This brings the desktop memory worktree onto local main after main advanced with opener, IM, zoom, and autonomous goal work. The conflict resolution keeps both the selected-chat reference imports and memory event settings entrypoint in MessageList, and keeps both open-targets and memory API routes in the server router.
Constraint: main and feat/desktop-memory-surface diverged across desktop chat and server routing.
Constraint: desktop Vitest runs src-tauri tests under Vitest, so the Tauri config test cannot import bun:test.
Rejected: Fast-forward merge | main carried newer local commits not present on the memory branch.
Confidence: high
Scope-risk: moderate
Directive: Keep chat selection references, memory cards, open-target routes, and memory routes together when touching these files.
Tested: bun run check:desktop
Tested: bun run check:server
The /goal slash command was developed and verified in an isolated worktree, while local main has advanced with unrelated desktop fixes. Merge the feature branch rather than rebasing so both lines of local history remain intact.
Constraint: local main contains newer independent commits after the feature base
Rejected: Fast-forward merge | main is not an ancestor of the feature branch
Confidence: high
Scope-risk: moderate
Directive: Keep /goal continuation limits bounded and provider-agnostic when modifying the loop
Tested: bun test src/commands/headless.test.ts src/goals/goalState.test.ts src/goals/goalEvaluator.test.ts
Tested: bun run check:server
Tested: bun run verify
Tested: real DeepSeek v4 pro CLI and desktop /goal validation in /tmp business fixtures
Not-tested: push or remote PR creation
Tauri WebView was allowed to call the desktop sidecar API, but the CSP still blocked image loads from that same local server, so extracted IDE icons degraded to the generic code glyph in packaged builds. macOS detection also trusted command shims before app bundles, which let stale JetBrains Toolbox scripts keep removed IDEs visible.
Constraint: Opener icons are served by the per-run desktop sidecar on localhost or 127.0.0.1
Rejected: Bundle a fixed static icon set | the product should use detected installed application icons across platforms
Rejected: Trust macOS command shims for detection | stale shims survive app removal and make the menu lie
Confidence: high
Scope-risk: narrow
Directive: Keep macOS IDE detection anchored to real app bundles before accepting command shims
Tested: bun test src/server/__tests__/open-target-service.test.ts src/server/__tests__/open-target-api.test.ts desktop/src-tauri/tauri-config.test.ts
Tested: cd desktop && bun run test -- src/api/client.test.ts src/api/openTargets.test.ts src/stores/openTargetStore.test.ts src/components/layout/OpenProjectMenu.test.tsx src/components/layout/TabBar.test.tsx
Tested: cd desktop && bun run lint
Windows and Linux need to follow the same product rule as macOS: show icons from the user's installed applications instead of maintaining bundled brand artwork. The opener service now resolves command paths for Windows, walks from launcher scripts to the real IDE executable, extracts associated executable icons through the existing PNG endpoint, and resolves Linux .desktop Icon entries through common XDG application and icon theme locations.
Constraint: Do not redistribute third-party IDE trademark assets in the app bundle.
Constraint: Keep the frontend platform-agnostic by preserving the existing iconUrl contract.
Rejected: Add static per-IDE icon assets | impossible to keep complete and creates unnecessary trademark redistribution risk.
Confidence: medium
Scope-risk: moderate
Directive: Prefer platform installation metadata over bundled brand art when adding more opener targets.
Tested: bun test src/server/__tests__/open-target-service.test.ts src/server/__tests__/open-target-api.test.ts
Tested: cd desktop && bun run test -- src/stores/openTargetStore.test.ts src/components/layout/OpenProjectMenu.test.tsx src/components/layout/TabBar.test.tsx
Tested: cd desktop && bun run lint
Tested: bun run check:server
Not-tested: Native Windows and Linux VM smoke tests for real installed IDE icons
Some memory-only project directories have no session JSONL metadata, so the previous label recovery still fell back to reversing sanitizePath(). That reverse mapping is lossy for Chinese characters, spaces, and punctuation, leaving paths rendered with repeated slashes. The memory API now performs a bounded filesystem-guided recovery: it walks only roots whose sanitized prefix can still match the project id and only descends through candidate directories that preserve that prefix match.
Constraint: Do not change the existing ~/.claude/projects/<sanitized>/memory storage layout.
Rejected: Full-home recursive search | too expensive and unnecessary because sanitizePath prefix matching gives a tight traversal boundary.
Confidence: high
Scope-risk: narrow
Directive: Keep project ids as opaque storage keys; recover display labels from real metadata or existing directories first.
Tested: bun test src/server/__tests__/memory.test.ts
Tested: local server curl /api/memory/projects against the real PicTacticAgent legacy memory project
The first project opener UI used generic code glyphs for every IDE, which made the menu visually noisy and did not match the native desktop expectation. This changes the macOS path to expose transparent PNG icons generated from each detected local .app bundle and renders them directly in the toolbar/menu, falling back to the existing glyph only when an icon cannot be loaded.
Constraint: Do not redistribute third-party IDE trademark assets in the app bundle.
Constraint: Keep icon detection local to already detected open targets and cache only runtime PNG results.
Rejected: Bundling downloaded IDE logos | creates asset maintenance and trademark redistribution risk.
Rejected: Adding simple-icons as a dependency | it is not a faithful desktop app icon source and still requires brand permission checks.
Confidence: high
Scope-risk: narrow
Directive: Prefer local bundle icons on macOS; add curated official assets only as a cross-platform fallback layer.
Tested: bun test src/server/__tests__/open-target-service.test.ts src/server/__tests__/open-target-api.test.ts
Tested: cd desktop && bun run test -- src/stores/openTargetStore.test.ts src/components/layout/OpenProjectMenu.test.tsx src/components/layout/TabBar.test.tsx
Tested: cd desktop && bun run lint
Tested: bun -e openTargetService.getTargetIcon for vscode and finder returned PNG data
Tested: cd desktop && bun run build
Coding sessions need a native way to keep working until a stated completion condition is actually met. This adds a CLI /goal command, goal state and evaluator logic, and a headless-safe command filter so interactive and print-mode runs can both start a goal and continue when the evaluator says more work remains.
Constraint: The evaluator must work with saved third-party provider runtimes that may reject structured output schemas.
Constraint: Completion evidence must come from visible transcript text, not hidden goal prompts or assistant thinking blocks.
Rejected: Implement /goal as a bundled skill only | it would not own the query loop or support native continuation semantics.
Rejected: Expose all local-jsx commands to headless mode | interactive UI commands would become available in print mode.
Confidence: high
Scope-risk: moderate
Directive: Keep /goal evaluator evidence scoped to visible transcript text before adding richer completion signals.
Tested: bun test src/commands/headless.test.ts src/goals/goalState.test.ts src/goals/goalEvaluator.test.ts
Tested: bun run check:server
Tested: bun run verify
Tested: Real provider /goal smoke with MiniMax-M2.7-highspeed, including a two-turn continuation that logged Goal continuation #1 and completed with GOAL_SMOKE_LOOP_DONE.
Not-tested: Durable cross-process goal persistence for resumed sessions.
The CLI stores project directories with sanitizePath(), which intentionally replaces every non-alphanumeric byte with '-' for cross-platform safety. That storage key is lossy, so the desktop memory page must not treat it as the only source of display truth. The memory API now prefers the current cwd or the cwd/workDir/projectPath recorded in recent session JSONL files, and only falls back to the lossy reverse mapping when no metadata exists.
Constraint: Keep the existing ~/.claude/projects/<sanitized>/memory layout unchanged for CLI compatibility.
Rejected: Change sanitizePath to preserve Unicode | would break existing transcript and memory directory lookup semantics.
Confidence: high
Scope-risk: narrow
Directive: Treat project ids as storage keys, not user-facing path labels.
Tested: bun test src/server/__tests__/memory.test.ts
Tested: cd desktop && bunx vitest run src/__tests__/memorySettings.test.tsx
Tested: local server curl /api/memory/projects with a Chinese-space project path
Not-tested: bun run check:server still has two unrelated cron scheduler launcher timeout failures in src/server/__tests__/cron-scheduler-launcher.test.ts
Desktop sessions need a fast local escape hatch that opens the same materialized cwd the agent is editing, without showing unavailable IDE choices or persisting detection state. This adds a local open-targets API with silent in-memory detection for common IDEs and platform file managers, then wires a compact Codex-style toolbar menu into the desktop TabBar for active session workdirs.
Constraint: The first version is local IDE/editor and Finder/Explorer/file-manager only, no terminal targets or IDE plugin integration.
Constraint: The opened path must come from the active session workDir so isolated worktrees open the actual agent editing surface.
Rejected: Persisting detected applications | detection is cheap and temporary state avoids stale app inventory.
Rejected: Rendering unavailable IDEs as disabled menu rows | the user asked to show only detected targets and fall back to Finder/Explorer when no IDE is available.
Confidence: high
Scope-risk: moderate
Directive: Keep this path session-workdir based; do not switch it to repository root without proving isolated worktree behavior.
Tested: bun test src/server/__tests__/open-target-service.test.ts src/server/__tests__/open-target-api.test.ts
Tested: cd desktop && bun run test -- src/stores/openTargetStore.test.ts src/components/layout/OpenProjectMenu.test.tsx src/components/layout/TabBar.test.tsx
Tested: cd desktop && bun run lint
Tested: bun run check:desktop
Tested: bun test src/server/__tests__/h5-access-auth.test.ts -t 'allows local desktop H5 access settings under explicit server auth with a valid bearer'
Not-tested: bun run check:server full suite had one unrelated H5 auth integration timeout in the full concurrent run; the timed-out test passed when rerun alone.
The CLI already emits memory_saved events and stores Markdown memory files, but the desktop app had no usable surface for seeing or editing those writes. This adds a project memory API, a Settings memory editor, chat memory event cards, and routing from /memory or /context into the memory UI.
Constraint: Memory files live under Claude project storage and must remain plain Markdown editable by users.
Rejected: Only expose raw filesystem links | users need an in-app review and edit flow from chat.
Confidence: high
Scope-risk: moderate
Directive: Keep memory storage project-scoped and preserve unknown Markdown content when editing.
Tested: bun test src/server/__tests__/ws-memory-events.test.ts src/server/__tests__/memory.test.ts
Tested: bun run check:server
Tested: bun run check:desktop
Tested: agent-browser E2E for chat memory card, Open Memory navigation, and responsive Markdown editor layout
Not-tested: Live model auto-memory trigger rate with real provider credentials
Desktop search results were visually misleading when many directories shared the same basename, because the UI emphasized the basename and the server ranking allowed deep basename matches to outrank direct path-prefix matches. This makes search rows show the insertable relative path and ranks direct path-prefix matches before unrelated same-name directories.
Constraint: Desktop @ file suggestions should match the CLI mental model for path search.
Rejected: Keep basename plus parent-path rows | it hides the distinction between src/, src/hooks/, and desktop/src in the primary text.
Confidence: high
Scope-risk: narrow
Directive: Preserve path-prefix ranking ahead of basename-only matches for directory searches.
Tested: bun test src/server/__tests__/filesystem.test.ts
Tested: cd desktop && bun run test -- FileSearchMenu.test.tsx ChatInput.test.tsx
Tested: bun run check:server
Tested: bun run check:desktop
Desktop @ file search was recursively walking the filesystem with a local skip list, which let Python and Node generated directories leak into results and diverged from CLI behavior. Route candidate discovery through the same git-first model: tracked files, untracked files with exclude-standard, and ripgrep fallback for non-git folders, then derive selectable directories from those candidates.
Constraint: Desktop picker must select both files and directories without surfacing ignored project artifacts.
Constraint: No new dependencies; reuse the existing git, ripgrep, settings, and ignore utilities.
Rejected: Maintain a hardcoded directory denylist | it would drift from CLI and miss project-specific ignore rules.
Rejected: Full recursive readdir scanning | it ignores git index semantics and makes large dependency trees visible.
Confidence: high
Scope-risk: moderate
Directive: Keep desktop @ file candidate discovery aligned with src/hooks/fileSuggestions.ts before changing ranking or ignore behavior.
Tested: bun test src/server/__tests__/filesystem.test.ts
Tested: cd desktop && bun run test -- FileSearchMenu.test.tsx ChatInput.test.tsx
Tested: cd desktop && bun run lint && bun run test -- --run && bun run build
Tested: bun run check:server
Tested: bun run check:coverage
Tested: bun run verify
Not-tested: Native Windows/Linux manual UI smoke; path handling relies on cross-platform git/ripgrep wrappers and normalized relative paths.
The pure white appearance option needed to avoid warm-theme leakage while preserving the existing warm classic brand theme. This adds the white theme mode, keeps local browser startup from reusing stale H5 server URLs in dev, and moves visible legacy warm surfaces onto theme tokens.
Constraint: H5 server auth policy, CORS policy, SDK routes, adapter routes, and IM access paths must not change for a visual theme fix
Rejected: Rename the original light theme to pure white | the original theme is a warm classic palette, not a neutral white workspace
Confidence: high
Scope-risk: moderate
Directive: Keep structural white-theme borders neutral; reserve the warm brand color for selected states, primary actions, and small accents
Tested: cd desktop && bun run test -- src/__tests__/generalSettings.test.tsx src/lib/desktopRuntime.test.ts src/lib/persistenceMigrations.test.ts src/stores/uiStore.test.ts src/stores/settingsStore.test.ts
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run build
Tested: bun test src/server/__tests__/settings.test.ts src/server/__tests__/h5-access-policy.test.ts src/server/__tests__/h5-access-auth.test.ts src/server/middleware/cors.test.ts
Tested: Browser smoke at http://127.0.0.1:5173 with data-theme=white, no H5 token prompt, /status inspector visible, inspector border #DDE3EA
Not-tested: Full bun run verify gate
Non-streaming adapter only read `output_tokens`/`input_tokens` (Anthropic-style),
but many OpenAI-compatible providers (e.g. Jiutian MoMA) return standard
`completion_tokens`/`prompt_tokens` fields. This caused a null reference
error when reading `output_tokens` from such responses.
The streaming adapter already handles `completion_tokens` correctly, and
the Azure adapter uses the same fallback pattern applied here.
Fixes#383
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The H5 browser shell can load while deeper Settings tabs still exercise separate API routes. This adds MCP, plugin, and Agents endpoints to the H5 policy and integration matrices so the QR-token path is protected against page-level regressions.\n\nConstraint: H5 must expose existing desktop settings surfaces only after local opt-in and a valid H5 token.\nRejected: Rely on manual browser clicks alone | they missed the stopped-port failure mode and would not guard future regressions.\nConfidence: high\nScope-risk: narrow\nDirective: Add new Settings API surfaces to this H5 matrix when exposing them in the browser UI.\nTested: bun test src/server/__tests__/h5-access-policy.test.ts src/server/__tests__/h5-access-auth.test.ts\nTested: bun run check:server
The desktop sidecar can bind on LAN addresses for phone access, so the H5 settings switch must be an authorization boundary for remote capability routes, not only a token-mode toggle. Remote browser API, proxy, websocket, and SDK routes now fail closed while H5 is disabled; local desktop, Tauri, WebUI, adapter, and internal SDK paths remain tokenless. When H5 is enabled, remote API, proxy, and websocket requests must use the H5 token carried by the QR link, and the server API key cannot substitute for that H5 token.
Constraint: Desktop sidecar binds 0.0.0.0 while reporting loopback to local UI.
Constraint: Client-controlled Host and Origin headers cannot prove a local request; the boundary uses Bun requestIP instead.
Constraint: Static H5 shell and /health must still load so browser bootstrap can show a recovery flow.
Rejected: Trust loopback Host headers | LAN clients can spoof Host and Origin.
Rejected: Use ANTHROPIC_API_KEY as a remote H5 credential | it is not the phone pairing token and would weaken the QR-token boundary.
Confidence: high
Scope-risk: moderate
Directive: Do not make h5Enabled=false an open remote state for /api, /proxy, /ws, or /sdk routes.
Tested: bun test src/server/__tests__/h5-access-policy.test.ts src/server/__tests__/h5-access-auth.test.ts src/server/middleware/cors.test.ts
Tested: bun run check:server
Remote browser access now needs the H5 token once LAN mode is enabled, but changing that mode remains a local desktop control-plane operation even when deployment auth is explicitly enabled. CORS also no longer treats a non-local same-origin alias as trusted unless it is the configured H5 public origin or allowlisted origin.
Constraint: H5 exposes desktop control surfaces on the LAN and must not let remote browsers enable their own access.
Constraint: Desktop WebView, localhost WebUI, SDK websocket, and IM adapter loopback calls must remain tokenless.
Rejected: Allow authenticated remote control-plane writes under SERVER_AUTH_REQUIRED | violates the local opt-in boundary for H5 exposure.
Confidence: high
Scope-risk: moderate
Directive: Do not relax /api/h5-access mutating endpoints for browser origins without adding a reviewed desktop-only trust signal.
Tested: bun test src/server/__tests__/h5-access-auth.test.ts
Tested: bun test src/server/__tests__/h5-access-policy.test.ts src/server/middleware/cors.test.ts
Tested: bun run check:server
Wire request-scoped H5 auth and CORS decisions through the server so LAN browser API, proxy, and websocket access require the H5 token while loopback desktop, Tauri WebView, local adapter, SDK, and static shell paths keep their existing bootstrap behavior.
Constraint: Task 2 scope is limited to server auth/CORS wiring and the four requested files
Rejected: Global auth toggle for all requests | would incorrectly block loopback desktop and bootstrap asset flows
Directive: Keep static H5 shell and assets tokenless unless the browser-side bootstrap model changes
Confidence: high
Scope-risk: narrow
Tested: bun test src/server/__tests__/h5-access-policy.test.ts src/server/middleware/cors.test.ts
Tested: bun test src/server/__tests__/h5-access-policy.test.ts src/server/__tests__/h5-access-auth.test.ts src/server/middleware/cors.test.ts | h5-access-auth blocked by missing zod dependency in src/server/api/haha-oauth.ts
Tested: LSP diagnostics clean for src/server/index.ts, src/server/middleware/cors.ts, src/server/__tests__/h5-access-auth.test.ts, src/server/middleware/cors.test.ts
Not-tested: h5-access-auth runtime assertions because this checkout is missing zod for haha-oauth imports
The classifier should not treat every /sdk path as trusted by pathname, because LAN-exposed servers already have a narrower local SDK trust boundary.
Constraint: Desktop local chat, Tauri WebView, loopback adapters, and local SDK routes must remain tokenless.
Rejected: Path-only /sdk trust | a non-local caller could reach the SDK route through a LAN-bound server.
Confidence: high
Scope-risk: narrow
Directive: Keep remote /sdk and hostile-origin loopback requests in the H5/browser bucket unless a dedicated SDK auth layer is verified.
Tested: bun test src/server/__tests__/h5-access-policy.test.ts
Add a small standalone request classifier and focused server tests so later H5 token enforcement can target LAN/browser requests without changing the current desktop, SDK, or adapter paths.
Constraint: Task scope is limited to a standalone classifier plus tests in two server files
Constraint: Local desktop WebView, loopback integrations, and /sdk routes must remain tokenless for now
Rejected: Integrate the classifier into auth middleware now | this task only introduces the reusable classification seam
Confidence: high
Scope-risk: narrow
Directive: Keep local-trusted and internal-sdk classifications tokenless unless the server auth integration changes with matching regression coverage
Tested: bun test src/server/__tests__/h5-access-policy.test.ts
Tested: per-file TypeScript diagnostics for src/server/h5AccessPolicy.ts and src/server/__tests__/h5-access-policy.test.ts
Not-tested: Full server auth middleware integration
Not-tested: bun run check:server remains red from pre-existing repo dependency and unrelated test failures
The sidebar had only single-session deletion, which made stale or noisy
session lists expensive to maintain. This adds a batch-management lane in
the desktop UI and a server endpoint that deletes multiple sessions while
preserving per-session failure reporting and cleanup behavior.
Constraint: Session deletion must preserve existing transcript and adapter cleanup semantics
Constraint: Desktop UI should reuse the shared confirmation dialog instead of introducing a second modal surface
Rejected: Delete all selected sessions through repeated client DELETE calls | weaker partial-failure handling and duplicated cleanup orchestration
Rejected: Use a generic checkmark entry icon | it did not communicate batch deletion clearly enough
Confidence: high
Scope-risk: moderate
Directive: Keep batch deletion routed through the server batch endpoint so rollback and adapter cleanup stay centralized
Tested: bun test src/server/__tests__/sessions.test.ts -t batch-delete
Tested: bun run check:server
Tested: cd desktop && bun run check:desktop
Tested: browser UI smoke for group select, shift range select, delete confirmation, 7/30 day cleanup, Cmd+A filtered selection, and Escape exit
Not-tested: Live provider-backed session generation; this change covers session list management after sessions already exist
The H5 token gate still caused chat startup and runtime failures after the desktop startup hotfix, so the default server path now stays open for browser and LAN access. Explicit auth remains available through SERVER_AUTH_REQUIRED=1 or --auth-required for deployments that intentionally need it.
Constraint: Current H5 token state is blocking active users from normal chat usage
Rejected: Keep automatic remote-host auth with broader client-side token handling | still leaves existing upgraded clients vulnerable to stale or missing token state
Confidence: high
Scope-risk: moderate
Directive: Do not re-enable default H5 token auth without a migration and end-to-end browser chat verification
Tested: bun test src/server/__tests__/h5-access-auth.test.ts src/server/middleware/cors.test.ts
Tested: cd desktop && bun run test src/lib/desktopRuntime.test.ts src/__tests__/generalSettings.test.tsx
Tested: bun run check:server
Tested: bun run check:desktop
The H5 LAN release made the desktop sidecar bind broadly and require auth for non-localhost hosts. The Tauri WebView still talks to the loopback control URL without an H5 bearer token, so startup requests were rejected before the app could load.
Constraint: Desktop WebView control requests must stay tokenless for the local sidecar.
Rejected: Reuse H5 token bootstrap for Tauri | would add remote-access state to the native app startup path.
Confidence: high
Scope-risk: narrow
Directive: Tauri WebView bypass is limited to loopback control hosts; do not extend it to LAN or public hosts without re-reviewing H5 auth boundaries.
Tested: bun test src/server/__tests__/h5-access-auth.test.ts
Tested: bun run check:server
The detached worktree fix corrects stale private-network H5 public URLs after the desktop server chooses a new ephemeral LAN port. Merge it into local main while leaving unrelated local release edits untouched.
Constraint: Local main already had unrelated unstaged desktop release metadata edits.
Rejected: Broad staging before the merge | would risk committing unrelated local release work.
Confidence: high
Scope-risk: narrow
Tested: bun test src/server/__tests__/h5-access-service.test.ts
Tested: bun run check:server
Not-tested: physical phone browser on local WiFi after merge
H5 access could keep returning a previously saved private-network URL even after the packaged desktop server moved to a new ephemeral LAN port. The public settings now prefer the current auto-detected LAN URL for local/private stored addresses while preserving explicit public reverse-proxy URLs.
Constraint: Packaged desktop H5 access uses an ephemeral server port and auto-publishes a LAN URL for phones on the same network.
Rejected: Always overwrite stored publicBaseUrl | public reverse-proxy deployments still need manual stable URLs.
Confidence: high
Scope-risk: narrow
Tested: bun test src/server/__tests__/h5-access-service.test.ts
Tested: bun test src/server/__tests__/h5-access-service.test.ts src/server/__tests__/h5-access-api.test.ts src/server/__tests__/h5-access-auth.test.ts
Tested: bun run check:server
Tested: manual LAN smoke with 0.0.0.0:5188 correcting stale 192.168.0.102:5179 to 192.168.0.102:5188
Not-tested: physical phone browser on local WiFi
The previous fix compared the venv's pyvenv.cfg `home` field with the
custom interpreter's directory, but Python's venv module skips the
wrapping venv when the source interpreter is itself in a venv (conda /
pyenv / hand-built venvs — which is the common case). The recorded
`home` then points to the base Python install, not the user-provided
path, causing the comparison to permanently fail and the UI to keep
showing "venv not ready" even right after a successful rebuild.
Switch to a marker file (~/.claude/.runtime/venv-base-interpreter.txt)
that records the exact `config.pythonPath` used at venv creation time.
Status check compares it verbatim with the current config. Falls back
to "match if both empty" when the marker is missing, which preserves
behavior for legacy users who never set a custom path.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The dependency check ignored the custom Python path: it only verified
the global venv's stamp file, so switching to a fresh custom interpreter
still showed deps as installed even though the new interpreter's
environment was empty. Setup also reused the old venv built from a
different base interpreter.
Verify the existing venv via pyvenv.cfg's home field, but only when a
custom path is configured — users without a custom path keep the
original code path untouched, protecting existing installations.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Packaged desktop builds started the sidecar on loopback only, so phones on the same WiFi could not connect to the H5 URL. The server now binds to LAN interfaces while keeping the desktop control URL on loopback, serves the bundled H5 frontend, and reports a LAN public URL when H5 access is enabled.
Constraint: Packaged Tauri resources place ../dist under Contents/Resources/_up_/dist.
Rejected: Bind the desktop control URL to 0.0.0.0 | desktop clients should keep using loopback for local control.
Confidence: high
Scope-risk: moderate
Directive: Keep packaged H5 resource lookup aligned with Tauri resource mapping when changing bundle resources.
Tested: bun run check:server
Tested: cd desktop/src-tauri && cargo test --lib
Tested: cd desktop && bun run test -- desktopRuntime.test.ts generalSettings.test.tsx
Tested: SKIP_INSTALL=1 ./desktop/scripts/build-macos-arm64.sh
Tested: Packaged claude-sidecar smoke served H5 HTML while listening on *:39876
Not-tested: Physical phone browser test on the local WiFi network
Providers like DeepSeek with 1M context windows can report
input_tokens that, when combined with output_tokens, exceed the
declared context window — causing the UI to show >100% usage.
Add an optional contextWindow parameter to calculateCurrentContextTokenTotal
that clamps the result to the window size. Pass it from analyzeContext
and sessionService where the window size is already available.
Backward compatible: callers that omit the parameter keep existing behavior.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Continuing an already materialized worktree session should behave like a normal existing session. The launch path now checks whether the session has already moved into its worktree before showing the startup bubble or passing worktree creation flags again.
Constraint: Existing session metadata can retain repository.worktree after the worktree transcript becomes the active session file
Rejected: Frontend-only suppression | the server could still launch the CLI with duplicate worktree creation flags
Confidence: high
Scope-risk: narrow
Tested: bun run check:server
Tested: bun test src/server/__tests__/conversations.test.ts -t worktree --timeout 30000
Tested: bun test src/server/__tests__/sessions.test.ts -t "newest duplicate session file" --timeout 30000
Not-tested: bun run verify still fails on unrelated desktop Settings.tsx and MessageList.tsx gates from the broader worktree
Integrate the Token usage settings page on top of the newer local main branch, preserving the existing H5 access router surface while adding the activity stats API route.
Constraint: Local main already contained H5 access router changes after the Token usage worktree base.
Rejected: Prefer either router branch during conflict resolution | both API surfaces are independent and must remain registered.
Confidence: high
Scope-risk: narrow
Directive: Keep h5-access and activity-stats routes registered together in src/server/router.ts.
Tested: Conflict marker scan for src/server/router.ts
Not-tested: Post-merge desktop/server checks pending
Expose local Claude Code CLI transcript usage in Settings so users can inspect recent token consumption and daily activity without leaving the desktop app.
The page uses server-side transcript aggregation for session, message, tool, model-token, and subagent token data. Daily token buckets use assistant message timestamps, and daily session counts use active parent sessions for the same date bucket so resumed sessions and cross-midnight work do not produce token-only days. Cache accounting is bumped to v5 to force recomputation under the corrected daily semantics.
Constraint: Usage data must come from local Claude Code CLI transcripts rather than mock/demo data.
Constraint: Desktop navigation keeps Token usage directly above Diagnostics.
Rejected: Bucket all token usage by session start date | hides resumed-session and cross-midnight consumption from the actual day it was spent.
Confidence: high
Scope-risk: moderate
Directive: Keep daily token and daily session counts on the same date-bucketing semantics.
Tested: bun run check:desktop
Tested: bun run check:server
Tested: Browser verification for Token usage in English and Chinese locale date labels
Not-tested: Full bun run verify quality gate
Merge the H5 access branch into local main so the desktop app can expose an opt-in browser chat surface with token-gated remote access, mobile chat layout, and full-width bottom sheets for mobile pickers. The merge keeps desktop and Tauri behavior behind runtime checks while preserving local WebUI development bypasses for loopback and private-LAN server targets.
Constraint: H5 is opt-in and local browser development must not require a desktop-generated token
Rejected: Cherry-pick only the final UI commit | it omitted earlier auth, CORS, settings, and H5 token service dependencies
Confidence: high
Scope-risk: moderate
Directive: Keep mobile-only layout branches guarded by browser viewport plus non-Tauri runtime checks
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run test -- src/__tests__/generalSettings.test.tsx src/__tests__/pages.test.tsx src/api/client.test.ts src/api/websocket.test.ts src/components/chat/ChatInput.test.tsx src/components/controls/PermissionModeSelector.test.tsx src/components/layout/AppShell.test.tsx src/components/layout/Sidebar.test.tsx src/components/shared/RepositoryLaunchControls.test.tsx src/hooks/useMobileViewport.test.tsx src/lib/desktopRuntime.test.ts src/pages/ActiveSession.test.tsx src/pages/EmptySession.test.tsx src/stores/settingsStore.test.ts
Tested: bun test src/server/__tests__/h5-access-api.test.ts src/server/__tests__/h5-access-auth.test.ts src/server/__tests__/h5-access-service.test.ts src/server/middleware/cors.test.ts
Tested: git diff --check HEAD^ HEAD
Not-tested: bun run verify full quality gate
Not-tested: mobile Chrome smoke was not rerun after merging into local main
H5 browser use needs to keep local desktop WebUI testing open while giving remote phones a usable chat-first surface. This change scopes browser auth by origin and target host, hides non-chat desktop chrome on H5 mobile, compacts the empty-session composer, and unifies mobile selectors behind a full-width bottom sheet with an explicit close action.
Constraint: Local browser development may bypass H5 auth only for loopback or private LAN server hosts; public or remote hosts still require an H5 token.
Constraint: Mobile H5 scope is chat-first; settings and scheduled-task surfaces stay hidden instead of fully redesigned.
Rejected: Reuse desktop popovers on mobile | hover/anchored dropdown behavior leaves key actions hard to dismiss and cramped on narrow screens.
Rejected: Let any localhost origin bypass remote auth | it would let arbitrary local pages access public H5 servers without a token.
Confidence: medium
Scope-risk: moderate
Directive: Keep future H5 selector popups on MobileBottomSheet unless a selector needs a reviewed custom mobile interaction.
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run test -- src/pages/EmptySession.test.tsx src/components/chat/ChatInput.test.tsx src/lib/desktopRuntime.test.ts src/components/shared/RepositoryLaunchControls.test.tsx src/components/controls/PermissionModeSelector.test.tsx src/components/controls/ModelSelector.test.tsx src/components/shared/DirectoryPicker.test.tsx src/__tests__/pages.test.tsx
Tested: bun test src/server/__tests__/h5-access-auth.test.ts src/server/middleware/cors.test.ts
Tested: Chrome mobile viewport smoke for context, model, permission, project, branch, and worktree bottom sheets with no horizontal overflow.
Not-tested: Full bun run verify quality gate.
Some Anthropic-compatible providers can return successful responses with
missing or partial usage metadata, especially after falling back to
non-streaming requests. Normalize usage at the provider response boundary so
transcript writes, side queries, and cost tracking continue without treating
token accounting as required response content.
Constraint: Third-party Anthropic-compatible providers do not always return complete usage fields.
Rejected: Add provider-specific handling | this is a protocol compatibility issue across proxies and gateways
Confidence: high
Scope-risk: moderate
Directive: Keep provider response normalization vendor-neutral; do not branch on provider names for missing usage fields.
Tested: bun test src/services/api/emptyUsage.test.ts src/utils/__tests__/thinking.test.ts src/server/__tests__/title-service.test.ts src/server/__tests__/conversations.test.ts
Tested: bun run check:server
Tested: bun run check:persistence-upgrade
Tested: git diff --check
Not-tested: Live Shengsuan Cloud provider replay
Remote browser access needs an explicit path that preserves localhost desktop behavior while allowing H5 tokens for REST and client websocket upgrades. This keeps the existing synchronous Anthropic auth helper intact, adds async request validation for H5 tokens, and centralizes origin handling so preflight, API, proxy, websocket, and health responses behave consistently.
Constraint: Existing validateAuth(req) must stay synchronous for current tests and callers
Constraint: Default localhost and Tauri desktop behavior must remain unchanged
Rejected: Replacing validateAuth with an async-only API | would break existing sync tests and call sites
Rejected: Allowing wildcard or fallback remote origins | violates the H5 explicit-origin security model
Directive: Restore ProviderService serverPort in server integration tests that call startServer with custom ports to avoid leaking global proxy runtime state
Confidence: high
Scope-risk: moderate
Tested: bun test src/server/__tests__/h5-access-auth.test.ts --timeout 30000
Tested: bun test src/server/__tests__/h5-access-service.test.ts src/server/__tests__/h5-access-api.test.ts src/server/__tests__/settings.test.ts --timeout 30000
Tested: bun test src/server/__tests__/h5-access-auth.test.ts src/server/__tests__/conversation-service.test.ts src/server/__tests__/cron-scheduler-launcher.test.ts --timeout 30000
Tested: bun test src/server/__tests__/e2e/full-flow.test.ts --timeout 30000
Not-tested: Full non-live server gate end-to-end; an attempted broader run was interrupted after uncovering and then fixing a test-side ProviderService port leak
Move cc-haha managed settings writes behind one shared transaction helper so
H5 access mutations and provider settings sync serialize against the same file.
Also treat persisted H5 state without a valid token hash as disabled so origin
allowlists cannot remain active after partial corruption.
Constraint: H5 and provider config both share ~/.claude/cc-haha/settings.json
Rejected: Keep per-service locks | concurrent read-modify-write still drops unrelated fields
Rejected: Only harden isOriginAllowed | corrupted enabled state would still surface as active in settings reads
Confidence: high
Scope-risk: narrow
Directive: Any future writer of cc-haha/settings.json must go through the shared managed settings transaction path
Tested: bun test src/server/__tests__/h5-access-service.test.ts src/server/__tests__/h5-access-api.test.ts src/server/__tests__/providers.test.ts src/server/__tests__/provider-presets.test.ts
Not-tested: Full repo typecheck remains blocked by the existing bun-types/tsconfig environment issue in this worktree
Persist H5 access state in cc-haha managed settings, expose a narrow server API
for enable/disable/regenerate/verify flows, and keep token responses sanitized
so the raw secret is only returned at generation time.
Constraint: H5 config must live in ~/.claude/cc-haha/settings.json and preserve unknown fields
Rejected: Store raw token for later display | violates hash-only persistence requirement
Confidence: high
Scope-risk: narrow
Directive: Do not move H5 settings into ~/.claude/settings.json or expose tokenHash through the API
Tested: bun test src/server/__tests__/h5-access-service.test.ts src/server/__tests__/h5-access-api.test.ts
Not-tested: Full repo typecheck via tsc is blocked locally by missing bun-types and a TS 6 baseUrl deprecation in the current config
Bring the detached worktree fix onto local main after verifying the title-generation regression path and local quality checks.
Constraint: The worktree commit was based on an older main commit, while local main already contained later chat scroll fixes.
Confidence: high
Scope-risk: moderate
Directive: Keep title sanitation shared across CLI, server, and desktop read paths.
Tested: bun run check:server
Tested: bun run check:desktop
Tested: bun run verify (7 passed, 1 failed on existing aggregate agent-utils coverage baseline)
Not-tested: live model title-generation smoke
Slash-command and skill prompts can enter the title-generation path as internal XML breadcrumbs. The async title request previously treated that transport metadata as user prose, so a generated ai-title could persist raw command tags and override the normal session title.
This routes title sources and generated title output through a shared sanitizer, preserving user-visible command names and arguments while dropping unrelated internal XML metadata. Desktop fallback title rendering now applies the same cleanup so existing transcripts with bad ai-title entries recover on read.
Constraint: Session titles can be produced by both desktop server logic and the SDK generate_session_title control request.
Rejected: Only clean desktop session display | leaves CLI and remote title generation able to persist bad titles again
Confidence: high
Scope-risk: moderate
Directive: Keep generated-title input and persisted-title readback on the shared sanitizer; do not add a title path that reads command XML directly.
Tested: bun test src/utils/__tests__/sessionTitle.test.ts src/utils/__tests__/sessionTitleText.test.ts
Tested: bun run check:server
Tested: bun run check:desktop
Tested: bun run verify (7 passed, 1 failed on existing aggregate agent-utils coverage baseline)
Not-tested: live model title-generation smoke
Bring the detached worktree fix into the local main line while preserving the newer main behavior that passes session context into interactive chat message blocks.
Constraint: Local main already contains six unpublished commits on top of origin/main
Constraint: Merge conflict only affected MessageList message rendering around sessionId propagation
Rejected: Overwrite main's MessageBlock call shape | would regress AskUserQuestion and permission interactions that need sessionId
Confidence: high
Scope-risk: moderate
Directive: Keep MessageBlock sessionId propagation when editing chat rendering; AskUserQuestion depends on it
Tested: cd desktop && bun run test -- MessageList.test.tsx
Not-tested: Full verify after merge; previous verify remains blocked by agent-utils global coverage baseline
Chat sessions need predictable navigation when users switch tabs or read history during streaming. This preserves each session's scroll position, defaults fresh sessions to the latest message, and adds a compact jump-to-latest affordance when auto-follow is paused.
The PR gate also exposed that provider-scoped scheduled tasks must force the sdk-cli entrypoint when launched through the sidecar, so the same change records that runtime contract and its regression assertion.
Constraint: Desktop chat should not force-scroll while the user is reading older messages
Constraint: Scheduled task provider env must not inherit stale parent model runtime values
Rejected: Persist scroll positions in localStorage | session scroll is transient UI state and should not survive app restarts
Confidence: high
Scope-risk: moderate
Directive: Do not remove the sdk-cli entrypoint marker from provider-scoped cron tasks without rerunning the sidecar launcher regression
Tested: cd desktop && bun run test -- MessageList.test.tsx
Tested: cd desktop && bun run lint
Tested: bun test src/server/__tests__/cron-scheduler-launcher.test.ts src/utils/__tests__/themeWords.test.ts
Tested: bun run check:desktop
Tested: bun run check:server
Tested: bun run check:native
Not-tested: bun run verify still fails on pre-existing agent-utils global coverage baseline below threshold
Not-tested: Chrome extension E2E blocked by Codex Chrome Extension communication timeout after plugin diagnostics passed
Scheduled task subprocesses should normalize their child environment instead of inheriting a stale parent entrypoint. Setting the default task entrypoint keeps launcher tests and non-OAuth task execution on the SDK CLI path while still allowing the official OAuth environment overlay to provide its desktop-specific entrypoint.
Constraint: Official OAuth tasks intentionally overlay provider-specific environment after the default child task environment is built
Rejected: Preserve parent CLAUDE_CODE_ENTRYPOINT | stale parent values can leak into scheduled task launches
Confidence: high
Scope-risk: narrow
Directive: Keep default task launcher environment deterministic before provider-specific overlays are applied
Tested: bun test src/server/__tests__/cron-scheduler-launcher.test.ts
Tested: bun run check:server
Not-tested: Full verify still blocked by existing agent-utils global coverage baseline
Agent definitions are cached across command invocations, so edits made through the shared file helper must invalidate both the agent definition cache and the underlying markdown file cache. This keeps the TUI creation path and /agents command output consistent after creating, updating, or deleting custom agents.
Constraint: The TUI agent wizard and /agents command share cached agent definition loaders
Rejected: Clear only getAgentDefinitionsWithOverrides | the nested markdown file memoization can still return stale files
Confidence: high
Scope-risk: narrow
Directive: Any future agent file mutation path must call clearAgentDefinitionsCache after a successful write or delete
Tested: bun test src/tools/AgentTool/loadAgentsDir.cache.test.ts
Tested: bun run check:server
Not-tested: Full verify still blocked by existing agent-utils global coverage baseline
Computer Use setup could fail on machines where PATH discovery misses a valid Python installation, especially Windows or conda-style environments. Store an optional interpreter path, prefer it during environment checks and venv creation, and expose a desktop settings control for selecting or clearing it.
Constraint: Python discovery is environment-specific and cannot always be inferred from PATH.
Rejected: Continue falling back to PATH after an invalid custom path | hides a user-selected broken interpreter and makes diagnosis ambiguous.
Confidence: high
Scope-risk: moderate
Directive: Preserve unknown Computer Use config fields and keep blank interpreter paths normalized to automatic detection.
Tested: bun test src/utils/computerUse/preauthorizedConfig.test.ts src/server/__tests__/computer-use-python.test.ts src/server/__tests__/computer-use-api.test.ts
Tested: cd desktop && bun run test src/pages/ComputerUseSettings.test.tsx
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run build
Tested: Computer Use browser smoke saved /opt/homebrew/bin/python3 and backend persisted then reset pythonPath to null
Not-tested: Full bun run check:server; existing cron-scheduler-launcher test expects CLAUDE_CODE_ENTRYPOINT=sdk-cli but received undefined.
Related: https://github.com/NanmiCoder/cc-haha/issues/331
- Remove pr-checks lane from quality gate (not needed locally)
- Fix launcherRouting tests to pass explicit null envAppRoot to avoid CLAUDE_APP_ROOT pollution
- Fix cron-scheduler-launcher test: CLAUDE_CODE_ENTRYPOINT is correctly set to sdk-cli for scheduled tasks, update assertion and restore env var in cleanup
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
CLI transcripts persist native worktree-state entries for resume, so the desktop git-info endpoint should treat that state as authoritative for worktree identity. Desktop launch metadata can be stale after placeholder cleanup or reopen, and should only fill gaps.
The displayed business branch still prefers the desktop launch branch because CLI originalBranch means the source checkout before worktree creation, not necessarily the selected base ref.
Constraint: CLI worktree-state carries runtime identity; Desktop repository metadata carries launch intent.
Rejected: Let desktop repository metadata override worktree-state | stale placeholder metadata can resurrect the wrong slug, path, or source cwd after reload.
Confidence: high
Scope-risk: narrow
Directive: Worktree identity fields should prefer CLI worktree-state; only branch display may prefer desktop launch intent.
Tested: bun test src/server/__tests__/sessions.test.ts --test-name-pattern worktree-state|worktree identity|git-info
Tested: cd desktop && bun run lint
Tested: git diff --check
Desktop sessions already persist repository launch metadata, but CLI worktree sessions also write native worktree-state entries to the JSONL transcript for resume. Relying only on desktop session-meta leaves reopened or older transcripts vulnerable to losing the footer worktree identity when in-memory state is gone.
This teaches SessionService to recover the latest CLI worktree-state entry and lets git-info use it as a fallback for worktree badge fields. Desktop repository metadata still wins when present because it carries the selected business branch, while the CLI state keeps reopened sessions stable.
Constraint: CLI records worktree state as transcript metadata, not only as cwd.
Rejected: Infer worktree identity only from .claude/worktrees path | path heuristics miss hook-based worktrees and do not carry original cwd or slug.
Confidence: high
Scope-risk: narrow
Directive: Prefer explicit transcript metadata over path guessing for worktree session identity.
Tested: bun test src/server/__tests__/sessions.test.ts --test-name-pattern CLI worktree state
Tested: bun test src/server/__tests__/sessions.test.ts --test-name-pattern worktree-state|worktree identity|git-info
Tested: cd desktop && bun run lint
Tested: git diff --check
