* Support overriding the Codex executable path
* Replace save-as-template prompts with an in-app dialog
* Seed local packaged app config from workspace
* Fix packaged config and connection test overrides
* Keep tools-pack mac config seeding self-contained
* Require absolute CODEX_BIN overrides
* docs(readme): add @nexudotio X link + Stay in the loop section
- Add X/Twitter follow badge in the header next to Discord, so visitors
can subscribe to release notes and milestone threads in one click.
- Add a short "Stay in the loop" section above "Star us" pointing to
@nexudotio on X. Discord is for chat, X is for the public milestones
(releases, new skills, new design systems).
No other changes.
* feat(web): add Follow @nexudotio pill to entry sidebar foot
- Surface the official X account next to the Settings/Pet/Language pills
so users can subscribe to release notes from inside the app, not only
the README.
- Uses the existing .foot-pill style (already supports <a>, has
text-decoration: none) and the existing 'external-link' Icon — no new
CSS, no new icons, no i18n key required (single short English label).
- Opens in a new tab with rel="noreferrer noopener".
Pairs with the README badge added in this same PR.
---------
Co-authored-by: Tuola-waj <ge@nexu.io>
* feat(settings): add connection test for providers and CLI agents
Adds a "Test" action in the Settings dialog that verifies the configured
provider (Anthropic/OpenAI/Azure/Google) or CLI agent without sending a
real chat. Backed by a new daemon endpoint and shared contracts, with
categorized inline statuses and i18n strings across all supported locales.
* fix(settings): address connection test review feedback
* fix(daemon): pass empty MCP servers for connection probes
* fix(connection-test): address review blockers
* fix(daemon): fail json stream runs on structured errors
* fix(contracts): build connection test subpath export
* Use draft CLI env in agent connection tests
* fix(i18n): add fallback ids for new curated content
* fix(web): add hover tooltips to Design Files action buttons (#283)
The batch-download, select-all, and clear-selection buttons in
DesignFilesPanel had no title attribute, so users hovering them saw no
tooltip. The other action buttons (refresh, new sketch, paste, upload)
already had titles. Added titles to the three missing ones using the
existing translation keys, so hover behavior is consistent across the
panel.
Closes#283.
* docs: point pi-ai links to pi-mono packages (#275)
The pi project moved from a standalone repo to the pi-mono monorepo.
The old URL https://github.com/mariozechner/pi-ai now 404s. Replaced
both shapes of reference:
- The reference-style [piai]: definition now points at
https://github.com/badlogic/pi-mono/tree/main/packages/ai
(the multi-provider LLM API package).
- Inline links whose visible text is the CLI tool 'pi' or 'Pi' now
point at
https://github.com/badlogic/pi-mono/tree/main/packages/coding-agent
(the interactive coding-agent CLI), so a reader clicking 'pi' in
the daemon-discovery section lands on the actual binary's docs.
Affected: README.md and 10 translated READMEs, plus docs/spec.md,
docs/architecture.md, docs/references.md, docs/roadmap.md.
Closes#275.
* fix(daemon): expand $HOME / ${HOME} in OD_DATA_DIR (#390)
Some launchers (systemd unit files, NixOS modules, certain Docker
entrypoints) pass OD_DATA_DIR with a literal '$HOME' or '${HOME}'
because no shell ever expands them. resolveDataDir previously only
handled '~/' shorthand, so '$HOME/.open-design' fell through to
path.resolve(PROJECT_ROOT, '$HOME/.open-design') and produced paths
like /opt/open-design/$HOME/.open-design.
resolveDataDir now expands '~', '~/...', '$HOME', '$HOME/...',
'${HOME}', and '${HOME}/...' to os.homedir() before the absolute /
relative branch runs. Rebuilds via path.join so the platform separator
is correct on Windows even when the input used forward slashes.
Tests: 7 unit tests cover empty/undefined, '~', '~/...', '$HOME',
'$HOME/...', '${HOME}/...', absolute paths, and relative paths.
Closes#390.
* fix(daemon): accept backslash separators + hermetic resolve-data-dir tests
Round 1 review feedback on PR #530.
The previous regex only matched forward-slash separators, so a Windows
launcher passing OD_DATA_DIR=$HOME\.open-design or ${HOME}\.open-design
fell through to path.resolve(projectRoot, ...) and produced a directory
named $HOME or ~ under projectRoot. The regex now accepts both forward
and back slashes for the home-prefix separator.
The previous tests called the real resolveDataDir against literal
~/od-test, $HOME/od-test, etc., which created and write-checked
directories under the developer's or CI runner's actual home. The tests
now stub os.homedir() with vi.spyOn to a per-test mkdtemp directory and
remove it in afterEach, so no test ever writes outside its own sandbox.
Added explicit fixtures for the Windows backslash forms ($HOME\od-test,
${HOME}\od-test, ~\od-test) so launcher coverage stays cross-platform.
12/12 resolve-data-dir tests pass, daemon typecheck clean.
* fix(docs,daemon): apply pi-mono links to README.es and await test cleanup
Round 2 review feedback on PR #530.
README.es.md was added in upstream #552 after my pi-mono link sweep
landed, so the daemon-discovery paragraph (line 222), the [piai]
reference (line 684), and the Pi table row (line 709) still pointed
at the broken https://github.com/mariozechner/pi-ai URL. Applied the
same replacements: the [piai] ref now points at packages/ai, and the
inline Pi link now points at packages/coding-agent. Spanish readers
get the same coverage as the other 11 locales.
The absolute-path test in tests/resolve-data-dir.test.ts dropped its
fixture via void rm(abs, ...), so a failed async removal could leak
rdd-abs-* directories from the suite. The test is now async and
awaits the rm in the finally block, matching the awaited cleanup in
afterEach. 12/12 resolve-data-dir tests still pass, daemon typecheck
clean.
* fix(daemon): share $HOME expander between OD_DATA_DIR and OD_MEDIA_CONFIG_DIR
Round 3 review feedback on PR #530.
resolveDataDir (server.ts) now expands $HOME / ${HOME} / ~, but
media-config.ts had its own resolveOverrideDir that only handled ~/.
Because configFile() falls back to OD_DATA_DIR when OD_MEDIA_CONFIG_DIR
is unset, setting OD_DATA_DIR=$HOME/.open-design split state: SQLite,
projects, and artifacts went to the expanded path while
media-config.json stayed under <projectRoot>/$HOME/.open-design.
Stored provider keys then appeared missing on the next read.
Extracted the home-prefix expansion into apps/daemon/src/home-expansion.ts
so resolveDataDir and resolveOverrideDir share one resolver. Both now
recognize ~ / $HOME / ${HOME} (bare tokens) and ~/, ~\, $HOME/, $HOME\,
${HOME}/, ${HOME}\ (prefix forms with either separator).
Three new media-config routing tests cover the OD_DATA_DIR fallback for
$HOME/..., ${HOME}/..., and the OD_MEDIA_CONFIG_DIR explicit-override
$HOME/... case so the co-location guarantee is locked down by tests.
Daemon typecheck clean. Tests pass on Linux CI; the existing pattern in
the file uses process.env.HOME which os.homedir() reads on POSIX.
Resolve-data-dir tests stay hermetic via vi.spyOn.
* docs(daemon): media-config comments reflect full $HOME / ${HOME} expansion
Round 3 review feedback on PR #530 (lefarcen, P3 non-blocking).
The file-header and resolveOverrideDir() function comment said only
~/ expands. Updated both to mention the shared expandHomePrefix()
helper and the full set of forms it handles (~, $HOME, ${HOME} with
either separator), so a future reader does not need to chase the
implementation to understand what env values are accepted.
* test(daemon): stub os.homedir() in media-config routing tests
Round 4 review feedback on PR #530.
The new $HOME / ${HOME} routing tests relied on process.env.HOME being
read by os.homedir(), which works on POSIX but is unreliable on Windows
(Node prefers USERPROFILE / profile APIs there). The tests would expand
to the real user home while fixtures were written under the per-test
homeDir, causing platform-specific failures in the same area this PR
is making cross-platform.
The inner describe block now stubs os.homedir() via vi.spyOn to return
the per-test homeDir, matching the pattern in resolve-data-dir.test.ts.
Restored in afterEach. The four $HOME-form routing tests now pass on
both POSIX and Windows.
Daemon typecheck clean. The two OAuth fallback test failures unrelated
to this change (real ChatGPT/Codex tokens in the local env) remain
out-of-scope.
* fix(i18n): drop duplicate uk.ts promptTemplates keys after rebase
Upstream #674 added the same Ukrainian translations my earlier
commit added. The rebase landed both copies; tsc rejects duplicate
property names. Drop my copies so #674 (which is now upstream) is
the single source for these keys.
---------
Co-authored-by: Nagendhra <nagendhra405@gmail.com>
* fix(daemon): unbreak Claude Design ZIP import on Node 24 and raise file ceiling
- Skip inflateRawSync when an entry's central-directory uncompressedSize is 0;
Node 24 rejects { maxOutputLength: 0 } with ERR_OUT_OF_RANGE, which silently
killed the entire import for any zip containing an empty file or a streaming
entry whose size is only present in the data descriptor.
- Raise MAX_FILES from 500 to 5000. Real-world design-system exports commonly
exceed 500 files; MAX_TOTAL_BYTES (100 MB) and MAX_FILE_BYTES (25 MB) already
cap pathological inputs.
- Add regression tests for both: zero-byte deflate entry, central directory
advertising uncompressedSize=0, and a 600-file zip.
Refs #590
* fix(daemon): preserve real payload when central uncompressedSize is 0
Reviewers correctly flagged that the previous Buffer.alloc(0) fast-path
trusted the central directory's uncompressedSize, which is unreliable for
streaming/data-descriptor zips: an entry whose central record reports 0
can still carry real deflated bytes. The earlier fix wrote empty files to
disk in that case, and the post-condition body.length !== uncompressedSize
check still passed because both sides were 0.
- Inflate streaming entries with maxOutputLength = MAX_FILE_BYTES when the
central directory advertises 0, so legitimate non-empty payloads decode
fully instead of being silently truncated.
- Move size enforcement post-decode: per-file and total-byte budgets are now
computed from the actual decoded length, and the strict equality check is
skipped when central was 0 (i.e., genuinely unknown).
- Keep the empty-deflate degenerate case (compressed.length === 0) safe by
short-circuiting before zlib instead of relying on uncompressedSize.
Tests:
- New: streaming-zip case with central uncompressedSize=0 + non-empty body
asserts the on-disk file matches the original bytes (would have been
silently truncated under the previous fix).
- New: oversized streaming entry (> MAX_FILE_BYTES) is still rejected even
though the central directory under-reports.
- The original 0-byte and >500-file regressions remain covered.
* fix(daemon): surface OpenCode error frames + treat empty-output runs as failed
Closes#691. OpenCode runs would silently complete in ~3 seconds without
producing any visible chat output and still be rendered as a successful
turn — three independent bugs along the structured-stream path conspired
to produce this silent-failure shape.
## Bug 1 — `apps/daemon/src/json-event-stream.ts:85-91`
OpenCode emits structured error frames on stdout (e.g. provider auth
failures, network errors, schema mismatches) and still exits 0. The
parser was downgrading these to `{type: 'raw', line: ...}`, which the
chat UI does not render as an assistant message. The error string was
discarded as "no-op output."
Fix: emit a proper `{type: 'error', message, raw}` event matching the
qoder-stream contract that the daemon's existing error-handling path
already recognises.
## Bug 2 — `apps/daemon/src/server.ts:4199-4205`
Even after Bug 1 was fixed, the json-event-stream branch wired the
parser to a bare `(ev) => send('agent', ev)` lambda — bypassing the
`sendAgentEvent` wrapper that interprets `type:'error'` events and
sets the `agentStreamError` flag the close handler reads to flip the
run to `failed`. So an emitted `error` event would just be forwarded
as a no-op `agent` SSE event with no lifecycle effect.
Fix: route json-event-stream through `sendAgentEvent`, mirroring the
qoder-stream-json wiring at line 4175.
## Bug 3 — `apps/daemon/src/server.ts:4220-4234`
Even after Bugs 1 and 2 are fixed, there's still a class of runs where
OpenCode never emits any error frame, never emits any substantive
event, and exits 0. Pre-fix this was marked `succeeded` and the user
saw a blank chat with no diagnostic.
Fix: track `agentProducedOutput` inside `sendAgentEvent` (set on
`text_delta`, `thinking_delta`, `tool_use`, `tool_result`, `artifact`
— deliberately NOT on `status` / `usage`, since a model can emit
token-usage numbers for an empty completion). When the close handler
sees `code === 0 && trackingSubstantiveOutput && !agentProducedOutput`
the run is marked `failed` with an explicit AGENT_EXECUTION_FAILED
SSE error so the chat shows a clear reason instead of a silent
empty turn.
The check is gated by `trackingSubstantiveOutput` so it only fires
on streams that actually contribute to the output flag (currently
qoder-stream-json and json-event-stream). ACP sessions and plain
stdout streams keep their existing success/failure determination.
## Tests
- 3 new unit tests in `apps/daemon/tests/json-event-stream.test.ts`
pin the OpenCode error event shape: full repro
(`error.data.message`), `error.name` fallback, and the
generic-fallback shape when `error` is empty.
- All 60 daemon test files (851 tests) pass on `pnpm --filter
@open-design/daemon test`. All 42 web test files (309 tests) pass
on `pnpm --filter @open-design/web test`.
- Full repo `pnpm typecheck` clean.
## Live verification
Verified end-to-end via a stub `opencode` binary that mimics each of
the failure shapes against `pnpm tools-dev run web`:
1. Stub emits `{"type":"error",...}` then `exit 0` — run now ends as
`failed` with the OpenCode error message surfaced as an SSE
`error` event. Pre-fix this was `succeeded` with an empty chat.
2. Stub emits nothing then `exit 0` — run now ends as `failed` with
"Agent completed without producing any output…" diagnostic.
Pre-fix this was `succeeded` with an empty chat.
3. Stub emits a normal `step_start` / `text` / `step_finish` sequence
then `exit 0` — run still succeeds. (Regression check.)
## Out of scope (mentioned for the next person)
- `claude-stream-json` and `copilot-stream-json` still wire to a bare
`(ev) => send('agent', ev)` and don't currently parse `type:'error'`
frames. If their CLIs ever start emitting structured error events
the same pattern (route through `sendAgentEvent` + emit proper
`type:'error'`) applies. Not in scope here because we have no
evidence those CLIs do this today, and changing the wiring without
a confirmed failure mode risks regressing currently-working flows.
- ACP sessions (`pi-rpc`, `acp-json-rpc`) own their own success /
failure determination via `acpSession?.hasFatalError()` and the
empty-output guard explicitly skips them via
`trackingSubstantiveOutput`.
- Plain stdout streams have no event-level tracking, so the empty-
output guard skips them too. Diagnosing a no-output plain-stream
agent is a separate problem that needs different signals.
* chore: retrigger CI on top of green main (post #697 i18n backfill)
* feat(linux): add headless mode for install/start/stop operations
* docs(linux): document headless mode commands and usage
* refactor(linux-headless): write web-root.json instead of polling IPC for URL
* fix(linux-headless): fail start when web identity never appears instead of returning success
* docs(linux-headless): add use-case context and clarify launcher path dependency
* fix(linux-headless): ensure launcher, identity and shutdown align with tools-pack
- Bake OD_DATA_DIR into launcher so manual runs use the same paths as tools-pack
- Validate web-root.json fields before accepting to reject stale identity
- Remove web-root.json on successful stop
- Add IPC server for graceful STATUS/SHUTDOWN handling
* fix(linux-headless): create IPC server before writing web-root.json
* docs: fix broken pi-ai links, point to correct pi-mono packages
All links to https://github.com/mariozechner/pi-ai returned 404 after
the project was restructured into the badlogic/pi-mono monorepo.
- "pi" / "Pi" (the CLI tool the daemon scans for) now points to
packages/coding-agent
- "pi-ai" (the multi-provider LLM API) now points to packages/ai
via the shared [piai] reference definitions
Closes#275.
* fixup! Merge remote-tracking branch 'upstream/main' into docs/fix-pi-pi-ai-links
Fix [piai] reference in README.ar.md and README.es.md: was incorrectly
pointing to packages/coding-agent (pi CLI) instead of packages/ai (pi-ai
provider library).
* fixup! fix row order in README.uk.md: move Pi after DeepSeek TUI to match English README
* feat: add accent color control and launcher for Open Design
* fix: remove launcher binary from PR
* test: cover accent appearance edge cases
---------
Co-authored-by: ferasbusiness666 <ferasbusiness666@users.noreply.github.com>
* ci: notify Discord #resolved on issue close-via-merged-PR
* ci: address review feedback on Discord #resolved workflow
P1:
- Add contents:read permission (required by listPullRequestsAssociatedWithCommit)
- Drop cross-referenced timeline fallback to eliminate false positives from
plain mentions; closed-event+commit_id is now the only resolver path
(also fixes the cross-repo number-collision concern Codex raised)
P2:
- Validate webhook URL prefix before POST (reject misconfigured secrets)
- Retry on Discord 429 up to 3 times honouring Retry-After header,
bounded 1..60s, with sane default if header missing
P3:
- allowed_mentions: { parse: [] } so issue/PR titles can't @everyone or
ping roles/users in #resolved
* feat(skills): add social-media-dashboard skill + Totality Festival design system
- New skill 'social-media-dashboard': single-screen creator analytics
dashboard with platform switcher (X / GitHub / LinkedIn / YouTube /
Instagram), KPI row, growth chart with annotations, top-post / top-PR
preview, trending topics, and top comments. Includes a self-contained
example.html (Totality Festival styled, X + GitHub tabs, live KPI
ticker, GitHub contributors grid, world-map audience geography).
- New design system 'totality-festival': cosmic-premium dark glassmorphic
system with amber corona highlights and cyan atmospheric accents.
Mirrors Google Labs' design.md spec example so skills can be
validated against an upstream reference.
- Fix swatches parser in apps/daemon/src/design-systems.ts so it
recognises the '- **Name:**' bold-with-inner-colon form used by
several existing systems (ant, totality-festival, ...). Previously
only the '**Name** (`#hex`)' form was matched, which left their
picker thumbnails empty.
* feat(skills): polish social-media-dashboard example + add PR preview
- Top Post media block: replace empty gold frame with an inline SVG
thumbnail (radial glow + ascending data curve + amber/cyan pulse dots
+ play icon + 'LIVE · 0:22' caption). Visually echoes the live-artifact
story the post copy is selling.
- Hoist the brand-mark linearGradient into a global SVG defs block at
the top of <body> so all three avatars (header, user, top-post) can
reference url(#brandRing) and render the teal arrow consistently.
Previously only the header SVG carried the gradient definition, so
the user and post avatars rendered as empty rings under headless
capture.
- Add hero.png preview to .preview/ for the PR description.
---------
Co-authored-by: Tuola Ge <gexingli@refly.ai>
* feat(daemon): export project transcript to disk for downstream consumption
Adds exportProjectTranscript(db, projectsRoot, projectId, options?) — a
pure function that walks SQLite-backed conversation history and writes a
structured, lossless JSONL transcript to <projectDir>/.transcript.jsonl.
This is the input primitive that #450's "Finalize design package"
synthesis step needs. Landed ahead of the synthesis endpoint as a small,
reviewable, well-tested unit — no HTTP route, no LLM call, no web UI in
this diff. PR 2 will wire POST /api/projects/:id/finalize on top of it.
Format: JSONL with header line + per-conversation marker lines +
per-message lines. Compact encoding saves ~20–30% on synth-call tokens
vs pretty-printed JSON. schemaVersion field reserved on the header for
incompatible changes later.
Coalescing: events_json carries streaming text_delta / thinking_delta
chunks plus tool_use / tool_result / thinking_start markers and
telemetry. The export collapses runs of same-type deltas into terminal
text / thinking blocks via arrival-order with type-change flush,
preserving any interleaving with tool blocks. Telemetry (status, usage,
raw) is dropped. thinking_start is treated as an explicit flush trigger
so multiple thinking blocks in one message survive intact.
Content fallback: user-typed messages persist as plain text in
messages.content with events_json = NULL because user input does not
flow through the streaming pipeline. When event-derived blocks come
back empty, fall back to a single text block from content so user
prompts are not silently dropped.
Atomic write: tmp filename includes pid + crypto-random suffix so
concurrent exports for the same project cannot collide. fsync before
rename so a crash between rename and power loss cannot lose bytes.
Hidden file: leading dot keeps .transcript.jsonl out of listFiles
(projects.ts:54), the archive zip, and the gallery.
Tests: 14 unit cases — empty project, text/streaming coalescing, tool
ordering, telemetry filtering, type-change flush, text↔thinking↔text
interleaving, thinking_start flush, multi-conversation chronological
order, atomic-write hygiene, content fallback (both directions),
malformed events_json, and unsafe project id rejection.
Refs: nexu-io/open-design#450
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(daemon): address PR #493 review feedback for transcript export
Addresses every blocker, P2, and P3 raised on
https://github.com/nexu-io/open-design/pull/493:
- Blocker (event-shape mismatch): coalescer now switches on the
PersistedAgentEvent kind discriminator (text/thinking/tool_use/
tool_result), reading the shared `text` field for content kinds,
matching what apps/web/src/providers/daemon.ts:347-394 actually
persists into messages.events_json. Empirical confirmation: live
SQLite contains zero `type` keys.
- Removed @ts-nocheck from the source file; added inline types for
Db (Database.Database), ConversationRow, MessageRow, AttachmentRef,
CommentAttachmentRef, and Block. Tests retain @ts-nocheck per
codebase convention. Note: db.ts still uses @ts-nocheck, so the
new types catch drift inside transcript-export.ts itself, not at
the SQLite-helper boundary.
- parseEvents now distinguishes null / malformed / not_array / ok
cases; non-null-but-unparseable rows emit a console.warn with
project+message id before falling back to content.
- Switched temp-write from writeSync (which can return short) to
writeFileSync({flag:'wx'}); explicit fsync via reopen before
rename, per reviewer concern about partial-write durability.
- Added per-project lockfile (.transcript.lock) acquired with
openSync(..., 'wx') and released in finally; concurrent exports
throw the new TranscriptExportLockedError. Stale-lock recovery
is documented as a known limitation in the file header.
- Header gains attachmentCount, commentAttachmentCount, and explicit
attachmentsInlined: false. Per-message lines gain attachments /
commentAttachments references (paths only, not bytes; synthesis
reads files from disk by path). schemaVersion bumped 1 -> 2 so
the change is explicit; v1 was never consumed.
- mkdirSync(dir, { recursive: true }) at entry covers projects
with DB rows but no on-disk directory yet (codex bot finding).
- Refactored node:fs imports from named to default
(import fs from 'node:fs') so vitest spies in tests #15-#17 can
redefine properties on the underlying CJS exports object. ESM
namespace imports of node:fs produce a frozen Module Namespace
Object that vi.spyOn cannot mutate; default-import returns the
CJS module.exports which is mutable.
- Inline PersistedAgentEvent union: the daemon tsconfig does not
resolve the `@open-design/contracts/api/chat` subpath export, so
the union is restated in the source. Schema-mismatch tests cover
the case where the contract would diverge.
- Test count 14 -> 24: failure injection for writeFileSync /
fsyncSync / renameSync, existing-file replacement, lockfile
contention (lockfile-pre-create design — synchronous API can't
race via Promise.allSettled), parse-warning cases (malformed +
not-array), attachments header + per-message coverage, missing-
project-dir case.
Refs nexu-io/open-design#450 (does not close).
* fix(daemon): preserve thinking-segment boundaries on status thinking-start
Codex flagged this as a P2 on PR #493a39d430:
https://github.com/nexu-io/open-design/pull/493#discussion_r3188524878
The web translator emits `{ kind: 'status', label: 'thinking' }` at every
thinking_start (apps/web/src/providers/daemon.ts:367-369). The previous
default branch dropped all status events without flushing the active
accumulator, so two thinking segments separated only by that marker
merged into one block — losing the original boundary and making the
transcript non-lossless for downstream synthesis.
coalesceBlocks now matches `status` explicitly: when `label === 'thinking'`
the prior accumulator flushes; other status labels and usage / raw drop
without flushing as before. Behavior fix within schemaVersion 2; no
shape change.
Test #25 verifies the boundary preservation:
[thinking 'a', thinking 'b', status 'thinking', thinking 'c', thinking 'd']
→ blocks: [thinking 'ab', thinking 'cd'] (two blocks, not one)
Existing test #5 still passes because it uses status with label 'streaming'
which remains pure telemetry and does not flush.
Suite: 506 -> 507 daemon tests, all green.
Refs nexu-io/open-design#493
---------
Co-authored-by: DevForgeAI CI/CD Engineer <devforge-ai@development.ai>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(skills): add 5 Orbit briefing templates
Introduces a new "orbit" scenario family in the Examples gallery for
morning-briefing surfaces. Each template lives at the top of "我的设计"
and aggregates yesterday's connector activity into a single page.
- orbit-general: adaptive bento dashboard that fans across 12-16
connectors, where each module picks its own UI form by data type
(list / avatar stack / status ring / heatmap / file grid / alert
card / kanban / etc.)
- orbit-github: GitHub-flavored single-connector digest mirroring the
Notifications + PR-diff visual language
- orbit-gmail: Gmail-flavored digest rendered as a Daily Digest email
inside the three-pane inbox
- orbit-linear: Linear-flavored digest in the dark Inbox + cycle-
progress layout
- orbit-notion: Notion-flavored digest authored as a native Notion
page (callout / toggle / database table)
The new scenario value 'orbit' surfaces as a filter pill in
ExamplesTab automatically; no UI code change required.
* fix(skills): reframe Orbit skill descriptions as pipeline-triggered
The original descriptions framed each skill as a standalone "X-flavored
briefing template" the user picks. They are actually skills the Orbit
daily-digest pipeline selects automatically based on which connectors
the user has authenticated, then runs against live connector data.
Rewrites both `description` and `example_prompt` for all 5 templates:
- orbit-general: invoked when 2+ connectors are connected; aggregates
the past 24h across every authenticated source
- orbit-github / orbit-gmail / orbit-linear / orbit-notion: invoked
when the named connector is the user's only connection (or scope is
explicitly limited to it); pulls the past 24h from that connection
alone
All 5 now state explicitly that they are not user-triggered — the
Orbit scheduler invokes them.
* feat(examples): add Orbit pill to the mode filter row
Surfaces the Orbit briefing skills as a top-level "type" filter in the
Examples gallery, alongside Prototype · Desktop / Mobile / Slide deck /
Docs & templates. Filter matches skills with scenario === 'orbit'.
- ExamplesTab: extend ModeFilter and MODE_PILLS with 'orbit'; teach
matchesMode and modeCounts about it
- i18n: add 'examples.modeOrbit' to Dict and to all 16 locale files
('Orbit' is left untranslated as a brand name)
* polish(orbit-general): real Figma preview image + revised comment
Replaces the empty gray placeholder in the Figma module with an
Unsplash UI-design photo, and rewrites the mock comment to read like
a substantive design-review note rather than a nit about button
placement.
* feat(examples): eager-load card previews via IntersectionObserver
Card previews previously only loaded on hover, leaving the example
gallery showing 'Hover to preview' placeholders for everything below
the fold. Now each card observes the viewport and prefetches its HTML
800px before scrolling into view, so the iframe is ready by the time
the user reaches it.
Hover remains as a fallback path (and for browsers without
IntersectionObserver, the card loads immediately on mount).
Also reverts the Unsplash photo on the orbit-general Figma module
back to the gray placeholder — the stock image semantically misread
as a Photoshop screenshot rather than a Figma artboard.
* feat(orbit-general): drop Figma connector module
Removes the Figma bento card and its scoped CSS, plus the orphaned
Top-3 entry that referenced a Figma comment. Reassigns Top-3 #2 to
a Notion document review so the priority list stays aligned with
the connectors actually rendered.
* i18n(skills): translate Orbit example prompts to English
The example_prompt is what gets injected into the chat input when a
user clicks 'Use this prompt', and is read by the agent verbatim. It
should match the SKILL.md description language (English), not the UI
locale. Replaces the Chinese drafts with English equivalents across
all 5 Orbit skills, and drops the Figma reference from orbit-general
since that connector module was removed earlier.
* fix(skills): rewrite Orbit SKILL.md bodies with reproducible specs
Earlier the bodies were too abstract (only a connector→UI mapping
table and a one-line style note), so agents running the skill could
not reproduce the shipped example.html and got stuck in long retries.
Each SKILL.md body now contains:
- exact color tokens lifted from the example.html
- type stack and font sizes
- a section-by-section page spec (top-to-bottom)
- chip / pill / icon rendering rules
- forbidden list
The example_prompt is collapsed back to a one-line user intent so the
skill body is the source of design truth.
Covers all 5 templates: orbit-general, orbit-github, orbit-gmail,
orbit-linear, orbit-notion.
* feat(orbit): make every connector item clickable
Each Orbit briefing template now links its rows / cards to the matching
source URL so users can jump straight from the morning digest to the
underlying connector.
- orbit-general: each bento card gains an 'Open in {connector} ↗' CTA
built from a connector→URL map; each Top 3 card becomes an anchor
- orbit-github: every event row opens the corresponding github.com
pull/issue URL parsed from the row identifier; the header logo links
to the repo
- orbit-linear: each issue row gains a small ↗ button that opens
linear.app/{team}/issue/{ID}
- orbit-gmail: action and reply buttons jump to a Gmail search URL
scoped to the sender
- orbit-notion: page-link spans wrap as anchors and database rows are
click-to-open against notion.so
All links use target="_blank" rel="noopener noreferrer".
* fix(skills): force agents to mirror example.html 1:1
Earlier skills told the agent the example was 'source of truth' but
left phrasing soft, so agents felt free to add extra UI elements
(snoozed-mail row, extra yellow stars on inbox rows, etc.) that
were not in example.html.
Each Orbit SKILL.md now opens with a 'Source-of-truth protocol' that
forces the agent to:
1. read example.html before writing any output
2. mirror its DOM structure / class names / module count / element
order 1:1
3. only refresh mock values; never invent additional UI elements,
rail entries, sections, badges, or chrome ornaments
The reference sections that follow stay informative for tokens and
visual language but are explicitly demoted from spec to commentary.
* fix(orbit-gmail): remove three-pane / left-rail / inbox-list claims
The example.html is a single-column page: Gmail top header + the
opened Orbit Daily Digest email (toolbar / subject / sender / digest
body / reply bar). Earlier copy described a Gmail three-pane app with
Compose button, label list, Categories tabs, and an inbox listing —
none of which exist in the actual asset.
- example_prompt: drops 'three-pane inbox' phrasing
- description: same
- body: rewrites Page sections to mirror the real header → email-chrome
layout, top to bottom; explicitly forbids left rail, inbox list, and
Categories tab strip
* feat(orbit): forbid external design systems in all 5 skills
Each Orbit briefing template ships its own complete visual language
baked into example.html (Gmail / GitHub / Linear / Notion / Open
Orbit's editorial bento). Adds an explicit 'Design system policy'
block telling the agent to:
- ignore any DESIGN.md attached to the active project
- ignore brand tokens or Figma files supplied via chat
- use exclusively the colors / fonts / radii from example.html
This is a hard constraint: an Orbit briefing must look like the
connector it represents, not like the user's brand.
* feat(newproj): hide design-system picker for skills that opt out
Skills can declare 'od.design_system.requires: false' in SKILL.md to
opt out of DESIGN.md injection (the Orbit briefing skills do this —
their example.html ships with a complete connector-native visual
language). When the active default skill for a tab opts out, hide the
design-system picker so we don't ask the user to attach a brand we'll
then ignore.
Existing tabs that don't host a default skill (template, other) keep
the picker. The check only fires for prototype / live-artifact / deck.
* review: address P2 reviewer feedback
P2 — Connector family coverage gaps (orbit-general):
Adds Finance, CRM/Sales, Support, Analytics, Infrastructure, Security
rows to the connector→UI mapping table (now 16 families). Adds a
'Fallback heuristics' subsection so unknown connectors are routed by
data shape (numbers + time series → Alerts, rows + status field →
Task mgmt, etc.).
P2 — 'Forbidden' rules too vague (all 5 skills):
Rewrites every Forbidden section as a paired 'Don't / Do' constraints
table so each negative is paired with a concrete positive. Replaces
obvious bans (lorem ipsum) with substantive ones (real-shaped mock
copy, plausible identifiers, dev-team label hues, etc.).
* ci: register orbit skills in de/ru/fr en-fallback lists
The localized-content coverage test asserts that every skill in
skills/ is either translated or explicitly declared as falling back
to English in the LOCALIZED_CONTENT_IDS bundle. The 5 new orbit
skills weren't in any bundle, so the workspace validation job failed
on the de/ru/fr coverage assertions.
Adds the 5 orbit-* ids to DE/FR/RU_SKILL_IDS_WITH_EN_FALLBACK so
those locales explicitly fall back to the SKILL.md English copy
(matching the minimal-change posture chosen earlier in this PR).
* fix(daemon, packaged): unbreak GUI-launched agent detection on minimal PATHs (#442)
GUI-launched daemons (Finder/Dock on macOS, .desktop on Linux) inherit a
stripped PATH from launchd / the desktop session and don't read the
user's interactive shell rc files, so any CLI installed via `npm i -g`
under a sudo-free prefix like ~/.npm-global was silently undetected.
Two layers maintained their own copies of the user-toolchain bin list
(`apps/daemon/src/agents.ts:userToolchainDirs` for the resolver,
`apps/packaged/src/sidecars.ts:resolvePackagedPathEnv` for the packaged
sidecar PATH builder) and had already drifted on `~/.asdf/shims` and
`~/Library/pnpm`. Adding ~/.npm-global to one side would have
preserved the same anti-pattern.
Extracts `wellKnownUserToolchainBins` into @open-design/platform as the
single source of truth, has both layers consume it, and extends the
list to cover ~/.npm-global/bin, ~/.npm-packages/bin, plus
$NPM_CONFIG_PREFIX/bin / $npm_config_prefix/bin for users with a
non-standard prefix. New vitest coverage in the platform package and
a regression test in apps/daemon/tests/agents.test.ts modelled on the
existing mise case.
Verified end-to-end: under PATH=/usr/bin:/bin:/usr/sbin:/sbin (the
launchd default a `.app` actually inherits), `resolveAgentExecutable`
now returns ~/.npm-global/bin/gemini instead of null.
* fix(daemon): isolate OD_AGENT_HOME resolution from $NPM_CONFIG_PREFIX leakage
Address review feedback on PR #614:
- mrcfps spotted that the daemon wrapper called wellKnownUserToolchainBins
without passing `env`, so the helper read its default process.env. A
developer or CI runner with NPM_CONFIG_PREFIX / npm_config_prefix
exported would inject that real <prefix>/bin into resolveOnPath() even
while the OD_AGENT_HOME hook pointed home at a temp fixture, making
agent-detection tests environment-dependent. Reproduced locally: with
OD_AGENT_HOME=<tmp> + NPM_CONFIG_PREFIX=/Users/me/.npm-global,
resolveAgentExecutable({ bin: 'codex' }) returned the real machine's
binary instead of null. Wrapper now passes `env: {}` whenever
homeOverride is set, alongside the existing includeSystemBins gate.
- lefarcen suggested also handling whitespace-only NPM_CONFIG_PREFIX
values (e.g. NPM_CONFIG_PREFIX=" ") so the helper does not emit a
bogus "<whitespace>/bin" entry. Added a .trim() check before
appending.
- lefarcen also suggested a comment pointer from the daemon wrapper to
the platform helper so readers don't have to grep. Added the
reference inline.
Coverage:
- packages/platform/tests/index.test.ts: new whitespace-prefix case.
- apps/daemon/tests/agents.test.ts: new env-isolation regression
asserting that OD_AGENT_HOME + NPM_CONFIG_PREFIX cannot leak the
real prefix bin into the sandbox.
* test(daemon): preserve $NPM_CONFIG_PREFIX across the env-isolation case (#614)
Address mrcfps's second-round review on PR #614: the env-isolation
regression sets `process.env.NPM_CONFIG_PREFIX = realPrefix` in its
body and then unconditionally `delete`s it in `finally`. On a developer
machine or CI runner that already exported `NPM_CONFIG_PREFIX`, that
mutates the worker-wide env for every later test, making downstream
env-sensitive assertions order-dependent.
Move the save/restore into the file's existing afterEach hook (mirroring
the OD_AGENT_HOME / OD_DAEMON_URL / OD_TOOL_TOKEN pattern) and drop the
in-test `delete`. Same coverage, no worker-state mutation.
* fix(platform): prioritise $NPM_CONFIG_PREFIX over the conventional npm guesses (#614)
Address mrcfps's third-round review on PR #614: when the user has
explicitly configured a prefix via $NPM_CONFIG_PREFIX (or
$npm_config_prefix), that's where `npm i -g` puts the *current*
binaries. The conventional guesses ~/.npm-global / ~/.npm-packages
often hold *stale* installs from an older prefix the user has since
rewritten — searching the env-driven prefix first matches npm's own
resolution order (env > .npmrc > default) and gives "explicit beats
convention" semantics.
Move the env-driven push above the conventional `dirs.push(.npm-global,
.npm-packages)`. Add a vitest case in the platform package that asserts
$NPM_CONFIG_PREFIX/bin's index in the result is strictly less than
~/.npm-global/bin's and ~/.npm-packages/bin's.
`resolveOnPath()` and the packaged PATH builder both preserve insertion
order, so first hit wins and the new ordering propagates to both
layers.
* fix(platform): lift $NPM_CONFIG_PREFIX above every conventional bin (#614)
Address mrcfps's fourth-round review on PR #614: the previous fix only
moved $NPM_CONFIG_PREFIX/bin ahead of ~/.npm-global / ~/.npm-packages,
but ~/.local/bin still appeared earlier in the array. Under a minimal
GUI-launch PATH a stale agent in ~/.local/bin (also a shared dumping
ground for pip --user / cargo install / hand-built binaries) could
outrank the user's *current* explicit npm prefix.
Move the env-driven push to the head of `dirs` so the explicit prefix
wins over every conventional location below — ~/.local/bin included.
Matches npm's own resolution order (env > .npmrc > default) across the
whole list, not just the npm-prefix block.
Tightened the existing order test to assert `explicitIdx === 0` and
that ~/.local/bin's index is strictly greater than the explicit
prefix's index, so a future drift would fail loudly.
* feat(media): add Nano Banana image provider
* fix(media): support Gemini API key headers for Nano Banana
* refactor(media): move Nano Banana model override flag into provider metadata
PerishCode