Three issues reported on Windows + one user-requested limit bump:
1. Dashboard CPU/RAM/Network all at 0
handleSystemMetrics read /proc/* exclusively. Replaced with a
platform-split:
- metrics_unix.go (!windows): existing /proc reading code.
- metrics_windows.go: kernel32!GetSystemTimes for CPU
(delta of idle vs kernel+user FILETIMEs) and
kernel32!GlobalMemoryStatusEx for memory. Network left at zero
for now — MIB_IF_ROW2 is too version-sensitive to parse by hand.
handlers_info.go::handleSystemMetrics reduced to one delegating
call.
2. Terminal black screen on Windows
creack/pty/v2 returns "unsupported" on Windows; the v0.7.1 pipe
fallback works but pipes don't carry TTY signals, so cmd/pwsh/wsl
go silent. Implemented native ConPTY:
- terminal_conpty_windows.go: CreatePseudoConsole + STARTUPINFOEX
+ PROC_THREAD_ATTRIBUTE_PSEUDOCONSOLE wiring via
windows.NewProcThreadAttributeList. CreateProcessW launches
child with the PC attached, full ANSI / line discipline /
resize.
- canUseConPTY() probes once at startup (Win10 1809+ check).
- Restructure: terminal_session.go now holds just the interface
+ ptySession + pipeSession structs. terminal_session_unix.go
wires creack/pty. terminal_session_windows.go tries ConPTY
first, falls back to pipeSession.
3. Agent stops after 15 tool calls
MaxToolIterations bumped 15 → 500. Doc comment explains why the
cap exists at all (infinite-loop safety) and that 500 is well
above realistic usage.
- internal/version/version.go: 0.7.5 → 0.7.6
- CHANGELOG.md: v0.7.6 entry covers the three fixes
Three Windows install/launch issues reported by the user:
1. Double-click on Desktop shortcut → dialog "This is a command line
tool. You need to open cmd.exe and run it from there."
Cause: charmbracelet/huh detects no TTY when launched via Explorer
and aborts. Fix:
- cmd/muyue/commands/root.go: skip RunFirstTimeSetup when
os.Stdin is not a character device; persist config.Default()
and let the React onboarding wizard handle first-run UX.
- ci-{main,develop}.yml: build Windows binaries with
-ldflags="-H=windowsgui" so the .exe is a GUI subsystem app —
no console window flashes on double-click.
2. CLI sub-commands (`muyue scan`, `muyue install-shortcuts`, etc.)
would lose all output under -H=windowsgui when launched from
cmd.exe / PowerShell. Mitigation:
- cmd/muyue/console_windows.go (new, build-tagged): on init(),
call kernel32!AttachConsole(ATTACH_PARENT_PROCESS). If the
parent has a console, rebind os.Stdout/os.Stderr/os.Stdin to
it and call log.SetOutput(os.Stderr) so existing log.Printf
calls surface. If no parent console (Explorer), exit silently.
3. After install, `muyue` not recognized in PowerShell.
Causes: (a) the extracted binary is muyue-windows-amd64.exe, not
muyue.exe; (b) the user PATH update by install-shortcuts doesn't
propagate to the existing PowerShell session.
Fix in install-shortcuts:
- Copy self to <installDir>/muyue.exe (rename impossible — the
running .exe is locked on Windows) so `muyue` resolves once
PATH is set.
- Update Desktop + Start Menu .lnk to target the canonical
muyue.exe rather than the platform-suffixed binary.
- Print the line `$env:Path += ';<installDir>'` for the user to
paste, refreshing the current session immediately.
- ci-main.yml install snippet bumps to 5 lines, last being
`$env:Path += ";$dest"`.
- internal/version/version.go: 0.7.4 → 0.7.5
- CHANGELOG.md: v0.7.5 entry covers all three fixes
Logo dropped at project root by user. Bake it everywhere it matters:
Assets:
- assets/muyue.ico — multi-res (16/24/32/48/64/128/256) generated via PIL
- assets/muyue-{16,32,64,128,256,512}.png — clean PNG resizes
- LogoMuyue.png kept at root as the source of truth
Windows binary (.exe):
- CI runs `rsrc -ico assets/muyue.ico -arch {amd64,arm64} -o cmd/muyue/rsrc_windows_{amd64,arm64}.syso`
before `go build` (both ci-main.yml and ci-develop.yml)
- Go automatically links *.syso files matching the target GOOS/GOARCH —
no code change in the cmd/muyue main package
- .syso files are gitignored: regenerated at every build, never committed
- Existing install-shortcuts subcommand already uses IconLocation =
"$exe,0" so the embedded icon flows automatically into Desktop +
Start Menu .lnk files
Web UI:
- web/public/favicon-{16,32}.png + muyue.png + muyue-64.png
- web/index.html: real <link rel="icon"> tags (16/32 PNG + apple-touch),
replacing the placeholder SVG hexagon
- App.jsx header: 22×22 logo image rendered next to the "MUYUE" wordmark
(rounded 4px corners for visual consistency with the source logo)
Install snippet (ci-main.yml changelog template):
- Idempotent first line: `New-Item -ItemType Directory -Force -Path $dest`
to handle the case where the user re-runs after a partial install
Versioning unchanged (still v0.7.3 — these additions stay on the same
release branch / PR #9).
Two user-reported pain points:
1. First-run setup proposed only MiniMax (no MiMo step). Onboarding
now offers both keys side-by-side under a single "apikey" step,
with per-key Validate buttons. At least one must be valid to
proceed; the rest of the providers (OpenAI/Anthropic/Z.AI/Ollama)
are not shown in the wizard — they're configured later via the
Config tab. Active provider = MiniMax if valid, else MiMo.
2. Windows install instructions failed: Move-Item to C:\Windows
requires admin. Replaced with a no-admin 4-line snippet that
installs to %LOCALAPPDATA%\Muyue and calls a new subcommand
`muyue install-shortcuts` to create Desktop + Start Menu .lnk
files and add the install dir to the user PATH. Shortcut creation
uses WScript.Shell COM via PowerShell — keeps Go binary
dependency-free. Folder paths resolved through
[Environment]::GetFolderPath so OneDrive/redirected profiles
work too.
- cmd/muyue/commands/install_shortcuts.go: new file
- web/src/components/OnboardingWizard.jsx: 2-key apikey step
- .gitea/workflows/ci-main.yml: updated install snippet
- internal/version/version.go: 0.7.2 → 0.7.3
- CHANGELOG.md: v0.7.3 entry
When at least one browser_test session is connected, every chat
message in Studio now auto-enables advanced reflection regardless of
the user toggle. The intent: during AI-driven UI testing, having a
second model produce a preliminary [RAPPORT PRÉALABLE] materially
improves which clicks the active model decides to perform and the
quality of the final ✓/✗ report.
- handlers_chat: derive wantReflection from body.AdvancedReflection
OR (browserTestStore has any active session). The user toggle still
works for normal conversations; tests just override it.
- Silent fallback when no inactive provider is configured (no error,
no behaviour change for single-provider setups).
- Tests.jsx: add a hint explaining the auto-on behaviour so the user
understands why the Studio toggle appears bypassed.
- Version 0.7.1 → 0.7.2 + CHANGELOG entry.
The terminal tab was unusable on Windows: creack/pty has no native
Windows ConPTY support, so pty.Start() returned "operating system not
supported" and the WebSocket closed immediately on any tab click —
even though the menu detection (wsl --list --quiet, pwsh, cmd) worked.
Introduce a termSession interface with two implementations selected at
runtime:
- ptySession (unix): unchanged behaviour, real PTY via creack/pty,
resize works, vim/top behave normally.
- pipeSession (windows): plain stdin + merged stdout/stderr pipes,
forwarded to the WebSocket. Resize is a no-op (no SIGWINCH without a
TTY), so full-screen TUIs misbehave in this mode — but launching
wsl.exe, pwsh, or cmd works for line-based interaction, which is
what the menu shortcuts target.
handleTerminalWS now goes through startTermSession(cmd); the unix path
is unchanged, the windows fallback kicks in only when pty.Start would
have failed.
Bump v0.7.0 → v0.7.1; CHANGELOG entry added.
The suffix _test.go makes Go treat the file as a test file (only
compiled during 'go test'), so server.go could not see the exported
BrowserTestStore / NewBrowserTestStore / RegisterBrowserTestTool and
the four handler methods at build time. Rename to a non-test name
keeps the same package-level visibility but compiles into the regular
api package.
Audit corrections (security, concurrency, stability):
- chat_engine: bound resp.Choices[0] access, release tool slot per-iteration
- conversation_multi: synchronous save under existing lock (was racy fire-and-forget)
- workflow/engine: short-circuit on failed deps (no more infinite busy-wait); track failed/skipped status
- handlers_workflow: rune-aware truncate for plan goal (UTF-8 safe)
- server: CORS limited to localhost origins (was wildcard)
- handlers_info / terminal: mask API keys and SSH passwords as "***" in GET responses; preserve stored secret if "***" sent on update
- terminal: sshpass uses -e + SSHPASS env var (was both -p and -e)
- handlers_chat: MaxBytesReader 50 MB on /api/chat
- image_cache: 10 MB cap per image
- handlers_config: font size <= 72; profile-save unmarshal errors propagated
- handlers_info: /lsp/auto-install ProjectDir restricted to user home
- Shell.jsx: parenthesized resize-condition (operator precedence)
- orchestrator_test: CleanAIResponse capitalization (fixes failing vet)
New features:
- platform: detect OS name (Debian, Ubuntu, Windows 11, macOS X.Y) and inject in Studio system prompt next to the date
- agents: default timeout 30 min for crush_run/claude_run (cap also 30 min)
- agents: new cwd, wsl_distro, wsl_user params; on Windows hosts launch via "wsl -d <distro> -u <user> --cd <cwd> --"
- agents: new claude_run tool (mirror of crush_run for Claude Code CLI)
- terminal: list installed WSL distros individually in new-tab menu (Windows only)
- studio: system prompt rewritten around BMAD-METHOD personas + mandatory delegation template
- studio: "Réflexion avancée" toggle — inactive provider produces a preliminary report injected as [RAPPORT PRÉALABLE] context for the active provider
- studio: "Historique compressé" toggle — collapses past tool calls to last action only, with "Tout afficher" expansion
Replace message-count context windows with token-budget based ones for both
studio and shell. Add /api/ai/task endpoint for background tool
check/install/update. Enhance sudo blocking to catch piped/chained elevation
commands. Add SSH password support via sshpass and connection editing UI.
Remove realTokens persistence in favor of consumption tracking. Bump to 0.4.1.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Replace message-count context windows with token-budget based ones for both
studio and shell. Add /api/ai/task endpoint for background tool
check/install/update. Enhance sudo blocking to catch piped/chained elevation
commands. Add SSH password support via sshpass and connection editing UI.
Remove realTokens persistence in favor of consumption tracking. Bump to 0.4.1.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
- Fix token count reset on app restart: persist realTokens in conversation.json
- Fix token/context window values: Studio 150K (summarize at 120K), Terminal 100K
- Fix table rendering in terminal tab: correct thead/tbody display model
- Fix copy button always top-right in Studio code blocks
- Add markdown horizontal rule (---) support in Studio and Terminal
- Fix bullet list double dot: remove CSS ::before duplicate bullet point
- Add image attachments support (VLM description, file mentions @file.ext)
- Add sudo detection with cache (sync.Once)
- Fix message content serialization (TextContent wrapper)
- Guide AI to use read_file instead of cat in studio prompt
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
- Block sudo/doas commands when not running as root
- Add real token counting from API responses
- Track and display consumption by provider/day
- Add Mermaid diagram rendering in Shell and Studio
- Add copy-to-clipboard buttons for code blocks
- Support tables in AI message rendering
- Update system prompt with context (date, time, root status)
💘 Generated with Crush
Assisted-by: MiniMax-M2.7 via Crush <crush@charm.land>
- Fix terminal default fontSize from 6px to 14px across all references
- Add terminal tool to shell AI via ChatEngine with tool_call streaming
- Fix provider key detection (apiKey → api_key, baseURL → base_url)
- Add mimo provider migration and validation endpoint
- Bump version to 0.4.0
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
All fallbacks were still using 12px. User confirmed 6px is the
correct baseline on their display.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
- Base font size reduced from 12px to 10px
- New tabs now initialize directly when added (was waiting for
tab switch because the MutationObserver only fired on visibility
changes, not on tab additions)
- Zoom level applied to newly created terminals
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
- Ctrl+/Ctrl-/Ctrl+0 to zoom in/out/reset terminal font size
- Zoom badge indicator in tab bar
- All shell shortcuts now shown in statusbar footer
- Added i18n labels for search, zoom, switch tab, next tab
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Add xterm addons from Vercel Hyper terminal: WebGL renderer with DOM
fallback, search bar (Ctrl+Shift+F), Unicode 11 grapheme support, and
inline image protocol. All existing functionality preserved.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
- Guarantee minimum 24x80 dimensions on WebSocket open
- Force reflow before init attempts
- Multiple fit attempts with increasing delays (0/50/100/200/400ms)
- Validate saved tabs structure from localStorage
- Resize active tab after closing another tab
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Replace Z.AI quota display with MiMo provider in the API Quota card.
ZAI is now a hidden fallback and should not appear in the dashboard.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Add MiMo-V2.5-Pro from Xiaomi Token Plan as a new AI provider with
base URL https://token-plan-ams.xiaomimimo.com/v1. The /model change
command now switches between MiniMax and MiMo only. ZAI is always
placed last in the fallback chain as the provider of ultimate resort.
Config panel shows MiniMax and MiMo cards.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
height:100% on .content>div fails because .content uses flex:1
without explicit height. Switch to position:absolute;inset:0 which
correctly fills the content area and gives xterm proper container
dimensions for fitAddon.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
xterm captures all keyboard input which prevents standard clipboard
operations. Add custom key handler to intercept Ctrl+Shift+C for
copy (selection) and Ctrl+Shift+V for paste, without interfering
with Ctrl+C (SIGINT) or browser devtools shortcut.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Stop propagation of Enter keydown in AI input and defer terminal
focus to next event loop tick to prevent xterm from capturing the
same key event.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
Use min-height:0 on xterm-wrapper (flex child) instead of height:100%
to properly fill available space in flex layout. Add delayed fit()
calls after initialization to let the layout stabilize before
calculating terminal cell dimensions.
💘 Generated with Crush
Assisted-by: GLM-5.1 via Crush <crush@charm.land>
