run behavior.

* test(amr): three-layer end-to-end coverage for the AMR login + turn flow

The PR up to this point shipped runtime + UI code with unit-level Vitest
coverage. This commit adds the cross-layer regression net the live demo
relied on:

1. apps/daemon/tests/integrations/vela.routes.test.ts (HTTP, Vitest)
   Spins up the real daemon Express app via `startServer({port:0,...})`,
   persists `agentCliEnv.amr.VELA_BIN = <fake>` into app-config.json,
   and exercises every /api/integrations/vela/* endpoint against the
   extended fake-vela stub:
     - status reads ~/.vela/config.json under various states
     - login spawns the fake, waits for config.json to appear, returns
       pid + startedAt + profile
     - 409 already-running guard with the stub's delay knob
     - logout removes the file (idempotent)
     - secrets (runtimeKey / controlKey) never leak in the projection
     - login → status round-trip flips loggedIn=false → true

2. e2e/tests/amr/turn.test.ts (tools-dev orchestrated, Vitest)
   Boots a namespaced daemon + web pair through `createSmokeSuite`,
   inlines a self-contained fake `vela` binary that handles BOTH
   `vela login` (writes ~/.vela/config.json) and
   `vela agent run --runtime opencode` (ACP stdio with the
   `session/set_model must precede session/prompt` gate the real binary
   enforces), then drives a complete /api/runs lifecycle for
   `agentId: 'amr', model: 'default'` and asserts the assistant message
   captures the fake's streamed text. This is the test that would have
   surfaced today's plugin-default-model regression (the `set_model
   before prompt` error) at PR time instead of demo time.

3. e2e/ui/amr-login-pill.test.ts (Playwright)
   Mocks /api/agents + /api/integrations/vela/{status,login,logout}
   to drive the Settings AMR card through the full Sign in → Signed in
   → Sign out cycle. Pins the AmrLoginPill polling contract and the
   aria-label semantics (the pill's accessible name is "Sign out" once
   logged in, regardless of which label the hover-state text shows).

fake-vela.mjs extensions:
   - Handles `vela login` argv by writing
     ~/.vela/config.json for the active VELA_PROFILE and exiting 0 —
     mirrors real vela's on-disk side-effect without the device-auth
     loop.
   - FAKE_VELA_LOGIN_DELAY_MS knob so route tests can observe the
     in-flight state of the spawn lifecycle.
   - FAKE_VELA_LOGIN_USER_EMAIL / _USER_PLAN to assert the surfaced
     user fields end-to-end.

Validated:
   - `pnpm guard` + `pnpm typecheck` (all workspace projects)
   - `pnpm --filter @open-design/daemon test`: 2998 / 2998 passing,
     including the new 8-test integration suite.
   - `cd e2e && pnpm test tests/amr`: 1 / 1 passing.
   - `cd e2e && pnpm exec playwright test ui/amr-login-pill.test.ts`:
     1 / 1 passing (6.7s).

* feat(amr): package native cli and refine login ui

* feat(amr): wire vela cli beta packaging

* docs(amr): document vela ci packaging review

* docs(amr): refine vela ci integration review

* fix(ci): refresh nix pnpm dependency hashes

* fix(pack): clean up Vela CLI packaging

* fix(pack): bundle Vela CLI support files

* fix(amr): recover login attempts from stale auth state

* test: expand AMR and automations coverage

* fix(amr): address review follow-ups

* test(web): align tasks fixtures with contracts

* fix(daemon): type wildcard route params

* fix(ci): refresh PR merge validation

* fix(amr): clear env credentials on logout

* feat(settings): inline local CLI model configuration

* fix(amr): recognize daemon env credentials

* [codex] Fix Vela companion packaging (#2979)

* Fix Vela companion packaging

* Update Nix pnpm dependency hashes

* [codex] Surface AMR account failures (#2980)

* fix: surface AMR account failures

* fix: cover AMR recovery error guidance

* chore: bump beta base version to 0.8.1 (#2990)

* Fix AMR profile and packaged runtime review issues

* Detect packaged AMR OpenCode companion tree

* feat(web): polish AMR frontend flows

* Polish AMR onboarding card

* fix: read AMR login state from dot-amr config (#3048)

* test: tighten AMR credential and packaging coverage

* test: restore AMR executable test env helper

* [codex] Fix packaged mac Dock identity and AMR label (#3076)

* Fix packaged mac sidecar Dock identity

* Rename AMR assistant label

* Fix AMR live models and dot-amr login state (#3073)

* fix: read AMR login state from dot-amr config

* fix: load live AMR models before runs

* fix: point AMR onboarding link to production wallet

* fix: address AMR model review feedback

* fix: persist live AMR model fallback

* [codex] Fix AMR link catalog model ids (#3088)

* Fix packaged mac sidecar Dock identity

* Rename AMR assistant label

* Fix AMR link catalog model ids

* Fix AMR model normalization typecheck

* Use live AMR model for default runs

* fix: polish AMR runtime settings UI

* Accelerate AMR startup defaults (#3092)

* Surface AMR insufficient balance wallet URL (#3099)

* fix(web): polish onboarding controls (#3112)

* fix(web): show CLI scan loading state

* Avoid duplicate AMR wallet recharge links (#3117)

* Avoid duplicate AMR wallet recharge links

* Use Vela CLI 0.0.3 test package

* chore(nix): refresh pnpm deps hash

* Fix AMR wallet guidance display

---------

Co-authored-by: open-design-bot[bot] <282769551+open-design-bot[bot]@users.noreply.github.com>

* chore(pack): pin Vela CLI 0.0.3-test.1 (#3127)

* chore(nix): refresh pnpm deps hash

* chore(pack): pin Vela CLI 0.0.3

* chore(nix): refresh pnpm deps hash

* fix(web): suppress AMR exit 130 fallback (#3136)

* feat(web): nudge users to hosted AMR on model/auth/quota failures (#3083)

* feat(web): nudge users to hosted AMR on model/auth/quota failures

When a non-AMR agent run fails with an auth / quota / upstream model
error, surface an inline nudge under the error pill linking to Open
Design's hosted AMR gateway (https://open-design.ai/amr). The nudge
fires `surface_view` (element=run_failed_toast) on impression and
`ui_click` (element=go_amr) on the link.

Also teach the daemon to classify CLI-agent auth/quota/upstream failures
(Claude Code, codex, ...) into specific API error codes
(AGENT_AUTH_REQUIRED / RATE_LIMITED / UPSTREAM_UNAVAILABLE) instead of
the generic AGENT_EXECUTION_FAILED, so both the error message and the
nudge key off accurate codes. AMR's own runs are excluded from the
nudge — they keep the dedicated sign-in / recharge affordances.

* feat(web): rework failed-run AMR guidance into per-case error UI

Replace the single inline nudge with a per-case failed-run experience
driven by the run's error code + agent:

- The error card is now neutral gray (was red) and always carries a
  retry button; it is driven by the persisted per-message error event so
  it survives a reload.
- Non-AMR agent hitting a model/auth/quota wall: a theme-color promotion
  card under the error card offers "switch to AMR & retry" — switches the
  run to AMR, opens Settings on the AMR card, and auto-retries once the
  account signs in (ProjectView polls vela login status, independent of
  the Settings pill lifecycle, with success / 5-min-timeout / unmount
  exits).
- AMR agent unauthorized: clearer copy + an "authorize & retry" button.
- AMR agent out of balance: clearer copy + a "top up" button to the AMR
  wallet, with manual retry.
- Settings AMR card: when opened from the nudge, it scrolls into view and
  pulses, and an authorize-button coachmark (a fake hand cursor that
  rises in and dismisses on hover) points at the sign-in control when not
  yet authorized.

analytics: surface_view (run_failed_toast) on the promotion card and
ui_click (go_amr) on its action are retained. i18n adds chat.amrCard.*
and chat.amrError.* (en / zh-CN / zh-TW translated; other locales fall
back to en) and drops the old chat.amrErrorGuidance keys.

* fix(daemon): require status context for numeric service-failure codes

Per review on #3083: the model-service classifier matched bare HTTP
status numbers (`500`, `502`, `429`, `401`), so ordinary CLI output like
`line 500`, `read 502 bytes`, or `exit code 401` could be misclassified
as a provider outage / auth wall and wrongly surface the AMR nudge. Now
a status number only counts when it carries explicit context (`HTTP 500`,
`status 503`, `code: 401`, `502 Bad Gateway`); textual provider phrases
(overloaded, bad gateway, service unavailable, rate limit, …) are
unchanged. Adds fixtures proving unrelated numeric output stays null.

* fix(web): keep error pill for failed runs ChatPane's card doesn't cover

Per review on #3083: the per-message gray error pill was suppressed for
every persisted error status event, but ChatPane only renders the
replacement top-level error card for `retryableAssistantMessage` (the
last failed assistant). So a failed turn that is no longer last (after a
follow-up) or an older failed run in history showed neither the pill nor
the card — its error detail vanished, undercutting reload/history
survival. ChatPane now passes `errorCardOwnerId` (the assistant id whose
error the card represents); AssistantMessage suppresses only that one
pill and keeps rendering StatusPill for all other error events.

* fix(daemon): don't treat a process exit code as an HTTP status

Follow-up to review on #3083: the status-context helper accepted a bare
`code` prefix, so `exit code 401` / `process exited with code 429` still
matched and got classified as AGENT_AUTH_REQUIRED / RATE_LIMITED (the
very `exit code 401` case the comment calls out as noise). `code` now
only counts when qualified (`status code` / `error code` / `response
code`) or punctuation-bound (`code: 401`); bare `exit code N` no longer
matches. Adds fixtures for exit-code lines returning null.

* chore(web): translate AMR card / error keys for 16 remaining locales

PR #3083 added 10 new `chat.amrCard.*` / `chat.amrError.*` keys but only
provided en/zh-CN/zh-TW translations; the other 16 locales fell back to
English. Translate the card title/body, three chips, primary CTA, and
the AMR self-error (auth / balance) messages and buttons for ar, de,
es-ES, fa, fr, hu, id, it, ja, ko, pl, pt-BR, ru, th, tr, uk.

* fix(amr): address review feedback on #2355

Targeted fixes for the unresolved review threads on #2355. Each fix
includes / updates a focused test.

- runtimes/executables.ts: `packagedVelaOpenCodeCompanionTree` now
  verifies the inner `opencode` executable exists + is runnable, not
  just the directory. This closes the false-positive availability path
  that let `detectAgents()` surface AMR as available even when the
  packaged companion was empty / partially copied (mrcfps, 4 threads).

- runtimes/executables.ts: `resolveAmrOpenCodeExecutable` now prefers
  the bundled `<OD_RESOURCE_ROOT>/bin/libexec/opencode/opencode` over a
  stale `opencode` on the user's PATH, so packaged AMR builds can't be
  hijacked by a global installation.

- web/EntryShell.tsx: when the Local CLI scan returns an available
  agent and the previously-selected agent is AMR, switch the selection
  to the first available local agent so the runtime and persisted
  agent agree before Continue.

- server.ts (model-probe branch): for AMR, check `readVelaLoginStatus`
  BEFORE rejecting on an empty live-model catalog — a signed-out user
  was getting `AMR_MODEL_UNAVAILABLE` ("choose a model") instead of
  the correct `AMR_AUTH_REQUIRED` (sign-in affordance).

- server.ts (default model fallback): if the user asked for the AMR
  agent default and the cached id is no longer in the FRESH catalog,
  fall back to `liveModels[0]` from the probe instead of rejecting the
  run as `AMR_MODEL_UNAVAILABLE`.

- integrations/vela.ts: route `vela login` through
  `createCommandInvocation` so an npm/Node-style `vela.cmd` / `.bat`
  shim on Windows gets the correct `cmd.exe /d /s /c …` wrapping with
  verbatim args (matches `execAgentFile` / chat-run spawning).

- tools/pack/src/linux.ts: in containerized Linux builds, bind-mount
  the host directory of `OPEN_DESIGN_VELA_CLI_BIN` and rewrite the env
  to the container-side path. The host path was being passed in as-is
  even though the default container only mounts /project, /tools-pack
  and cache/home — `copyOptionalVelaCliBinary` saw a missing path.

Deferred (out of scope for this PR):
- `od amr status/login/logout/cancel` CLI subcommands (AGENTS.md
  UI/CLI dual-track rule, server.ts:5763) — sizable surface; tracked
  for a separate focused PR.
- Strict `--require-vela-cli` for Windows + mac-x64 beta builds:
  prematurely blocked — `@powerformer/vela-cli` only publishes the
  `darwin-arm64` platform binary today; adding the flag elsewhere
  would fail the builds. Revisit once win/x64/linux binaries ship.

* fix(amr): hoist sendAmrAccountFailure above the AMR catalog preflight (TDZ)

The new signed-out AMR branch in the catalog preflight at server.ts:10875
calls `sendAmrAccountFailure(...)` to emit AMR_AUTH_REQUIRED, but the
const declaration sat ~100 lines below at the outer function scope. Because
`const` is TDZ-aware, that branch would have thrown `ReferenceError:
Cannot access 'sendAmrAccountFailure' before initialization` for the
exact users it tries to help — defeating the original intent.

Hoist the helper to just above the AMR preflight block so it's available
to every AMR code path in this function. Behavior elsewhere is unchanged.

Also rerun the daemon test suite: `launch.test.ts > resolveAgentLaunch
uses packaged built-in Vela for AMR` was creating the
`<resourceRoot>/bin/libexec/opencode/` companion *directory* only, but
this PR's earlier tightening of `packagedVelaOpenCodeCompanionTree`
also requires the inner `opencode` executable. Add it to that fixture
to match the new contract; the test was a sibling of the executables /
env-and-detection fixtures already updated in 13fc4f4.

Addresses #2355 review (mrcfps, 2026-05-28).

* feat(web): add hover cancel for AMR login (#3158)

* feat(web): add hover cancel for AMR login

* fix(web): don't bounce AmrLoginPill back to 'Signing in…' after local cancel

Both codex-connector (P2) and looper (CHANGES_REQUESTED) on this PR
flagged the same race in the new local-cancel path: `handleCancelLogin`
dispatches `notifyAmrLoginStatusChanged('login-canceled')` immediately
after `/login/cancel` returns, but the `AMR_LOGIN_STATUS_EVENT` listener
unconditionally re-enters `refresh()` and then restarts polling
whenever `/api/integrations/vela/status` still reports
`loginInFlight: true`.

That is a real race because the daemon's `cancelVelaLogin()` only sends
SIGTERM (escalating to SIGKILL after `LOGIN_CANCEL_KILL_GRACE_MS` =
2000 ms) and keeps the child in `activeLoginProcs` until it actually
exits — so the first `/status` read after a successful cancel can
legally still come back as in-flight. Under that window the pill flips
back to 'Signing in…' and can later surface the timeout/error path even
though the user already canceled, defeating the behavior promised in
the PR description.

Fix the listener instead of every dispatch site: in the
`login-canceled` branch, after the local reset (stopPolling +
setPending(null) + clear refs), optimistically mark every subscribed
pill instance as not-in-flight (`setStatus((c) => c ? { ...c,
loginInFlight: false } : c)`) and `return` — skip the
refresh-and-reconcile branch below entirely. The next explicit refresh
(component mount, user interaction, or a `status-changed` event) will
pick up the daemon's confirmed state once the child has actually
exited.

Add a focused regression test that holds `/api/integrations/vela/status`
at `loginInFlight: true` even after a successful `/login/cancel`,
asserting that the pill stays at the Canceled → Authorize sequence and
never bounces back to 'Signing in…'. This test fails on the pre-fix
listener and passes on the new behavior; existing
'cancels an in-flight AMR sign-in…' and 'reconciles late AMR browser
completion to Signed in after local cancel' tests continue to pass.

Addresses review feedback on #3158 (chatgpt-codex-connector, nettee).

---------

Co-authored-by: lefarcen <935902669@qq.com>

---------

Co-authored-by: a1chzt <chizblank@gmail.com>
Co-authored-by: Amy <1184569493@qq.com>
Co-authored-by: Mason <jinmeihong0201@gmail.com>
Co-authored-by: Caprika <56862773+alchemistklk@users.noreply.github.com>
Co-authored-by: open-design-bot[bot] <282769551+open-design-bot[bot]@users.noreply.github.com>
2026-05-28 05:09:55 +00:00
lefarcen
c14baf07d3 Merge origin/main into release/v0.8.0
PR #2461 sync prep — resolves 14 conflicts merging 84 main-side commits
on top of 58 release-side commits accumulated during the 0.8.0 cycle.

Resolution summary:

Take main (theirs) where main carried deliberate forward progress:
- apps/web/src/components/PluginCard.tsx — 7 hunks, i18n migration:
  hardcoded English aria-labels/titles replaced with t() calls keyed
  on pluginCard.* (all 8 keys verified present in en.ts).
- apps/web/src/components/TasksView.tsx — 1 hunk, source-ingestion
  feature: sortedRoutines (newest-first), sourceIngestionTemplates,
  patchSourceForm, submitSourceIngestion. activeCount/pausedCount
  semantics preserved (now keyed on sortedRoutines, count unchanged).
- e2e/ui/app.test.ts — new node:fs/promises + tmpdir + path + @/timeouts
  imports needed by main-side test helpers.
- e2e/ui/settings-local-cli-codex-fallback.test.ts — menu-dismissal
  helper block added by main.

Keep both sides where each added a different field to the same object
literal:
- apps/web/src/components/ProjectView.tsx (locale + analyticsHints
  spread).
- apps/web/src/components/DesignSystemFlow.tsx (locale + analyticsHints).

Take release (ours) where release carried deliberate work that ships
0.8.0:
- CHANGELOG.md — release-side 0.8.0 entry + PR link refs; main's
  Unreleased section was the same body of work, now finalized.
- apps/landing-page/public/{apple-touch-icon,favicon}.png +
  apps/web/public/app-icon.svg — release-side visual refresh assets
  consistent with 0.8.0 stable ship.
- tools/pack/src/linux.ts — packageVersion const required by line 466;
  taking main's empty line would build-error.
- e2e/ui/project-management-flows.test.ts +
  e2e/ui/settings-api-protocol.test.ts +
  e2e/ui/settings-memory-routines.test.ts — release-side release-smoke
  hardening (shangxinyu1 + PerishFire) takes precedence on overlap.

Closes-issue / unblocks: PR #2461 sync release/v0.8.0 → main.
2026-05-23 12:17:18 +08:00
Marc Chan
a3872b97a9
fix(tools-dev): preserve web origin trust on web start (#2715)
* fix(tools-dev): preserve web origin trust on web start

Restart daemon/web when the trusted web port is missing, and reuse the active web port during repeated starts so run web and start web keep app-config origin checks aligned.

Generated-By: looper 0.0.0-dev (runner=worker, agent=opencode)

* fix(plugins): refresh official registry bundled count

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(tools-dev): preserve daemon/web reserved ports

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(tools-dev): preserve daemon reuse on web start

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(tools-dev): preserve running daemon port on web reuse

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(tools-dev): reserve explicit web port before daemon allocation

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* test(web): stabilize media provider reload flash timing

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(web): restore merged reattach workspace coverage

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* fix(tools-dev): reserve allocated daemon port

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)

* test(e2e): wait for artifact manifest persistence

Generated-By: looper 0.0.0-dev (runner=fixer, agent=opencode)
2026-05-23 00:25:43 +08:00
lefarcen
aedbb9dbe4 release: Open Design 0.8.0
Bumps 14 workspace package.json files from 0.7.0 to 0.8.0:
- root, apps/{web,daemon,desktop,landing-page}
- packages/{contracts,host,platform,sidecar,sidecar-proto}
- tools/{dev,pack,pr}, e2e

apps/packaged was already at 0.8.0 from the preview lane.
Independently versioned packages keep their own tracks.

Adds CHANGELOG [0.8.0] - 2026-05-20 entry covering the
305 PRs merged since 0.7.0 by 75 contributors:

- Plugin engine rebuild + Plugin Registry surface
- Headless by default (desktop is thin wrapper around CLI)
- Critique Theater Phases 9 through 16
- 149 design systems with structured tokens.css
- Italian locale + CJK font fallback
- Leonardo.ai, ElevenLabs, SenseAudio providers
- Windows packaged auto-update
- Visual refresh + Quick-brief discovery overhaul
- PostHog v2 analytics
- Manual edit UX overhaul
2026-05-20 21:22:17 +08:00
PerishFire
899c9fe4d8
Support nightly and preview package identity (#2437) 2026-05-20 19:46:39 +08:00
PerishFire
ad37fd30cf
Add desktop updater UI flow (#2270) 2026-05-19 21:36:51 +08:00
PerishFire
bd48c597b0
chore: pin dependency versions and harden CI caches (#2189)
* chore: pin dependency versions

* ci: enforce pinned dependency specs

* ci: fix pnpm executable invocation
2026-05-19 13:58:27 +08:00
PerishFire
4424f08be0
[codex] Add packaged desktop auto-update (#1375)
* Add packaged desktop auto-update

* Handle counted beta nightly update versions

* Refresh desktop auto-update branch for main

* Serialize desktop updater operations

* Refresh auto-update branch for packaged paths
2026-05-19 11:20:05 +08:00
lefarcen
2a0ebea50b release: Open Design 0.7.0
- bump 14 monorepo package.json files to 0.7.0 (root + apps/{web,daemon,desktop,packaged,landing-page} + packages/{contracts,platform,sidecar,sidecar-proto} + tools/{dev,pack,pr} + e2e); apps/packaged was already at 0.6.1 from beta lane, all others at 0.6.0
- add CHANGELOG.md [0.7.0] - 2026-05-12 entry covering 97 merged PRs since 0.6.0:
  - Critique Theater: Phase 7 web client state machine (#1307) + Phase 6.2 daemon artifact extraction (#1085)
  - Web/UI: thumbs-up/down feedback widget (#1308), Cmd+, opens Settings (#1173), Finalize design package + Continue in CLI (#974), fetch models button for BYOK (#1034), provider models alphabetical sort (#1097), collapsible MCP JSON field-mapping (#1136), design file rename (#894)
  - Daemon: auto-memory store with chat-protocol-aware extraction (#999), install/uninstall skills & design systems (#1003), HTTP 206 range requests for video/audio (#1105), scheduled routines (#1033), agent runtime + route registration refactor (#1063, #1043)
  - HyperFrames: HTML-in-Canvas across web + skills (#866)
  - Skills/design systems: generic skills + design-templates split + finalize-design API (#955), agent-browser skill (#1284), WeChat design system + login-flow skill (#1083), hud/loom/trading-terminal design systems (#1069), release-notes-one-pager skill (#873), tokens.css schema (#1231)
  - Packaging: macOS Intel (x64) build (#759), official Nix flake (#402), beta packaging cache (#1095)
  - Maintainer ops: tools-pr PR-duty workspace (#1259), MAINTAINERS.md (#1290), contributor card bot (#932), PR→issue linking discipline (#1263)
  - Changed: conversation run isolation (#1271), default English i18n fallback (#1270), Codex CLI exit diagnostics / empty-response handling / path fallback (#1267, #1244, #1205)
  - Fixed: ~30 web + desktop + daemon + packaging bugfixes
  - Internal: nightly UI/desktop regression coverage (#1256), e2e/release report hardening (#1140), entry/settings automation (#954)
- catch up [Unreleased] compare link to v0.7.0 and add missing [0.6.0] release link
- add 97 PR footnote refs ([#402]..[#1330])

Verified locally: pnpm install + pre-build contracts/daemon/desktop dist + pnpm typecheck (exit 0 across all 14 packages on Node 22.22 with engine-warning).

Release workflow validation runs after merge via release-stable.
2026-05-12 15:33:28 +08:00
Bryan A
587c783dc0
feat(web): add Finalize design package + Continue in CLI buttons (#451) (#974)
* feat(daemon): expose resolvedDir on GET /api/projects/:id (#451 prereq)

Native projects (no metadata.baseDir) live at <projects root>/<id>, where
projects root is daemon-side state. The web client cannot reconstruct an
absolute path on its own, and shell.openPath on a relative path is
undefined behavior. Without resolvedDir, the upcoming Continue in CLI
button (#451) would render permanently disabled for native projects.

Mirrors PR #832's pattern of exposing designMdPath in its response.
Computed via the existing resolveProjectDir(...) helper. No behavior
change to existing callers; they ignore the new field.

Adds ProjectDetailResponse contract type and a focused projects-routes
test covering imported-folder, native, and unknown-id paths.

* feat(web): add parseProvenance helper for DESIGN.md staleness checks

Pure helper that extracts Project ID, design system, current artifact,
transcript message count, and generated UTC timestamp from the
`## Provenance` section emitted by the daemon's finalize synthesis
prompt (apps/daemon/src/finalize-design.ts). Used by useDesignMdState
to derive the Continue in CLI button's stale/fresh state without an
additional daemon endpoint.

Handles missing section, "none" sentinels for design system /
artifact, and malformed timestamps without throwing. Tests cover all
four branches.

* feat(web): add buildClipboardPrompt template for Continue in CLI

Inline single-source-of-truth template per #451 spec §3.4. Names the
project, the working directory, and the DESIGN.md-first operating
contract for the receiving `claude` CLI session. Trailing TODO is
the blank task slot the issue body specifies — left empty so the user
fills it in before submitting.

Also lands the shared copyToClipboard helper (jsdom-safe canonical path
+ execCommand fallback) so the new button and any future caller share
one fallback path, mirroring the inline pattern in FileViewer.tsx.

Tests cover happy-path field rendering, "none"/"unknown" sentinels
when DESIGN.md fields are absent, and both clipboard branches.

* feat(web): add useProjectDetail + useDesignMdState hooks

useProjectDetail wraps GET /api/projects/:id, surfacing the resolvedDir
field and falling back to metadata.baseDir for older daemons that don't
include it. Continue in CLI needs an absolute working directory so the
desktop bridge can openPath it; the web client never reconstructs the
path itself.

useDesignMdState fetches the project's file list, downloads DESIGN.md
when present, parses the Provenance section, and computes a stale
verdict by comparing the recorded generatedAt against the max mtime of
non-DESIGN.md files and the max conversation updatedAt. Drives the
button's three-state UI (disabled / fresh / stale) without a
daemon-side endpoint.

Tests cover happy path, fallback, and both stale branches plus the
pure computeStale helper for the null-timestamp edge case.

* feat(web): add useFinalizeProject hook with cancel + error-code mapping

Wraps POST /api/projects/:id/finalize/anthropic for the Finalize design
package button. Three concerns:

  1. Lifecycle: idle → pending → success | error. Double-clicking the
     button aborts the prior in-flight request before starting a new
     one so the daemon never sees stacked finalize calls per project.

  2. Cancellation: AbortController plumbed through fetch + a 130 s
     timer (daemon timeout 120 s + 10 s buffer). Cancel returns to idle
     cleanly — it's a user gesture, not an error surface.

  3. Daemon error mapping: when the response is non-OK, body.error.code
     drives the canonical user-facing toast string (table covers all
     7 codes the daemon emits today plus a network-error catch-all).
     body.error.details, when a string, surfaces alongside the category
     message so account-usage-cap responses (Anthropic 400 →
     UPSTREAM_UNAVAILABLE) can show the upstream's own reason instead
     of just the daemon's category label — committed to lefarcen on
     #450 verification reply.

Tests cover request body shape, all 8 error codes via it.each, the
network-error path, the details-surfacing branch, the cancel ⇒ idle
flow, and the unknown-code → catch-all message branch.

* feat(web): add useTerminalLaunch with electron/web detection

Capability-detected wrapper around window.electronAPI.openPath. On
desktop the bridge forwards to shell.openPath, which opens the OS
file manager at the project working directory (per Electron's
contract for directory paths — it is NOT a terminal launcher;
spawning a terminal application is deferred per #451 Non-goals). On
browser builds the hook reports web-fallback so the caller renders
a manual-instruction toast naming the working directory.

Treats any non-empty string return from shell.openPath as ok: false
so platform-specific failures surface the manual fallback toast.
Behavior is exercised end-to-end by the upcoming
ContinueInCliButton tests.

* feat(desktop): expose shell.openPath via electronAPI bridge

Adds an openPath bridge method that the Continue in CLI button (#451)
uses to surface the project working directory in the OS file manager.
shell.openPath is part of Electron's contract and resolves to '' on
success / a non-empty error string on failure; the IPC handler
forwards the result so the renderer can decide between the success
toast and the manual fallback toast without a separate error channel.

Empty / non-string inputs short-circuit to a self-describing error
string so the renderer never needs to worry about undefined-input
crashes from the main process.

Web side: extracts Window.electronAPI into a single global declaration
at apps/web/src/types/electron.d.ts so future bridge methods land in
one place. Two pre-existing inline declare-global blocks
(NewProjectPanel.tsx, providers/registry.ts) are deleted in favor of
that single source of truth — the inline ones each carried a partial
shape of the bridge and were diverging from the desktop preload.

* feat(web): add FinalizeDesignButton, ContinueInCliButton, ProjectActionsToolbar

Project-level toolbar that hosts the two new actions from #451.
Mounted between AppChromeHeader and the chat/workspace split (wiring
lands in the next commit). Per-file actions (Export PDF/PPTX/ZIP,
Deploy) stay in the FileViewer share menu.

FinalizeDesignButton has three idle labels driven by DESIGN.md
existence + staleness, plus a pending state with a spinner and a
cancel link that maps to useFinalizeProject's AbortController. Error
toasts are owned by ProjectView so the button doesn't carry its own
toast surface.

ContinueInCliButton renders disabled with a Finalize-pointing
tooltip when DESIGN.md is missing (so the workflow is discoverable
rather than hidden), enabled when fresh, and enabled with a stale
chip otherwise. Chip text is the spec's canonical "Spec is stale —
regenerate?" — N-turns-ago is deferred per spec §4.6.

Toast.tsx is a tiny transient component that mirrors
PromptTemplatePreviewModal's state-based toast pattern; supports a
secondary details line so daemon error envelopes that carry an
upstream explanation (e.g. Anthropic account-usage cap) can surface
the real reason alongside the daemon's category label.

CSS appends one block to apps/web/src/index.css mirroring the
existing app-project-title token usage; no CSS modules in this
repo (verified by grep).

* test(web): cover ContinueInCliButton states + interaction wiring

Three rendered states (DESIGN.md missing → disabled with the
Finalize-pointing tooltip; DESIGN.md fresh → enabled, no chip;
DESIGN.md stale → enabled with the canonical "Spec is stale —
regenerate?" chip), plus three onClick branches (no-op when
disabled, fires once when fresh, fires once when stale).

Click-handler integration with clipboard / shell.openPath / toast
lives in ProjectView (the button is presentational and takes the
handler in via props), so those are covered by Phase K's wiring +
the manual smoke test rather than the per-component test.

* feat(web): wire Continue in CLI + Finalize buttons into ProjectView

Mounts the new project-actions toolbar between AppChromeHeader and
the chat/workspace split, hidden when workspaceFocused so the
focus-mode artifact view stays uncluttered.

Wires the four hooks (useProjectDetail, useDesignMdState,
useFinalizeProject, useTerminalLaunch) to a single shared toast
surface. handleFinalize reads the request body from the existing
config: AppConfig prop and uses effectiveMaxTokens(config) to match
the chat-flow's maxTokens defaulting; on success it refreshes
useDesignMdState so the toolbar re-renders with the new chip state.

handleContinueInCli builds the literal clipboard prompt, copies it,
opens the working directory via shell.openPath on desktop /
falls through to a manual-instruction toast on browser, and surfaces
shell.openPath failures with a fallback toast that names the path.

Errors lift into the same toast surface (a useEffect tied to
finalize.error) so the daemon's category message + body.error.details
reach the user as the spec's two-line render — covered by hook test
16a in the prior commit.

⌘+Shift+K (mac) / Ctrl+Shift+K (others) is the keyboard
accelerator for Continue in CLI; capture-phase, platform-gated,
no-op when DESIGN.md is missing. Mirrors the existing FileWorkspace
shortcut idiom and does not collide with ⌘+P (Quick Switcher).

* fix(web): distinguish timeout abort from user cancel in useFinalizeProject

Addresses codex P2 finding on PR #974: the catch block treated every
AbortError as a user-initiated cancel and reset to idle silently. If
the internal 130 s timeout fired, users saw no failure signal but the
daemon's synthesis call may still have been in flight.

Adds a timedOutRef set inside the setTimeout callback before
controller.abort(), and branches in the catch: timeout → status
'error' with new TIMEOUT code ("Finalize timed out after 130 s. The
daemon may still be running."), user cancel → existing idle reset.
Reset the ref at the start of every trigger() so a previous timeout
doesn't poison the next call.

Adds one test using vi.useFakeTimers() that advances past 130_001 ms
and asserts the TIMEOUT error surface.

* fix(web): surface clipboard failures by rendering the prompt in the toast

Addresses codex P2 finding on PR #974: handleContinueInCli ignored
copyToClipboard's return value, so when both clipboard paths failed
(restricted browser context / insecure origin) the toast still said
"paste the prompt" though nothing had been copied — leaving users
with no manual-copy recourse in exactly the environments where the
fallback should help.

handleContinueInCli now branches on copyToClipboard's boolean return.
On failure the toast renders the prepared prompt in a scrollable
<pre> block and pins itself open (no auto-dismiss) so the user has
time to select-and-copy manually. Includes a Dismiss button + the
working directory in the secondary details line so the user has the
information needed to proceed.

The folder-open call is skipped on copy failure because there's
nothing to paste yet; the user copies first, then re-clicks Continue
in CLI when they're ready.

Toast component grows an optional Updating VS Code Server to version 41dd792b5e652393e7787322889ed5fdc58bd75b
Removing previous installation...
Installing VS Code Server for Linux x64 (41dd792b5e652393e7787322889ed5fdc58bd75b)
Downloading:       0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  0%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  1%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  2%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  3%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  4%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  5%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  6%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  7%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  8%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9%  9% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 11% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 12% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 13% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 14% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 15% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 16% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 17% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 18% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 19% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 21% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 22% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 23% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 24% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 25% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 26% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 27% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 28% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 29% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 30% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 31% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 32% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 33% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 34% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 35% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 36% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 37% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 38% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 39% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 40% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 41% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 42% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 43% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 44% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 45% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 46% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 47% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 48% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 49% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 50% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 51% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 52% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 53% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 54% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 55% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 56% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 57% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 58% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 59% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 61% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 62% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 63% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 64% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 65% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 66% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 67% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 68% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 69% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 70% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 71% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 72% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 73% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 74% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 75% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 76% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 77% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 78% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 79% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 80% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 81% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 82% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 83% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 84% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 86% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 87% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 88% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 89% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 90% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 91% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 92% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 93% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 94% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 95% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 96% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 97% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 98% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99% 99%100%100%
Unpacking:   0%  1%  2%  3%  4%  5%  6%  7%  8%  9% 10% 11% 12% 13% 14% 15% 16% 17% 18% 19% 20% 21% 22% 23% 24% 25% 26% 27% 28% 29% 30% 31% 32% 33% 34% 35% 36% 37% 38% 39% 40% 41% 42% 43% 44% 45% 46% 47% 48% 49% 50% 51% 52% 53% 54% 55% 56% 57% 58% 59% 60% 61% 62% 63% 64% 65% 66% 67% 68% 69% 70% 71% 72% 73% 74% 75% 76% 77% 78% 79% 80% 81% 82% 83% 84% 85% 86% 87% 88% 89% 90% 91% 92% 93% 94% 95% 96% 97% 98% 99%100%
Unpacked 4009 files and folders to /home/bryan/.vscode-server/bin/41dd792b5e652393e7787322889ed5fdc58bd75b.
Looking for compatibility check script at /home/bryan/.vscode-server/bin/41dd792b5e652393e7787322889ed5fdc58bd75b/bin/helpers/check-requirements.sh
Running compatibility check script
Compatibility check successful (0) prop and the auto-dismiss
TTL is suppressed whenever code is present. CSS adds .od-toast-code
(monospace, max-height 240 with overflow-auto) and .od-toast-dismiss
styling.

Six new Toast tests cover details rendering, code rendering,
no-auto-dismiss when code is present, auto-dismiss when code is
absent, and the Dismiss button affordance.

* fix(web): make ContinueInCliButton disabled-state guidance visible

Addresses mrcfps's PR #974 review: native <button disabled> does
not fire hover/focus events in browsers we ship against, so a
`title` tooltip on the disabled button never surfaces. The only
guidance for the missing-DESIGN.md state was effectively invisible —
defeating the spec's "discoverable, not hidden" intent.

Renders the help text as a visible sibling <span> next to the
disabled button instead. Adds aria-describedby pointing the button
at the hint's id so assistive tech announces the explanation when
the disabled button gets focus. The native `disabled` attribute
stays so the button still can't be clicked or submitted.

CSS adds .project-actions-disabled-hint (muted italic, 11.5px,
matches the existing meta/secondary text style on this surface).

Test asserts the role="note" hint is in the DOM with the canonical
text and that the button's aria-describedby links to its id.

* fix(web): keep ProjectActionsToolbar at natural height inside the .app grid

The .app container was `grid-template-rows: auto 1fr` — only two
rows. Adding ProjectActionsToolbar as a third child between
AppChromeHeader and the chat/workspace split made the toolbar the
2nd grid item, so it took the `1fr` row (filling roughly half the
viewport) while the split got pushed into an implicit auto row at
its content's natural height. Surfaced as a screenshot from Bryan
showing the toolbar's background bleeding across most of the screen.

Extend grid-template-rows to `auto auto 1fr` and pin the split to
`grid-row: 3` explicitly. Now:
- Toolbar visible: row 1 = header (auto), row 2 = toolbar (auto),
  row 3 = split (1fr, fills remaining viewport).
- Toolbar hidden via hidden=workspaceFocused → ProjectActionsToolbar
  returns null, row 2 collapses to 0px (auto with no content), split
  still fills row 3.

No JS changes; existing 609 tests still green.

* fix(web): guard useFinalizeProject state writes against superseded triggers

Addresses mrcfps's PR #974 P1 review on useFinalizeProject.ts:132
(also called out as P1.3 in lefarcen's deep-dive review).

Calling trigger() twice in quick succession aborted the first
controller and swapped abortRef to the new one, but the first
request's later AbortError catch still unconditionally called
setStatus('idle') / setError(null). That cleared the spinner and
re-enabled both toolbar buttons while the replacement finalize was
still pending — defeating the de-duplication this hook was meant to
enforce.

Adds an isCurrent() closure (`abortRef.current === controller`)
and gates every state-write site after the await: success path,
non-OK envelope path, AbortError-timeout, AbortError-cancel, and
network-error all bail early when the trigger has been superseded.
Per mrcfps: "make every state write request-scoped."

Regression test triggers twice in quick succession with a
never-resolving fetch, awaits the first promise (it rejects with
AbortError), and asserts status stays 'pending' rather than
collapsing to 'idle' under the replacement's lifetime.

* fix(desktop): allowlist-validate shell.openPath against registered project roots

Addresses mrcfps's PR #974 P1 review on runtime.ts:305 (also called
out as P1.2 in lefarcen's deep-dive review): the new
`shell:open-path` IPC handler accepted any renderer-supplied
string and forwarded it straight into Electron's `shell.openPath`,
widening the renderer→main trust boundary so XSS or a compromised
renderer dependency could open arbitrary local paths to the user.

Adds an explicit gate around the bridge:

  1. validateExistingDirectory(p) — floor check that rejects empty
     strings, relative paths, files, apps, and non-existent paths;
     realpath-resolves so symlink games can't be used to register
     one path and reach another.

  2. createProjectRootGate() — Set-backed allowlist of
     daemon-validated project working directories. The renderer
     calls registerProjectRoot(absDir) once per project mount via
     a new IPC method (preload bridge); the main process only
     opens paths that pass both the floor check and the allowlist.

ProjectView wires the registration via a useEffect tied to
projectDetail.resolvedDir, so the active project's daemon-supplied
working directory is always the one being approved (not a renderer-
synthesized string).

Threat-model caveat documented in the runtime.ts comment block: an
attacker that fully controls the renderer can also call register
with arbitrary paths. Closing that gap fully requires a daemon-side
round-trip to derive the canonical resolvedDir from the daemon's
project registry, which is deferred to keep this PR focused.
Today's allowlist still defends against accidental misuse, bugs,
and common XSS payloads that don't know to call register first.

Adds apps/packaged/tests/desktop-project-root-gate.test.ts with 13
cases: floor-validation rejection cases (empty / relative / missing
/ file), happy-path resolution, symlink realpath canonicalization,
and the allowlist's register/isApproved/reset semantics. Mirrors
the existing apps/packaged/tests/desktop-url-allowlist.test.ts
pattern from PR #911 — the packaged workspace hosts the test
because apps/desktop has no vitest setup yet.

* fix(daemon): wire request-lifecycle abort signal through finalize route

Addresses mrcfps's PR #974 P1 review on
apps/daemon/src/server.ts:3831-3837 (also called out as P1.1 in
lefarcen's deep-dive review): `POST /api/projects/:id/finalize/anthropic`
called `finalizeDesignPackage(...)` without threading any
request-lifecycle abort, so cancelling the browser fetch only
aborted the UI-side request — the daemon's 60–120 s Anthropic call
kept running and still wrote DESIGN.md after the UI returned to idle.

Adds an AbortController inside the route handler, fired from
`res.on('close')`, and threads its signal into the existing
`signal?: AbortSignal` parameter on `FinalizeOptions`
(finalize-design.ts:70). `callAnthropicWithRetry` already passes
the signal through to the underlying fetch, so a client disconnect
now propagates all the way to the Anthropic SDK call.

Listener-event choice: `res.on('close')` is the canonical event
for "client disconnected before response was sent" in Express. The
common alternative `req.on('close')` fires whenever the *request*
stream finishes — for POST routes that means as soon as the
body-parser middleware drains the body, well before the route does
any work. Using req.on('close') would have flipped the abort
controller in every successful run; the test caught this empirically.

Caveat documented in the route's comment block: an abort fired
*after* the upstream response has been received but *before* the
atomic write completes still allows the write to land. The SDK
contract bounds the network round-trip, not the post-network disk
handoff.

Adds tests/finalize-route-abort.test.ts: spins up the test server,
mocks global fetch to capture the daemon-side AbortSignal at the
Anthropic call, sends the request via raw http (so we can destroy
the underlying socket), waits until the server reaches the
Anthropic call, then destroys the socket and asserts that the
daemon-side signal received an abort event within 5 s.

Three pre-existing project-watchers chokidar tests show flaky
timeouts under full-suite concurrency but pass in isolation;
unrelated to this fix.

* fix(daemon): refactor finalize-route-abort test to satisfy strict TS narrowing

The CI typecheck (`pnpm --filter @open-design/daemon typecheck`,
which runs both tsconfig.json and tsconfig.tests.json) caught what
my pre-push validation missed: TS narrowed `capturedSignal` to
literal `null` because vitest's mockImplementation closure can't
prove its callback runs, leaving the bare `let capturedSignal:
AbortSignal | null = null` permanently typed at its initial value.
At line 184 (`expect(capturedSignal?.aborted).toBe(true)`) the
right-hand side of the optional-chain became unreachable, and TS
flagged it as `Property 'aborted' does not exist on type 'never'`.

Switches to the standard ref-object pattern
(`const capture: { signal: AbortSignal | null } = { signal: null }`).
TS narrows let bindings inside closures conservatively but treats
object-property writes as opaque, so `capture.signal` reads
correctly across the closure boundary. Logic is unchanged.

(Pre-push oversight: ran `pnpm --filter @open-design/web typecheck`
but not the full repo `pnpm typecheck` after the daemon test
landed; the daemon's own typecheck would have caught this. Adding
`pnpm typecheck` back into the standard pre-push checklist.)

* fix(desktop): make shell.openPath gate daemon-controlled and reject .app bundles

Addresses lefarcen + mrcfps PR #974 P1 reviews on the previous path
allowlist (commit 8bf56597):

  - mrcfps (runtime.ts:45): `validateExistingDirectory` accepted
    macOS `.app` bundles because they're directories, so the gate
    would forward `/Applications/Safari.app` (or any other app
    bundle) into shell.openPath and *launch* the application — a
    stronger capability than the bridge's intended "reveal the
    project folder" feature.

  - lefarcen (runtime.ts:396): the allowlist was renderer-controlled.
    A compromised renderer could call `shell:register-project-root`
    with any existing absolute directory and then `shell:open-path`
    that same path; the IPC injection issue I'd documented as
    "deferred" was the central reviewer concern, not an acceptable
    caveat. Both reviewers asked for the gate to be derived from
    a daemon-authoritative source.

The redesign drops the renderer-controlled register/openPath pair
and replaces it with a single `openPath(projectId)` bridge call.
The desktop main process resolves the project ID by calling the
daemon's `GET /api/projects/:id` endpoint over the web sidecar
proxy (which already forwards `/api/*` to the daemon — verified
in apps/web/sidecar/server.ts:209 and apps/web/next.config.ts:77),
parses `resolvedDir` from the response, validates it against the
floor (absolute, exists, is-directory, not .app), and only then
forwards to `shell.openPath`. The renderer never names the path
directly, so a compromised renderer cannot escalate to opening
arbitrary local paths — it can only name a project the daemon
already knows about, and the canonical path comes from the daemon's
own response.

Surface changes:

  - `runtime.ts`: `createProjectRootGate` removed.
    `fetchResolvedProjectDir(webUrl, projectId, fetchImpl?)` added.
    `validateExistingDirectory` rejects `.app` suffix after the
    realpath check (so symlinked launders are caught too).
    `shell:open-path` handler signature changes from `(path)` to
    `(projectId)`; `shell:register-project-root` handler removed.

  - `preload.cts`: `openPath(projectId)`; `registerProjectRoot`
    removed from the bridge surface.

  - `apps/web/src/types/electron.d.ts`: type updated to match.

  - `useTerminalLaunch.ts`: `open(projectId)` instead of
    `open(dir)`.

  - `ProjectView.tsx`: passes `project.id` to
    `terminalLauncher.open`; the registerProjectRoot useEffect is
    deleted. Toast text still reads `projectDir` (from
    `useProjectDetail.resolvedDir`) for fallback messages — the
    *display* path is independent of the *open* mechanism.

  - `apps/packaged/tests/desktop-project-root-gate.test.ts`:
    rewritten to cover `validateExistingDirectory` (8 cases
    including the new `.app` suffix and symlinked-bundle rejection)
    and `fetchResolvedProjectDir` (8 cases including empty/invalid
    project ids, daemon HTTP success/failure, missing resolvedDir,
    network error, and URL canonicalization).

Total: 16 passing tests, ~330 LOC churn including test rewrites.

Lesson learned (from the iteration loop, not the code): when a
reviewer asks for "ideally X, or at least Y," shipping Y with a
deferred-X note flags the gap rather than fixing it. Either ship X
or argue Y is sufficient; don't middle-ground.

* feat(contracts,sidecar-proto): add desktop-auth IPC + fromTrustedPicker

Schema-only prep for the PR #974 round-3 fix. Adds the two type
extensions the daemon HTTP gate and the desktop main process will
build on:

- packages/sidecar-proto: SIDECAR_MESSAGES.REGISTER_DESKTOP_AUTH, with a
  base64-validated `{ secret }` payload + RegisterDesktopAuthResult.
  Updates normalizeDaemonSidecarMessage to accept the new message and
  pins both branches (accept + reject) in tests/index.test.ts.

- packages/contracts: ProjectMetadata.fromTrustedPicker — a marker the
  daemon stamps on folder-imported projects whose POST /api/import/folder
  passed the desktop HMAC gate. The marker is privileged in the same
  way as `baseDir`: only the gated import handler sets it, and the
  desktop main process refuses to forward `shell.openPath` for
  folder-imported projects whose metadata lacks it.

* fix(daemon): gate /api/import/folder on desktop HMAC token

Closes the renderer→arbitrary-baseDir→shell.openPath bypass chain
flagged by lefarcen and mrcfps in round 3 of PR #974. Both reviewers
converged on the same gap: the previous round only moved path
resolution into the daemon, but renderer JS could still POST
/api/import/folder with any absolute path, get a project ID back, and
then call openPath(projectId) to reveal the attacker-chosen path.

Daemon-side closure:

- New module-scope desktop auth secret + setter exported from
  apps/daemon/src/server.ts. The secret is null at boot (web/standalone
  mode unaffected) and gets set when the desktop main process
  registers it over the daemon's sidecar IPC.

- New `verifyDesktopImportToken` pure helper. Verifies tokens shaped
  `${nonce}~${exp}~${signature}` against HMAC-SHA256(secret, baseDir +
  "\n" + nonce + "\n" + exp). Field separator is `~` (not `.`) because
  ISO 8601 expiries embed dots; `~` is in neither base64url nor ISO
  8601 character sets. Rejects expired tokens, replayed nonces, and
  expiries beyond 2× the 60s TTL.

- New middleware on POST /api/import/folder. When the secret is set,
  every request must carry a valid `X-OD-Desktop-Import-Token` header
  bound to the requested baseDir. Rejected requests return 403 with
  FORBIDDEN. When the secret is unset (no desktop registered), the
  route is unchanged so web-only deployments and standalone daemons
  keep working.

- Trusted imports get `metadata.fromTrustedPicker: true` stamped on
  the project. POST /api/projects and PATCH /api/projects/:id reject
  any client-supplied `fromTrustedPicker` (privileged the same way as
  `baseDir`), and the PATCH preservation block re-stamps the marker
  on partial-metadata patches so it cannot be silently stripped.

- Daemon sidecar IPC handler: REGISTER_DESKTOP_AUTH calls
  setDesktopAuthSecret with the base64-decoded secret. The HTTP and
  IPC servers share a process so the registration takes effect
  immediately for the next inbound /api/import/folder call.

Tests:

- apps/daemon/tests/desktop-import-token-gate.test.ts (15 cases): web
  mode acceptance, no-token rejection, malformed-token rejection,
  wrong-secret rejection, wrong-baseDir rejection, expired rejection,
  oversized-window rejection, valid mint + trusted-picker stamp +
  replay rejection, plus 6 pure-helper cases for verifyDesktopImportToken.
  afterAll() clears the secret to keep the shared HTTP server clean
  for sibling test files.

- apps/daemon/tests/projects-routes.test.ts (+2 cases): POST and PATCH
  reject `fromTrustedPicker` in client-supplied metadata.

Existing folder-import-route.test.ts continues to pass because none of
those tests register a desktop secret; the gate stays dormant.

* fix(desktop,web): atomic pickAndImport replacing pickFolder; openPath trusted-picker check

Closes the renderer→arbitrary-baseDir bypass at the bridge boundary.
The renderer no longer receives a raw filesystem path from the main
process; the picker dialog and the import call live in a single
main-process transaction.

Desktop main:

- runDesktopMain generates a per-process 32-byte secret and registers
  it with the daemon over the daemon's sidecar IPC *before* the
  BrowserWindow is created. registerDesktopAuthWithDaemon retries a
  few times because tools-dev / tools-pack spawn daemon, web, and
  desktop as siblings, so the daemon may not be listening yet on
  desktop boot. A failed registration logs a warning and the runtime
  refuses pickAndImport calls (no secret → no token can be minted).

- runtime.ts replaces the `dialog:pick-folder` IPC with
  `dialog:pick-and-import`. The handler shows the picker, mints an
  HMAC token bound to the chosen path, POSTs /api/import/folder via
  the discovered web URL with the token + body, and returns the
  daemon's ImportFolderResponse to the renderer (or a structured
  failure envelope). Renderer never sees the path or the token.

- shell:open-path now consults a new pure helper
  `isOpenPathAllowedForProject` that refuses folder-imported projects
  whose metadata lacks `fromTrustedPicker: true`. This is the literal
  interpretation of mrcfps's round-3 follow-up: openPath is gated to
  projects whose resolvedDir came from the trusted-picker flow, not
  just transitively via the import gate. Native projects (no
  baseDir → daemon-owned <projectsRoot>/<id>) are always safe to open.

- fetchResolvedProjectDir now returns a `ResolvedProjectDirContext`
  with hasBaseDir + fromTrustedPicker so the openPath handler can
  enforce the marker check.

- New `signDesktopImportToken` pure helper mirrors the daemon-side
  signer with the same `~`-separated wire shape, exported for the
  packaged workspace's test file.

Preload bridge:

- `pickFolder` is deleted. The new `pickAndImport(init?)` returns the
  daemon's import response or a structured failure. `openPath` keeps
  its existing signature; its trust gate now lives in the main
  process.

Web renderer:

- electron.d.ts drops `pickFolder` and adds `pickAndImport` with the
  shared DesktopPickAndImportResult union pulled from contracts.

- NewProjectPanel: when running on Electron (pickAndImport bridge
  present), the "Open folder" button calls pickAndImport atomically
  and forwards the response through a new `onImportFolderResponse`
  prop. On web (no bridge), the existing manual baseDir input keeps
  working — browser builds have no shell.openPath surface so a
  renderer-named path cannot escalate.

- EntryView and App.tsx pass through the new callback. App's
  `handleImportFolderResponse` updates state from the response without
  a second fetch (the import already happened in the main process).

Tests (apps/packaged/tests/desktop-project-root-gate.test.ts):

- 3 cases for `isOpenPathAllowedForProject`: native allowed,
  trusted-picker allowed, legacy folder-import refused.

- 6 cases for `signDesktopImportToken`: shape (~-separated), determinism,
  signature flips when secret/baseDir/nonce/exp changes.

- Existing fetchResolvedProjectDir cases extended for the new
  `context` shape and additional cases that prove the metadata
  inspection (hasBaseDir, fromTrustedPicker) reads the daemon
  response correctly.

* fix(daemon): make desktop import-folder gate fail-closed (PR #974 round 4)

lefarcen P1 on round 3 of PR #974: the gate's `secret == null → accept`
branch (originally intended to keep web-only deployments unaffected)
let a renderer bypass the import boundary in two real desktop edges:

- Startup race: desktop's REGISTER_DESKTOP_AUTH IPC hasn't reached the
  daemon yet, but the renderer is already alive in the BrowserWindow
  and races to fetch /api/import/folder directly with arbitrary baseDir.
- Daemon restart mid-session: the new daemon process boots tokenless
  while a desktop is still running. Same shape: renderer fetches the
  route, daemon falls through to "web mode", accepts the untrusted
  baseDir. shell.openPath rejects (no fromTrustedPicker marker) but
  the daemon's other file APIs (read/write project files, list
  directories) operate on the attacker-chosen path.

Two coordinated mechanisms close that:

(1) Sticky in-process flag. `desktopAuthEverRegistered` flips to true
    on first non-null `setDesktopAuthSecret(...)` and never goes back.
    setDesktopAuthSecret(null) (used by tests) does NOT relax the gate
    so production code can never silently fall back to fail-open. Add
    `resetDesktopAuthForTests()` for vitest cleanup.

(2) Orchestrator-pinned mode via OD_REQUIRE_DESKTOP_AUTH=1 read at
    module load. tools-dev / tools-pack / apps/packaged set this when
    the daemon is spawned in a desktop-bundled flow (separate commits).
    With the env set, the gate is active from request 0 — a renderer
    racing /api/import/folder before registration completes gets a
    503 DESKTOP_AUTH_PENDING (transient, retry).

Standalone-daemon (web-only) deployments where neither mechanism fires
keep the gate dormant and the route's behavior unchanged.

Also addresses lefarcen P3 (whitespace HMAC mismatch): the desktop
signs the exact picker output, so the daemon must verify the same
string. The previous version trimmed `baseDir` before HMAC, which
would reject legitimate paths whose final component carried edge
whitespace. Use the raw request-body baseDir for verification; the
existing trim()+realpath() logic still normalizes for fs operations.

New error code: `DESKTOP_AUTH_PENDING` (HTTP 503, retryable).

Tests:

- `stays fail-closed (503 DESKTOP_AUTH_PENDING) after a registered
  secret is cleared` — exercises the sticky flag.
- `verifies the exact request-body baseDir, not a trimmed version` —
  pins the round-4 P3 fix.
- All existing desktop-import-token-gate cases continue to pass; the
  beforeEach/afterEach/afterAll resetters now use
  resetDesktopAuthForTests() to honor the sticky flag.

* fix(tools-dev,packaged): pin desktop import-auth on daemon spawn

PR #974 round-4 P1 follow-through. The daemon-side fail-closed gate
needs OD_REQUIRE_DESKTOP_AUTH=1 in the daemon's spawn env whenever
the daemon is paired with a desktop, so the gate is active from
request 0 and the daemon-restart-mid-session bypass cannot reopen.

tools-dev:
- spawnDaemonRuntime accepts a `requireDesktopAuth` option that
  appends OD_REQUIRE_DESKTOP_AUTH=1 to the spawn env.
- startDaemon takes the same flag and additionally checks whether a
  desktop runtime is already alive in this namespace; either branch
  pins the env (revival case where the daemon died mid-session and
  the user runs `tools-dev start daemon` to bring it back up).
- startApp threads the bundled-target list down so the daemon spawn
  knows when desktop is queued in the same orchestration even though
  the daemon starts first.
- The `start` / `restart` / `run` command actions pass the resolved
  target list into startApp.

apps/packaged:
- Packaged builds always pair a desktop with the daemon, so
  startPackagedSidecars unconditionally sets OD_REQUIRE_DESKTOP_AUTH=1
  in the daemon child env. Headless builds also flow through this
  same path, so the same gate applies.

Standalone-daemon flows unaffected: `tools-dev start daemon` (alone,
no desktop running, no desktop in the bundled target list) does not
set the env, and the daemon's gate stays dormant — current web-only
behavior is preserved.

* fix(desktop,web): align project-id regex with daemon; surface pickAndImport failures

mrcfps round-4 nits on PR #974.

apps/desktop/src/main/runtime.ts (mrcfps #1): the previous client-side
regex `^[a-zA-Z0-9_-]+$` rejected `.` even though the daemon's
canonical isSafeId / POST /api/projects accept `[A-Za-z0-9._-]{1,128}`.
Result: dotted ids like `my-project.v2` were valid backend-side but
got "project id contains disallowed characters" before
fetchResolvedProjectDir even hit the network, regressing Continue in
CLI / Finalize for those projects. Align the regex with the daemon's
shape, comment-tag the rationale.

apps/packaged/tests/desktop-project-root-gate.test.ts: add a
regression case for a dotted id and one for the 128-char length cap
(the new regex exposes both, the old regex obscured the dotted one).

apps/web/src/components/NewProjectPanel.tsx (mrcfps #2): the
`if (!result || result.ok !== true) return` branch swallowed every
non-OK pickAndImport shape (`desktop auth secret not registered`,
`web sidecar URL not available`, daemon HTTP errors with details)
the same way as the explicit `{ canceled: true }` cancel — leaving
the user with a silent no-op when the trusted-picker flow couldn't
even get off the ground. Reserve silent-return for the cancel case
only; surface every other reason via a Toast (existing component,
already used by ProjectView for related Continue-in-CLI flows).
The new `formatPickAndImportErrorDetails` helper flattens daemon
ApiError envelopes into a single readable secondary line so the
operator sees both the category ("Open folder failed: daemon
returned HTTP 503") and the upstream reason
("desktop auth required but secret not yet registered").

* docs(architecture): document desktop folder-import auth boundary

lefarcen P3 on PR #974 round 4: the `Folder import` section in
docs/architecture.md still documented only realpath / sandbox /
RUNTIME_DATA_DIR checks and omitted the new desktop HMAC trust
boundary, replay/TTL behavior, fail-closed semantics, daemon-restart
edge, and legacy-import migration note. Without that subsection it's
hard to review whether the 60s TTL, the `~`-separated token shape,
or the legacy folder-imports needing re-pick are intentional product
decisions or overlooked gaps.

Add a "Desktop folder-import auth (PR #974)" subsection covering:
- The trust handshake (32-byte secret over sidecar IPC at desktop boot).
- Token shape (`${nonce}~${exp}~${signature}`), HMAC payload, and
  why `.` cannot be the field separator (ISO 8601 expiries embed dots).
- TTL and replay behavior (60s, single-use, 2× TTL upper bound).
- Fail-closed mechanisms — sticky in-process flag and
  OD_REQUIRE_DESKTOP_AUTH env var pinning.
- Web-only deployments are unaffected (browser builds have no
  shell.openPath surface).
- The `metadata.fromTrustedPicker` marker and the openPath-side
  defense-in-depth check.
- Legacy folder-imports need re-pick to use the Continue-in-CLI button.
- Daemon-restart edge: 503 DESKTOP_AUTH_PENDING until desktop
  re-registers; restart desktop to recover.

* fix(packaged): skip desktop-auth gate in headless mode (PR #974 round 5 P2)

Round 5 (lefarcen P2): packaged headless mode (daemon+web only, no
Electron) was inheriting OD_REQUIRE_DESKTOP_AUTH=1 from the round-4
unconditional pin in startPackagedSidecars. Headless never runs desktop
main, so no client could ever register an HMAC secret and folder import
returned 503 DESKTOP_AUTH_PENDING permanently — even though headless has
no shell.openPath surface to exploit.

Plumb a required `requireDesktopAuth: boolean` option through
startPackagedSidecars: apps/packaged/src/index.ts (Electron entry)
passes true; apps/packaged/src/headless.ts passes false. Extract
buildPackagedDaemonSpawnEnv as a pure helper so vitest can pin both
branches without spawning a child process.

Tests added in apps/packaged/tests/sidecars.test.ts cover both branches
plus OD_LEGACY_DATA_DIR / daemonCliEntry env forwarding edges.

Refs: nexu-io/open-design#974

* fix(desktop,daemon): lazy auth retry + canonical HMAC binding (PR #974 round 5 P1+P3)

Round 5 (lefarcen P1, mrcfps): a daemon restart under
OD_REQUIRE_DESKTOP_AUTH=1 left desktop holding a stale secret while the
new daemon process required a fresh registration — folder import
returned 503 DESKTOP_AUTH_PENDING permanently until the user restarted
desktop. Same dead-end if the startup handshake missed its retry window.

Round 5 (lefarcen P3): the daemon verified the HMAC against raw
request-body baseDir, then trimmed before realpath(). A picker selection
of "/tmp/foo " could authorize an import of "/tmp/foo" — token bound to
a different path than the one imported.

Three coordinated fixes:

1. P1 lazy retry: extract pickAndImportFolder as a pure helper that
   takes injected fetch / mintToken / registerDesktopAuth deps. On 503
   DESKTOP_AUTH_PENDING from /api/import/folder, re-invoke the
   registration callback once, mint a fresh token (new nonce + new exp
   keeps replay protection), and POST again. Single retry, no infinite
   loop. Other failure shapes return immediately to the renderer.

2. P1 wiring: runDesktopMain now ALWAYS passes desktopAuthSecret to the
   runtime regardless of whether the initial handshake succeeded, plus
   a registerDesktopAuthWithDaemon callback the runtime invokes lazily.
   Soften the startup warning text to match the new recovery semantics.

3. P3 binding: trim picker output ONCE on the desktop side before both
   signing the HMAC and POSTing. Daemon-side verification stays against
   raw request-body baseDir (round-4 behavior); the daemon's defensive
   trim before realpath() is now a no-op for desktop traffic and only
   load-bearing for web-mode callers (path.isAbsolute("  /foo  ") is
   false). End-to-end: desktop-signed string == request body == HMAC-
   verified string == realpath() input.

Tests:

- apps/packaged/tests/desktop-pick-and-import.test.ts (NEW, 7 cases):
  lazy-retry happy path; lazy-retry exhausted (re-register WAS called);
  single-attempt happy path (no unnecessary IPC); optional-callback
  no-op; non-503 failures bypass retry; network errors; non-PENDING 503
  bypasses retry.

- apps/daemon/tests/desktop-import-token-gate.test.ts: replace round-4
  whitespace test with two round-5 binding tests — the trimmed string
  flows end-to-end (HMAC verifies, project metadata.baseDir equals
  realpath of trimmed input), and a request whose body baseDir diverges
  from the HMAC-bound string is rejected 403.

docs/architecture.md §"Desktop folder-import auth" — update the daemon-
restart-edge bullet to describe the lazy-retry recovery (round 4 said
"restart desktop to recover", which is now wrong) and add a headless-
packaged-mode bullet describing the round-5 P2 gate exclusion.

Refs: nexu-io/open-design#974

* feat(sidecar-proto,daemon): surface desktopAuthGateActive over STATUS IPC (PR #974 round 6 prep)

Round 6 (mrcfps): the split-start dev flow `tools-dev start daemon` ->
`tools-dev start desktop` was leaving the daemon ungated because
`OD_REQUIRE_DESKTOP_AUTH=1` is only injected when daemon and desktop
spawn in the same orchestrator invocation. To fix that, tools-dev needs
to introspect the running daemon's gate state before launching desktop
main — but the existing STATUS IPC didn't carry the flag.

This commit extends `DaemonStatusSnapshot` with a required
`desktopAuthGateActive: boolean` and wires the daemon sidecar's STATUS
handler (and the public `status()` method on the handle) to recompute
the value from `isDesktopAuthGateActive()` per request, since the flag
flips after `REGISTER_DESKTOP_AUTH` and stays sticky.

Extracted `withCurrentDesktopAuthGate(snapshot)` as a tiny pure helper
so the wiring is testable without booting a real IPC server. The new
test pins four scenarios:
- no secret registered (web-only mode) -> false
- after `setDesktopAuthSecret(buf)` -> true
- after `setDesktopAuthSecret(null)` (sticky) -> still true
- input snapshot's stale value is overridden by the live flag

The orchestrator-side consumer lands in the next commit
(`tools/dev/src/desktop-auth-gate.ts`).

Refs: nexu-io/open-design#974

* fix(tools-dev): auto-restart ungated daemon before desktop start (PR #974 round 6 mrcfps)

Round 6 (mrcfps): the split-start dev sequence
`tools-dev start daemon` -> `tools-dev start desktop` was leaving the
daemon running without `OD_REQUIRE_DESKTOP_AUTH=1`. The env var is
only injected when (A) daemon and desktop spawn in the same
orchestrator invocation (`startApp` line ~682) or (B) a desktop
runtime is already alive at daemon spawn time (`startDaemon` lines
~595-596). Neither fires for the split flow, so a renderer (or any
local HTTP client) could `POST /api/import/folder` directly with an
arbitrary `baseDir` before the desktop's first registration POST.
Round-5's lazy retry didn't help: it triggers on `503 DESKTOP_AUTH_PENDING`,
and the ungated daemon returns 200.

Close the gap by introspecting the running daemon's
`desktopAuthGateActive` (added to the STATUS IPC in the prior
commit) at the start of `startApp(DESKTOP, ...)`. When the daemon
reports the gate inactive, stop the daemon (and web, if running),
respawn the daemon with `requireDesktopAuth: true`, restart web,
then proceed with the desktop start. Restart order is critical and
pinned by tests: web stops FIRST (so the web->daemon proxy doesn't
serve a transient 502 against the down-then-up daemon), then daemon
stops, then daemon respawns gated, then web restarts.

The bundled-targets path (`pnpm tools-dev`) is unaffected because
trigger (A) already armed the gate at first daemon spawn — the
helper costs one ~800ms STATUS IPC roundtrip and returns no-op.

Helper lives in its own module (`tools/dev/src/desktop-auth-gate.ts`)
so the regression test can import it without triggering the
`cli.parse()` side effect at the bottom of `tools/dev/src/index.ts`.
Five `node:test` cases pin the call sequence — no daemon, gate
active, gate inactive + no web, gate inactive + web running, log
shape — so a future refactor can't silently regress the gate.

Two synthetic `DaemonStatusSnapshot` literals in `inspectAppStatus`
and `inspect` (used when the IPC is unreachable) get
`desktopAuthGateActive: false` to satisfy the now-required type
field — semantically correct since "no daemon answering" trivially
means "no gate active."

`docs/architecture.md` adds a new bullet under the Desktop folder-
import auth section describing this auto-restart behavior.

Refs: nexu-io/open-design#974

* fix(daemon): combine finalize request-abort + timeout signals (PR #974 round 7 lefarcen P1)

Round 6 wired the route handler to pass `finalizeAbort.signal` into
`finalizeDesignPackage`, but the helper only created its own
DEFAULT_TIMEOUT_MS controller when no caller signal was supplied. The
result: a client that stayed connected could hold the finalize lock and
upstream call indefinitely. Always create the timeout controller; when
the caller passes a signal, combine both via `AbortSignal.any` so
neither cancel path replaces the other.

Adds two regression tests in finalize-design.test.ts:
- timeout fires when caller signal never aborts
- pre-aborted caller signal still cancels

Adds an internal `timeoutMs` option to FinalizeOptions so tests can
exercise the abort path without a 120 s wait or fake-timer chains.
Production callers omit it; default remains DEFAULT_TIMEOUT_MS.

* fix(daemon): allow PATCH preserving existing fromTrustedPicker marker (PR #974 round 7 lefarcen P2)

The PATCH /api/projects/:id handler was rejecting any metadata that
contained `fromTrustedPicker`, including the unchanged `true` marker
that the linked-folder UI re-spreads when editing `linkedDirs`. Trusted
folder-imported projects could not update other metadata fields without
400-ing on their own marker.

Switch the rejection condition from `'in'` to a value comparison: only
reject when the incoming value differs from the persisted one
(`patch.metadata.fromTrustedPicker !== existingMeta?.fromTrustedPicker`).
That keeps acquisition (existing=undefined, patch true) and flip
(existing=true, patch false) attempts blocked while letting the UI
re-spread the existing marker.

POST /api/projects stays strict; that path has no existingMeta.

Adds two regression tests in desktop-import-token-gate.test.ts:
- allows PATCH preserving the existing fromTrustedPicker:true marker
- rejects PATCH that flips fromTrustedPicker on a trusted project

* fix(desktop,packaged): main-process api uses daemon URL not webUrl (PR #974 round 7 lefarcen P2)

Packaged builds load the renderer from `od://app/` and report that URL
through `discoverWebUrl`. But Node-side `globalThis.fetch` (undici) does
not route through Electron's registered `od://` protocol handler — that
handler runs in the renderer's protocol scope, not in main-process Node.
So `pickAndImportFolder` and `fetchResolvedProjectDir` calls from main
silently failed in packaged builds against the protocol scheme.

Add `discoverDaemonUrl` to `DesktopRuntimeOptions` and `DesktopMainOptions`.
The packaged shell already has the sidecar's real `http://127.0.0.1:<port>`
URL (`sidecars.daemon.url` from STATUS IPC) — thread it through to the
runtime. Main-process API calls now prefer the daemon URL and fall back
to the renderer URL for tools-dev (where it is itself http://127.0.0.1).

`PickAndImportFolderDeps.webUrl` renamed to `apiBaseUrl` so the boundary
is explicit at the type level; `fetchResolvedProjectDir`'s first
parameter renamed similarly. tools-dev callers see no behavior change —
their web URL is already an http://127.0.0.1 URL Node fetch can hit.

Test (`apps/packaged/tests/desktop-pick-and-import.test.ts`):
- existing 7 cases updated to the new prop name (no behavior change)
- new case pins URL composition: builds `${apiBaseUrl}/api/import/folder`
  and never produces a custom-protocol URL.

Note for review: this test pins URL composition; full Electron protocol
handler integration (renderer fetch through `od://`) is not exercised in
unit tests here.

* fix(tools-dev): preserve daemon/web ports across desktop-auth gate restart (PR #974 round 7 lefarcen P2)

Round 6 added the split-start auto-restart in ensureDaemonGateForDesktop
to close the dev-flow gap where `start daemon` then `start desktop`
left the daemon ungated. The restart was passing the current
`start desktop` CLI options to startDaemonGated/startWeb, which meant a
stack started with `--daemon-port 17456 --web-port 17573` could be
silently moved to random ports during the hardening restart, breaking
browsers and scripts pinned to those ports.

Extract the running ports from the STATUS snapshots (daemon.url and
web.url) and forward them as explicit `{ port }` callback args. The
closure in `tools/dev/src/index.ts` overrides the corresponding option
when a port was extracted; null falls back to the original CLI flags.

Adds three regression tests in tools/dev/tests/desktop-auth-gate.test.ts:
- preserves the running daemon port across the hardening restart
- preserves the running web port across the hardening restart
- falls back to caller options (port:null) when the URL has no port

* fix(web): refresh useDesignMdState on file/chat events (PR #974 round 7 mrcfps)

useDesignMdState() previously only recomputed on mount and on explicit
refresh() (called once after finalize). Once the user kept working —
editing files or sending more chat turns — the stale/fresh badge could
drift out of sync because file mtimes and conversation updatedAt moved
past the recorded generatedAt without the hook re-checking.

Hook accepts an optional `refreshKey: number` arg; ProjectView keeps a
counter and bumps it on three events:
- file-changed SSE (covers tool-emitted file mutations)
- live_artifact* SSE (covers chat turns that emit artifacts)
- streaming `true → false` edge (covers pure-text chat turns)

The hook treats refreshKey as a compute() dep; React's Object.is
comparison short-circuits the no-op renders, so each bump is a single
recompute pass.

Adds a regression test in useDesignMdState.test.tsx:
- flips stale state after a refreshKey bump without remounting

* fix(web): degraded-state useDesignMdState on malformed provenance (PR #974 round 7 mrcfps)

useDesignMdState used to report `{ isStale: false, staleReason: null }`
when the parser could not extract a comparison timestamp from the
DESIGN.md `## Provenance` section. The pinned test made that the
documented behavior. As mrcfps pointed out, that fails open exactly
when the freshness signal is most untrustworthy: any provenance-
formatting drift silently disables the staleness warning.

Extend `DesignMdStaleReason` with a third variant `'unknown-provenance'`.
On `generatedMs === null`, return `{ isStale: true, staleReason: 'unknown-provenance' }`.
ContinueInCliButton renders a distinct chip text "Spec freshness
unknown — regenerate to refresh signal" for that variant; the button
stays enabled because not-comparable is not the same as broken state.

Tests:
- modify the existing pinned test to assert the new degraded state
- add an end-to-end useDesignMdState test feeding a malformed Provenance
  section through compute() so a regression that re-pins fresh-on-null
  at the hook level (not just computeStale) fails fast
- add ContinueInCliButton render + click tests for the new chip

---------

Co-authored-by: DevForgeAI CI/CD Engineer <devforge-ai@development.ai>
Co-authored-by: lefarcen <935902669@qq.com>
2026-05-10 11:44:32 +08:00
Marc Chan
b03a504da6
release: Open Design 0.6.0 (#1080) 2026-05-09 19:58:11 +08:00
tenderpooh
109722de3a
feat(desktop): export artifacts directly to PDF (#532)
* feat(desktop): export artifacts directly to PDF

* fix(desktop): PDF 내보내기 기본 여백 제거
2026-05-08 23:42:12 +08:00
lefarcen
2bb029cb58
release: Open Design 0.5.0 (#820)
0.5.0 已从 c21cbc6 发布(https://github.com/nexu-io/open-design/releases/tag/open-design-v0.5.0);本次 squash 把版本 bump 与 CHANGELOG [0.5.0] 条目带到 main 历史,便于后续 0.5.1 release 在 main 上走标准 dispatch 流程。
2026-05-08 00:41:01 +08:00
PerishFire
cb92c93ae0
Migrate beta release publishing to R2 (#805)
* Prebundle standalone web packaged runtime

* Harden mac standalone prebundle policy

* Prebundle mac daemon packaged runtime

* Prune mac Electron locales

* Maximize mac release artifact compression

* Publish beta mac artifacts to R2

* Use remote R2 uploads for beta releases

* Fail fast on beta R2 access issues

* Use S3-compatible uploads for beta R2 releases

* Decouple beta versioning from GitHub releases

* Remove legacy beta metadata source

* Address release beta review notes
2026-05-07 19:13:52 +08:00
lefarcen
ae4a08773a
chore(release): prepare 0.4.1 (#659)
- bump remaining monorepo package.json files to 0.4.1 after apps/packaged was already bumped in #637
- add CHANGELOG.md [0.4.1] - 2026-05-06 entry covering the startup hotfix and 19 merged PRs since 0.4.0:
  - Added: manual edit mode (#620), Cmd/Ctrl+P quick file switcher (#556), resizable chat panel (#563), PI status/cancel updates (#618), accessibility and RTL/Bidi craft modules (#587, #595), i18n structure checks (#608)
  - Changed: first-PR README links now surface help-wanted issues (#605)
  - Fixed: packaged contracts runtime exports (#577), packaged runtime beta gating (#637), ACP/MCP/agent fixes (#604, #612, #627), conversation error recovery (#623), native mac quit (#637)
  - Documentation/Internal: OD_DATA_DIR migration docs (#570), Simplified Chinese QUICKSTART (#578), zh-TW/ko README syncs (#586, #619), generated metrics (#592)

Release workflow validation runs after merge via release-stable.
2026-05-06 18:05:56 +08:00
lefarcen
963bbf2500
release: Open Design 0.4.0 (#454) 2026-05-05 23:39:40 +08:00
PerishFire
bbdd4e84b5
chore: enforce test directory conventions (#496)
* chore: enforce test directory conventions

Move package, app, and tool tests out of src and add guard enforcement so source directories stay source-only.

* ci: use guard and package-scoped tests

Run the new repository guard in CI and keep test execution aligned with package-scoped commands after removing root aliases.

* ci: align stable release guard check

Use the new repository guard in stable release verification after replacing the residual-JS-only script.

* chore: tighten test layout enforcement

Enforce sibling tests directories, typecheck moved test suites with dedicated configs, and refresh remaining guidance that pointed at src-based tests.

* chore: clarify no-emit test tsconfigs

Explicitly disable declaration-only emit in test tsconfigs so review tooling sees they are no-emit typecheck configs.
2026-05-05 15:34:22 +08:00
lefarcen
016c08183f
release: Open Design 0.3.0 2026-05-03 23:07:28 +08:00
lefarcen
62b01a6dbf
release: Open Design 0.2.0 (#297) 2026-05-02 22:28:59 +08:00
PerishFire
c6d11018a0
Refresh desktop integration control plane (#123)
* feat(dev): add desktop tools-dev control plane

* refactor(sidecar): split Open Design contracts

Move Open Design-specific sidecar protocol definitions into @open-design/contracts so sidecar and platform can remain descriptor-driven primitives.

* refactor(daemon): organize package sources

Keep daemon app code, tests, and sidecar entrypoints in separate package directories so each layer can be built and verified independently.

* chore(repo): streamline maintenance entrypoints

Centralize agent guidance by directory and reduce root command chains while preserving the existing build scope.

* docs: translate agent guidance to English

* fix(sidecar): tolerate stale IPC sockets

Remove stale Unix socket files only after confirming no listener is active, so tools-dev can restart after unclean shutdowns.
2026-04-30 14:23:53 +08:00