Mac Studio AI Manager
{{ locale === 'zh' ? '请登录以继续' : 'Please sign in to continue' }}
{{ oauthError }}
⏳
{{ locale === 'zh' ? '等待管理员审批' : 'Pending Approval' }}
{{ locale === 'zh' ? '你的账号正在等待管理员审批，审批通过后即可使用。' : 'Your account is pending admin approval.' }}
🚫
{{ locale === 'zh' ? '账号已被拒绝' : 'Account Rejected' }}
{{ locale === 'zh' ? '你的账号已被管理员拒绝，请联系管理员。' : 'Your account has been rejected. Please contact the admin.' }}
Mac Studio AI Manager
{{ t.name }}
{{ sseStateText }}
{{ error }}
🖥️
{{ locale === 'zh' ? '添加你的第一台设备' : 'Add Your First Device' }}
{{ locale === 'zh' ? '你还没有注册任何设备。点击下方按钮添加一台 Mac Studio，开始使用 AI 推理服务。' : 'No devices registered yet. Add a Mac Studio to start using AI inference services.' }}
{{ locale === 'zh' ? '用户总数' : 'Total Users' }}
{{ adminStats.users?.total || 0 }}
{{ adminStats.users?.approved || 0 }} {{ locale === 'zh' ? '已审批' : 'approved' }} ·
{{ adminStats.users?.pending || 0 }} {{ locale === 'zh' ? '待审批' : 'pending' }} ·
{{ adminStats.users?.rejected || 0 }} {{ locale === 'zh' ? '已拒绝' : 'rejected' }}
{{ locale === 'zh' ? '设备总数' : 'Total Devices' }}
{{ adminStats.devices?.total || 0 }}
{{ adminStats.devices?.online || 0 }} {{ locale === 'zh' ? '在线' : 'online' }}
{{ locale === 'zh' ? 'USD 计费' : 'USD Billing' }}
{{ locale === 'zh' ? '手动充值' : 'Manual Top-up' }}
{{ locale === 'zh' ? '模型价格' : 'Model Price' }}
{{ locale === 'zh' ? '推理引擎版本' : 'Engine Versions' }}
{{ locale === 'zh'
? '为各推理引擎指定目标版本，一键推送到所有在线设备（设备自动判断首装/升级）。留空=不管理该引擎。'
: 'Set target versions per engine and push to all online devices (each device auto-decides install/upgrade). Empty = unmanaged.' }}
{{ locale === 'zh'
? 'Ollama 用 semver（0.5.1）；llama-server 用 release tag（b4321）；vLLM-MLX / oMLX 用 PyPI 版本（0.4.0 / 0.5.5）。仅协商支持的设备会收到指令。'
: 'Ollama: semver (0.5.1); llama-server: release tag (b4321); vLLM-MLX / oMLX: PyPI version (0.4.0 / 0.5.5). Only capable devices receive commands.' }}
{{ locale === 'zh' ? '设备版本' : 'Device Versions' }}
{{ locale === 'zh' ? '最新版本' : 'Latest' }}: {{ upgradeLatestVersion || '—' }}
{{ upgradeDisabledReason }}
{{ locale === 'zh' ? '推送结果' : 'Push Result' }}: {{ upgradeResult.pushed?.length || 0 }} {{ locale === 'zh' ? '台已推送' : 'pushed' }},
{{ upgradeResult.skipped_legacy?.length || 0 }} legacy,
{{ upgradeResult.skipped_busy?.length || 0 }} {{ locale === 'zh' ? '忙碌' : 'busy' }},
{{ upgradeResult.skipped_install_required?.length || 0 }} {{ locale === 'zh' ? '需重装' : 'install required' }},
{{ upgradeResult.skipped_up_to_date?.length || 0 }} {{ locale === 'zh' ? '已最新' : 'up-to-date' }},
{{ upgradeResult.skipped_offline?.length || 0 }} {{ locale === 'zh' ? '离线' : 'offline' }}
{{ locale === 'zh' ? '安装诊断报告' : 'Install Diagnostics' }}
{{ locale === 'zh' ? '客户端 install.sh 执行失败时会自动上报。仅显示最近 50 条。' : 'Auto-reported by install.sh on failure. Latest 50 only.' }}
| {{ locale === 'zh' ? '时间' : 'Time' }} | tenant | {{ locale === 'zh' ? '阶段' : 'Stage' }} | {{ locale === 'zh' ? '退出码' : 'Exit' }} | {{ locale === 'zh' ? '错误' : 'Error' }} | |
|---|---|---|---|---|---|
| {{ formatDate((r._ts || 0) * 1000) }} | {{ r.tenant_id }} |
{{ r.stage }} | {{ r.exit_code }} | ||
| {{ installReportsLoading ? '...' : (locale === 'zh' ? '暂无诊断报告' : 'No reports yet') }} | |||||
{{ locale === 'zh' ? '远程诊断' : 'Remote Diagnose' }}
{{ locale === 'zh' ? '对在线设备执行 Agent 侧白名单命令（只读诊断 / 针对性恢复）。命令由 Agent 代码内嵌，无 shell 直通。' : 'Run Agent-side whitelisted commands on an online device. Commands are hard-coded in Agent; no raw shell.' }}
{{ diagnoseError }}
status: {{ diagnoseResult.status }}
· rc: {{ diagnoseResult.rc ?? '-' }}
· elapsed: {{ diagnoseResult.elapsed_ms }}ms
· truncated: {{ diagnoseResult.truncated ? 'yes' : 'no' }}
stdout:
{{ diagnoseResult.stdout }}
stderr:
{{ diagnoseResult.stderr }}
reason: {{ diagnoseResult.reason }}
{{ t('admin.mano.title') }}
{{ manoHfTokenSet ? t('mano.hfToken.set') : t('mano.hfToken.unset') }}
{{ adminManoTokenStatus.msg }}
{{ manoState && manoState.public_enabled ? t('mano.public.enabled') : t('mano.public.disabled') }}
{{ locale === 'zh' ? '需先加载模型才能开启（关闭路径不限制）' : 'Load a model first before enabling (closing always allowed)' }}
mano-server
{{ statusText(engineStatus('mano-server')) }}
{{ locale === 'zh' ? '最近审计（20 条）' : 'Recent Audit (20)' }}
| {{ locale === 'zh' ? '时间' : 'Time' }} | tenant | command | status | rc | elapsed | args |
|---|---|---|---|---|---|---|
| {{ r.ts }} | {{ r.tenant_id }} |
{{ r.command }} |
{{ r.status }} | {{ r.rc ?? '-' }} | {{ r.elapsed_ms }}ms | {{ JSON.stringify(r.args) }} |
{{ e.name }}
{{ statusText(e.status) }}
{{ fmtMs(e.latency_ms) }}
{{ t('requests24h') }}
{{ summary.total_requests || 0 }}
{{ t('tokens24h') }}
{{ formatNumber(summary.total_tokens_in + summary.total_tokens_out) }}
In: {{ formatNumber(summary.total_tokens_in) }} / Out: {{ formatNumber(summary.total_tokens_out) }}
{{ t('avgLatency') }}
{{ fmtMs(summary.avg_latency_ms || 0) }}
{{ t('chart.requestTrend') }}
{{ t('chart.tokenByModel') }}
{{ t('recentRequests') }}
| {{ t('col.time') }} | {{ t('col.model') }} | Tokens | {{ t('col.latency') }} | {{ t('col.status') }} |
|---|---|---|---|---|
| {{ t('noRequests') }} | ||||
| {{ formatTime(r.time_str || r.timestamp) }} | {{ r.model }} | {{ r.tokens_in + r.tokens_out }} | {{ fmtMs(r.latency_ms) }} | {{ r.status_code }} |
{{ t('litellmRoutes') }}
| {{ t('col.modelName') }} | {{ t('col.source') }} | {{ t('col.backendModel') }} | {{ t('col.apiBase') }} | {{ t('col.actions') }} |
|---|---|---|---|---|
| {{ t('noRoutes') }} | ||||
| {{ r.model_name }} {{ t('autoSync') || 'Auto' }} | {{ getRouteSource(r).label }} | {{ r.litellm_params?.model || '-' }} | {{ r.litellm_params?.api_base || '-' }} | |
{{ t('audioApiRoutes') }}
| {{ t('col.route') }} | {{ t('col.behavior') }} | {{ t('col.backendModel') }} | {{ t('col.apiBase') }} | {{ t('col.actions') }} |
|---|---|---|---|---|
| {{ t('audioNoRoutes') }} | ||||
|
{{ route.route_name }}
{{ route.display_name }}
{{ locale === 'zh' ? '默认路由' : 'Default route' }}
{{ locale === 'zh' ? '当前激活' : 'Active model' }}
|
{{ route.endpoint }}
model=
{{ route.model_param }}
{{ locale === 'zh' ? '可省略 model，命中当前 active ASR' : 'model is optional and uses the current active ASR' }}
|
{{ route.backend_model }} |
{{ route.api_base }}
{{ route.files_endpoint || '/v1/files' }} / {{ route.endpoint }}
|
|
{{ t('modelTools') }}
Ollama
Llama Server
vLLM-MLX
Audio
Mano-P
oMLX
{{ t('storage.tab') }}
{{ selectedRegistryModel.name }}
{{ c }}
×
{{ t('selectSize') }}:
{{ s }}
{{ t('selectSize') }}:
{{ mt.tag }} · {{ mt.size }}
{{ locale === 'zh' ? '加载 tag 列表...' : 'Loading tags...' }}
{{ t('capAll') }}
Vision
Tools
Thinking
Embedding
{{ t('loadingModels') }}
{{ m.name }} Cloud
{{ m.description }}
{{ t('noRegistryResults') }}
{{ t('useCustom') }}: {{ pullName.trim() }}
{{ activePullText }}
{{ pullNotice.text }}
| {{ t('col.name') }} | {{ t('col.size') }} | {{ t('col.modified') }} | {{ t('col.actions') }} |
|---|---|---|---|
| {{ t('noModels') }} | |||
| {{ m.name }} | {{ formatSize(m.size) }} | {{ formatDate(m.modified_at) }} | |
⚠️ {{ locale === 'zh' ? 'llama-server 未安装，GGUF 模型无法加载。请运行 setup.sh 或手动编译 llama.cpp。' : 'llama-server not found. GGUF models cannot be loaded. Run setup.sh or compile llama.cpp manually.' }}
{{ llamaStatus.active_model_id || llamaStatus.alias || llamaStatus.loaded_model }}
Active
{{ llamaLoadingAlias || llamaLoadingModel }}
{{ llamaLoadingProgress >= 0 ? 'Loading ' + Math.round(llamaLoadingProgress * 100) + '%' : 'Loading...' }}
{{ locale === 'zh' ? '加载失败' : 'Load Failed' }}
{{ t('noModelLoaded') }}
{{ getLlamaWarningText() }}
{{ t('loadingModels') }}
{{ m.repo_id }}
{{ m.description }}
{{ t('noRegistryResults') }}
{{ selectedLlamaRepo.repo_id }}
× {{ t('cancel') }}
{{ v.quant_type || v.name }}
{{ v.shard_count }} {{ t('shards') || 'shards' }}
{{ formatSize(v.total_size) }}
{{ llamaDownloadStatus }}
| {{ t('col.name') }} | {{ t('quantType') }} | {{ t('col.size') }} | {{ t('col.status') }} | {{ t('col.actions') }} |
|---|---|---|---|---|
| {{ t('noGGUF') }} | ||||
| {{ m.name }}({{ m.shard_count }} shards) | {{ m.quant_type }} | {{ formatSize(m.total_size) }} | Loaded {{ t('llamaStatusBadgeStale') }} {{ llamaLoadingProgress >= 0 ? 'Loading ' + Math.round(llamaLoadingProgress * 100) + '%' : 'Loading...' }} Incomplete ({{ m.shard_count }}/{{ m.expected_shards }}) Error Available | |
⚠️ {{ locale === 'zh' ? 'vllm-mlx 未安装。请在 Python venv 中安装: pip install vllm-mlx' : 'vllm-mlx not installed. Install in venv: pip install vllm-mlx' }}
{{ vllmMlxStatus.model }}
Active
{{ locale === 'zh' ? '启动中，首次需下载模型...' : 'Starting, downloading model on first run...' }}
❌ {{ locale === 'zh' ? '启动失败' : 'Start failed' }}: {{ vllmMlxStatus.error_detail || '' }}
{{ locale === 'zh' ? '未运行' : 'Not running' }}
{{ t('loadingModels') }}
{{ m.repo_id }}
{{ m.description }}
{{ t('noRegistryResults') }}
{{ selectedMlxModel.repo_id }}
×
⚠️ {{ locale === 'zh' ? 'Audio Server 当前不可达。' : 'Audio Server unavailable.' }}
{{ locale === 'zh' ? '三步使用：下载 → 加载 → 激活。仅第一个加载的模型自动激活。' : 'Three steps: Download → Load → Activate. The first loaded model auto-activates.' }}
| {{ t('col.name') }} | {{ t('audioCol.type') }} | {{ t('col.size') }} | {{ t('audioCol.languages') }} | {{ t('audioCol.status') }} | {{ t('col.actions') }} | {{ t('audioCol.note') }} |
|---|---|---|---|---|---|---|
| {{ t('noModels') }} | ||||||
|
{{ m.display_name }}
{{ m.model_id }}
|
ASR TTS | {{ formatSize(m.total_size_bytes) }} | {{ (m.supported_languages || []).join(', ') || '—' }} | {{ t('audioStatus.tts_placeholder') }} {{ t('audioStatus.not_downloaded') }} {{ t('audioStatus.downloading') }} {{ (m.download_progress || 0).toFixed(0) }}% {{ t('audioStatus.downloaded') }} {{ t('audioStatus.loaded') }} {{ t('audioStatus.active') }} {{ t('audioStatus.failed') }} |
⚠️ {{ m.unavailable_reason }}
|
|
⚠️ {{ locale === 'zh' ? 'Mano-P Server 当前不可达。' : 'Mano-P Server unavailable.' }}
{{ manoState && manoState.public_enabled ? t('mano.public.enabled') : t('mano.public.disabled') }}
HF Token: {{ manoHfTokenSet ? t('mano.hfToken.set') : t('mano.hfToken.unset') }}
engine: {{ manoState.engine }}
{{ locale === 'zh' ? '生命周期：未下载 → 下载中 → 已下载 → 加载中 → 已加载 → 卸载 / 失败。' : 'Lifecycle: not_downloaded → downloading → downloaded → loading → loaded → unloaded / failed.' }}
| {{ t('col.name') }} | {{ t('col.size') }} | {{ locale === 'zh' ? '状态' : 'Status' }} | {{ locale === 'zh' ? '进度' : 'Progress' }} | {{ t('col.actions') }} | {{ locale === 'zh' ? '错误' : 'Error' }} |
|---|---|---|---|---|---|
| {{ t('noModels') }} | |||||
|
{{ m.display_name || m.model_id }}
{{ t('mano.modelState.active') }}
{{ m.model_id }}
|
{{ m.size_bytes ? formatSize(m.size_bytes) : '—' }} | {{ t('mano.modelState.not_downloaded') }} {{ t('mano.modelState.downloading') }} {{ ((m.progress || 0) * 100).toFixed(0) }}% {{ t('mano.modelState.downloaded') }} {{ t('mano.modelState.loading') }} {{ t('mano.modelState.loaded') }} {{ t('mano.modelState.unloaded') }} {{ t('mano.modelState.failed') }} | ⚠️ {{ locale === 'zh' ? '请联系管理员配置 HF token' : 'Please ask admin to configure HF token' }} {{ m.error }} | ||
oMLX
{{ omlxStatus ? (omlxStatus.status === 'running' ? (locale === 'zh' ? '运行中' : 'Running') : (locale === 'zh' ? '未运行' : 'Stopped')) : '...' }}
v{{ omlxStatus.version }}
:{{ omlxStatus.port }}
{{ locale === 'zh' ? '已上架模型' : 'Published models' }}: {{ omlxStatus.model_count }}
{{ locale === 'zh' ? '手工实例' : 'Legacy' }}
{{ locale === 'zh' ? omlxStatus.service_control_error : 'Engine supervision is unverified. Start, stop and restart are disabled; model management remains available.' }}
{{ omlxError }}
{{ locale === 'zh' ? 'Peer 节点（分压/分流）' : 'Peer Nodes (LB/Shard)' }}
| {{ locale === 'zh' ? '地址' : 'URL' }} | {{ locale === 'zh' ? '状态' : 'Status' }} | {{ locale === 'zh' ? '延迟' : 'Latency' }} | {{ locale === 'zh' ? '模型数' : 'Models' }} | {{ locale === 'zh' ? '模型' : 'Serving' }} | {{ locale === 'zh' ? '检查时间' : 'Checked' }} |
|---|---|---|---|---|---|
| {{ p.url }} | {{ p.status }} | {{ p.latency_ms != null ? fmtMs(p.latency_ms) : '—' }} | {{ p.model_count }} | {{ (p.models || []).join(', ') || '—' }} | {{ p.checked_at }} |
{{ locale === 'zh' ? '每 30 秒自动探测一次。offline 节点的模型请求会由 LiteLLM 路由到其他健康节点。' : 'Probed every 30s. Requests to offline peers are routed to healthy nodes by LiteLLM.' }}
{{ omlxText('模型管理连接', 'Model management connection') }}
{{ omlxLegacy ? omlxText('需要升级设备后台', 'Device backend upgrade required') : omlxAdmin?.connected ? omlxText('已连接', 'Connected') : omlxText('未连接', 'Not connected') }}
{{ omlxText('此设备后台尚未支持新版模型管理。旧列表只能确认模型已上架，不能确认内存加载状态；新增操作已禁用。', 'This device needs the new backend. The legacy list only confirms publication, not memory residency; new actions are disabled.') }}
{{ omlxDetail(omlxAdmin.error) }}
{{ omlxText('尚未验证管理能力，请刷新查看连接状态。', 'Management capability has not been verified. Refresh to check the connection.') }}
{{ omlxText('已下载模型', 'Downloaded models') }} · {{ currentTenantName }}
| {{ t('col.name') }} | {{ t('col.size') }} | {{ omlxText('状态', 'Status') }} | {{ t('col.actions') }} |
|---|---|---|---|
| {{ omlxError ? omlxText('无法确认模型清单', 'Inventory unavailable') : t('noModels') }} | |||
{{ m.name }} {{ m.runtime_id }} |
{{ formatSize(m.size_bytes) }} | {{ modelStateText(m) }} |
{{ omlxDetail(m.blocked_reason) }}
|
{{ omlxText('上架、加载与下载是独立状态。仅未上架且未被引擎使用的本地副本允许删除；卸载只释放内存。', 'Published, loaded and downloaded are separate states. Only unpublished local copies unused by the engine can be deleted; unloading only frees memory.') }}
{{ omlxText('模型设置', 'Model settings') }} · {{ currentTenantName }} · {{ omlxSettingsModel.name }}
{{ omlxText('只保存修改的字段，清空数值可恢复引擎默认。若引擎要求重新加载，会显示提示，不会自动重启。', 'Only changed fields are saved; clear a numeric value to restore the engine default. A required reload is reported and never started automatically.') }}
{{ omlxText('下载 HuggingFace 模型', 'Download a HuggingFace model') }}
{{ omlxText('由本设备的 oMLX 下载；完成后仍需验证架构和运行内存。', 'Downloaded by oMLX on this device; architecture support and runtime memory still need verification.') }}
{{ task.repo_id || task.model_id || omlxDownloadId(task) }}{{ task.status }}{{ Number(task.progress).toFixed(1) }}%{{ formatSize(task.downloaded_bytes || 0) }} / {{ formatSize(task.total_bytes) }}
{{ omlxDetail(task.error) }}
{{ omlxText('推理缓存', 'Inference cache') }}
SSD: {{ omlxCache?.ssd?.size_bytes != null ? formatSize(omlxCache.ssd.size_bytes) : omlxText('未知', 'Unknown') }}
{{ omlxText('使用引擎提供的缓存清理功能，保留模型权重。空间大小与模型统计分开。', 'Uses the engine cache controls and retains model weights. Cache space is separate from model size.') }}
{{ omlxText('删除中断的残留', 'Interrupted deletion remnants') }}
{{ row.id }} · {{ formatSize(row.size_bytes) }} · {{ row.status }}
{{ omlxDetail(row.blocked_reason) }}
{{ omlxText('设备操作任务', 'Device jobs') }}
{{ job.action }} · {{ job.model_id || 'oMLX' }} · {{ omlxJobStatus(job.status) }}
{{ omlxDetail(job.error) }}
{{ locale === 'zh' ? '服务日志' : 'Service Logs' }}
{{ omlxLogs.length ? omlxLogs.join('\n') : (locale === 'zh' ? '（暂无日志）' : '(no logs)') }}
{{ locale === 'zh' ? '系统中已下载、占用磁盘的模型（跨引擎汇总）。卸载不等于删除——删除才会释放空间。' : 'Downloaded models occupying disk across engines. Unloading ≠ deleting; only delete frees space.' }}
{{ locale === 'zh' ? '已统计模型大小' : 'Counted model size' }}:
{{ formatSize(diskUsage.total_size_bytes || 0) }}
{{ diskError }}
| {{ t('col.name') }} | {{ locale === 'zh' ? '引擎' : 'Engine' }} | {{ t('col.size') }} | {{ locale === 'zh' ? '状态' : 'Status' }} | {{ t('col.actions') }} |
|---|---|---|---|---|
| {{ t('noModels') }} | ||||
|
{{ row.name }}
{{ row.id }}
|
{{ storageEngineLabel(row.engine) }} | {{ formatSize(row.size_bytes) }} {{ omlxText('预估', 'Estimated') }} | {{ modelStateText(row) }} |
{{ omlxDetail(row.blocked_reason) }}
|
{{ omlxText('缓存与其他存储位置', 'Cache and other storage locations') }}
{{ omlxText('共享缓存可能包含上方已经统计的模型，以下大小不累加到模型总计。', 'Shared caches may include models counted above; these sizes are not added to the model total.') }}
{{ cache.name }} · {{ formatSize(cache.size_bytes) }}{{ omlxText('可能重复统计', 'May overlap') }}
{{ omlxDetail(cache.blocked_reason) }}
{{ cache.scope_note }} · {{ cache.root_source === 'default' ? omlxText('默认目录', 'Default location') : cache.root_source }}
{{ cache.error || omlxText('部分文件无法读取，当前大小可能不完整', 'Some files could not be inspected; the size may be incomplete') }}
{{ omlxText('查看模型缓存明细', 'View model cache details') }} ({{ cache.entries.length }})
| {{ t('col.name') }} | {{ t('col.size') }} | {{ omlxText('文件状态', 'File status') }} |
|---|---|---|
{{ entry.name }} {{ entry.relative_path }} | {{ formatSize(entry.size_bytes) }} | {{ entry.status_label }} · {{ omlxText('扫描不完整', 'Incomplete scan') }} |
{{ omlxText('删除残留占用', 'Deletion remnants') }}: {{ formatSize(diskUsage.remnant_size_bytes) }}
⚠️ {{ storageEngineLabel(eng) }}: {{ locale === 'zh' ? '探查失败' : 'probe failed' }} ({{ msg }})
ℹ️ {{ diskUsage.engines['vllm-mlx'].note }}
{{ t('pg.mode') }}
{{ t('pg.model') }}
{{ cap }}
{{ pgSelectedModelInfo.parameter_size }}
{{ t('pg.params') }}
System Prompt
{{ t('pg.audioModel') }}
ASR
{{ locale === 'zh' ? 'Audio API' : 'Audio API' }}
{{ locale === 'zh' ? '默认路由' : 'Default route' }}
{{ locale === 'zh'
? 'Playground 会调用 Audio OpenAI 兼容接口 /v1/audio/transcriptions。默认路由可省略 model，命中当前 active ASR。'
: 'Playground calls the Audio OpenAI-compatible /v1/audio/transcriptions endpoint. The default route omits model and uses the current active ASR.'
}}
{{ t('pg.params') }}
{{ t('pg.audioInputMode') }}
{{ pgAudioFile ? pgAudioFile.name + ' (' + formatSize(pgAudioFile.size) + ')' : t('pg.audioNoFile') }}
{{ t('pg.audioTwoStageHint') }}
{{ t('pg.audioMicAskingPerm') }}
{{ t('pg.audioMicUploading') }}
{{ pgAudioRecordingError }}
{{ t('pg.audioMicRecorded') }}: {{ formatSize(pgAudioRecordedBlob.size) }} ({{ pgAudioRecordedMime || 'audio' }})
{{ t('pg.mano.screenshot') }}
{{ t('pg.mano.dropHint') }}
{{ pgManoImageFile.name }} ({{ formatSize(pgManoImageFile.size) }})
{{ t('pg.mano.twoStageHint') }}
{{ pgManoTask.length }} / 4000
#{{ i + 1 }}
{{ t('pg.emptyTitle') }}
{{ t('pg.emptyDesc') }}
{{ msg.role }}
{{ t('pg.thinking') }} ({{ msg.thinking.length }} chars)
{{ msg.thinking }}
{{ msg.content }}
{{ pgLastReqInfo }}
{{ pgAudioLastReqInfo }}
{{ t('pg.audioMicAskingPerm') }}
{{ t('pg.audioMicUploading') }}
{{ pgAudioRecordingError }}
{{ t('pg.mano.empty.title') }}
{{ t('pg.mano.empty.desc') }}
{{ t('pg.mano.result') }}
{{ pgManoResult.error }}
{{ t('pg.mano.thought') }}
{{ pgManoResult.thought }}
{{ pgManoResult.action_desp }}
{{ t('pg.mano.action') }}:
{{ JSON.stringify(pgManoResult.action, null, 2) }}
{{ t('pg.mano.coordinates') }}:
![]()
{{ t('pg.mano.raw') }}
{{ pgManoResult.raw }}
{{ pgManoLastReqInfo }}
{{ pgManoImageFile ? (locale === 'zh' ? '已选图片' : 'Image selected') + ': ' + pgManoImageFile.name : t('pg.mano.dropHint') }}
{{ t('engineLatency') }}
{{ e.name }}
{{ fmtMs(e.latency_ms) }}
{{ t('col.status') }}: {{ statusText(e.status) }} ({{ e.status_code }})
{{ t('chart.latencyTrend') }}
{{ t('publicNetwork') }}
{{ t('noData') }}
{{ t2.name }}
{{ fmtMs(t2.latency_ms) }}
{{ t2.status }}
{{ locale === 'zh' ? '官方连通性测试' : 'Official Connectivity Tests' }}
{{ locale === 'zh' ? '不计额度' : 'No quota charge' }}
{{ connectivityError }}
{{ locale === 'zh' ? '目标设备' : 'Target Device' }}
{{ currentTenantName || currentTenantId || '-' }}
{{ currentTenantId || '-' }}{{ locale === 'zh' ? '通过' : 'Passed' }}
{{ connectivitySummary.pass }}
{{ locale === 'zh' ? '共' : 'Total' }} {{ connectivitySummary.total }}
{{ locale === 'zh' ? '失败' : 'Failed' }}
{{ connectivitySummary.fail }}
{{ connectivityResult ? fmtMs(connectivityResult.elapsed_ms || 0) : '-' }}
{{ locale === 'zh' ? '覆盖接口' : 'Suites' }}
{{ connectivitySuites.length || '-' }}
Manager / Inference / Audio / Files
{{ locale === 'zh' ? '正在测试公网入口...' : 'Testing public endpoints...' }}
{{ locale === 'zh' ? '尚未运行测试' : 'No test run yet' }}
{{ suite.name }} {{ suite.status }}
| {{ locale === 'zh' ? '检查项' : 'Check' }} | {{ locale === 'zh' ? '接口' : 'Endpoint' }} | {{ locale === 'zh' ? '期望' : 'Expected' }} | {{ locale === 'zh' ? '结果' : 'Result' }} | {{ locale === 'zh' ? '耗时' : 'Latency' }} | {{ locale === 'zh' ? '详情' : 'Detail' }} |
|---|---|---|---|---|---|
| {{ item.name }} |
{{ item.method }}
{{ endpointPath(item.url) }}
|
{{ item.expected }} | {{ item.status }} · {{ item.status_code || 0 }} | {{ fmtMs(item.elapsed_ms || 0) }} | {{ item.detail }} |
1h
6h
24h
7d
{{ t('col.requests') }}
{{ usageSummary.total_requests || 0 }}
Tokens In
{{ formatNumber(usageSummary.total_tokens_in) }}
Tokens Out
{{ formatNumber(usageSummary.total_tokens_out) }}
{{ t('avgLatency') }}
{{ fmtMs(usageSummary.avg_latency_ms || 0) }}
{{ t('chart.requestTrend') }}
{{ t('chart.tokenUsage') }}
{{ t('byModel') }}
| {{ t('col.model') }} | {{ t('col.requests') }} | Tokens In | Tokens Out | {{ t('avgLatency') }} |
|---|---|---|---|---|
| {{ t('noData') }} | ||||
| {{ m.model }} | {{ m.requests }} | {{ formatNumber(m.tokens_in) }} | {{ formatNumber(m.tokens_out) }} | {{ fmtMs(m.avg_latency_ms) }} |
🖥️
{{ locale === 'zh' ? '添加你的第一台 Mac Studio' : 'Add your first Mac Studio' }}
{{ locale === 'zh' ? '注册一台设备并在你的 Mac Studio 上运行安装命令，即可开始远程管理。' : 'Register a device and run the install command on your Mac Studio to get started.' }}
{{ locale === 'zh' ? '我的设备' : 'My Devices' }}
{{ locale === 'zh' ? '部分设备需要升级' : 'Some devices need upgrading' }}
{{ locale === 'zh' ? '标有「需手动升级」的设备无法接收自动更新。请点击对应设备的「重置 Token」按钮，在设备上执行新的安装命令即可完成升级。' : 'Devices marked "manual upgrade" cannot receive OTA updates. Click "Reset Token" for the device, then run the new install command on it.' }}
| {{ locale === 'zh' ? '名称' : 'Name' }} | {{ locale === 'zh' ? '状态' : 'Status' }} | {{ locale === 'zh' ? '版本' : 'Version' }} | {{ locale === 'zh' ? '租户 ID' : 'Tenant ID' }} | Token | {{ locale === 'zh' ? '最后连接' : 'Last Connected' }} | {{ locale === 'zh' ? '操作' : 'Actions' }} |
|---|---|---|---|---|---|---|
| {{ t.name }} | {{ t.effective_status === 'connected' ? (locale === 'zh' ? '在线' : 'Online') : t.effective_status === 'never_connected' ? (locale === 'zh' ? '从未连接' : 'Never Connected') : (locale === 'zh' ? '离线' : 'Offline') }} | {{ locale === 'zh' ? '需手动升级' : 'manual upgrade' }} {{ t.agent_version.slice(0,16) }} {{ locale === 'zh' ? '可更新' : 'update' }} {{ t.upgrade_status }} {{ locale === 'zh' ? '升级失败' : 'failed' }} — | {{ t.tenant_id }} |
{{ t.agent_token_prefix }}... |
{{ t.last_connected ? formatDate(t.last_connected) : '-' }} |
{{ locale === 'zh' ? '账户' : 'Account' }}
{{ locale === 'zh' ? '余额、充值和扣费流水' : 'Balance, recharge, and billing activity' }}
{{ userBillingError }}
{{ locale === 'zh' ? '当前余额' : 'Current Balance' }}
{{ formatUsdAmount(userBillingBalance?.balance, 2) }}
USD
{{ locale === 'zh' ? '请求状态' : 'Request Status' }}
{{ userBillingBalance?.can_request ? (locale === 'zh' ? '可请求' : 'Enabled') : (locale === 'zh' ? '需充值' : 'Recharge Required') }}
{{ locale === 'zh' ? '余额大于 0 时可继续调用 API' : 'API calls continue while balance is above 0' }}
{{ locale === 'zh' ? '累计充值' : 'Total Recharged' }}
{{ formatUsdAmount(userBillingBalance?.total_recharged, 2) }}
USD
{{ locale === 'zh' ? '累计用量' : 'Total Usage' }}
{{ formatUsdAmount(userBillingBalance?.total_usage, 2) }}
USD
{{ locale === 'zh' ? '充值' : 'Recharge' }}
{{ locale === 'zh' ? '在线充值入口稍后开放；现在请联系管理员处理充值。' : 'Online recharge will be added later. Contact the admin for top-ups now.' }}
微信：{{ billingAdminWechat }}
{{ locale === 'zh' ? '扣费流水' : 'Billing Activity' }}
| {{ locale === 'zh' ? '时间' : 'Time' }} | {{ locale === 'zh' ? '模型' : 'Model' }} | {{ locale === 'zh' ? '状态' : 'Status' }} | {{ locale === 'zh' ? '输入 / 输出 Tokens' : 'Input / Output Tokens' }} | {{ locale === 'zh' ? '金额' : 'Amount' }} | {{ locale === 'zh' ? '余额' : 'Balance' }} |
|---|---|---|---|---|---|
| {{ formatDateTime(tx.created_at) }} |
{{ tx.model_id || '-' }}
{{ tx.reason || (locale === 'zh' ? '充值' : 'Top-up') }}
|
{{ tx.status || '-' }} | {{ formatNumber(tx.input_tokens || 0) }} / {{ formatNumber(tx.output_tokens || 0) }} - | {{ formatUsdAmount(tx.amount, 8) }} | {{ formatUsdAmount(tx.balance_after, 2) }} |
{{ locale === 'zh' ? '暂无扣费流水' : 'No billing activity yet' }}
{{ locale === 'zh' ? '模型价格' : 'Model Pricing' }}
| {{ locale === 'zh' ? '模型' : 'Model' }} | {{ locale === 'zh' ? '输入 USD / 100万 Tokens' : 'Input USD / 1M Tokens' }} | {{ locale === 'zh' ? '输出 USD / 100万 Tokens' : 'Output USD / 1M Tokens' }} | {{ locale === 'zh' ? '来源' : 'Source' }} |
|---|---|---|---|
{{ p.id }} |
{{ formatUsdAmount(p.input_price_per_million_tokens, 3) }} | {{ formatUsdAmount(p.output_price_per_million_tokens, 3) }} | {{ p.source || '-' }} |
{{ locale === 'zh' ? '暂无模型价格' : 'No model prices yet' }}
{{ t('engineConfig') }}
| {{ t('col.engine') }} | URL | {{ t('col.port') }} | {{ t('col.status') }} |
|---|---|---|---|
| Ollama | http://127.0.0.1:11435 | 11435 | {{ statusText(engineStatus('ollama')) }} |
| llama-server | http://127.0.0.1:8000 | 8000 | {{ statusText(engineStatus('llama-server')) }} |
| LiteLLM | http://127.0.0.1:4000 | 4000 | {{ statusText(engineStatus('litellm')) }} |
| Audio Server | http://127.0.0.1:8200 | 8200 | {{ statusText(engineStatus('audio-server')) }} |
| Mano-P Server | http://127.0.0.1:8300 | 8300 | {{ statusText(engineStatus('mano-server')) }} |
| oMLX | http://127.0.0.1:8002 | 8002 | {{ statusText(engineStatus('omlx')) }} |
| Manager | http://127.0.0.1:8080 | 8080 | {{ statusText(managerLiveStatus) }} |
{{ t('publicUrls') }}
| {{ t('col.service') }} | URL |
|---|---|
| LiteLLM API | {{ publicBases.inference }} |
| Audio API | {{ publicBases.audio }} |
| {{ t('managerPanel') }} | {{ publicBases.manager }} |
API Keys
{{ locale === 'zh' ? '管理推理 API 访问密钥' : 'Manage inference API access keys' }}
| {{ locale === 'zh' ? '名称' : 'Name' }} | Key | {{ locale === 'zh' ? '创建时间' : 'Created' }} | {{ locale === 'zh' ? '最后使用' : 'Last Used' }} | {{ t('col.actions') }} |
|---|---|---|---|---|
| {{ k.name }} | {{ k.prefix }}... |
{{ formatDate(k.created) }} | {{ k.last_used ? formatDate(k.last_used) : '-' }} |
{{ locale === 'zh' ? '暂无 API Key' : 'No API keys yet' }}
{{ t('version') }}
Mac Studio AI Manager
v2.3.0
{{ locale === 'zh' ? '概述' : 'Overview' }}
{{ locale === 'zh' ? '安装' : 'Install' }}
{{ locale === 'zh' ? '卸载' : 'Uninstall' }}
{{ locale === 'zh' ? '架构原理' : 'Architecture' }}
{{ locale === 'zh' ? '故障排查' : 'Troubleshoot' }}
API
{{ locale === 'zh' ? '概述' : 'Overview' }}
{{ locale === 'zh' ? 'Mac Studio AI Manager 是一个远程管理面板，通过 WSS (WebSocket Secure) 中继技术，让你在任何地方通过浏览器管理局域网内的 Mac Studio AI 推理服务。无需公网 IP，无需端口转发。' : 'Mac Studio AI Manager is a remote management panel that uses WSS (WebSocket Secure) relay technology to manage your LAN-based Mac Studio AI inference services from anywhere via a browser. No public IP or port forwarding required.' }}
{{ locale === 'zh' ? '模型管理' : 'Model Management' }}
Ollama / llama-server / LiteLLM
{{ locale === 'zh' ? '在线测试' : 'Playground' }}
{{ locale === 'zh' ? '多模型对话、流式输出' : 'Multi-model chat, streaming' }}
{{ locale === 'zh' ? '性能监控' : 'Performance' }}
{{ locale === 'zh' ? '延迟探测、用量统计' : 'Latency probes, usage stats' }}
{{ locale === 'zh' ? '多设备' : 'Multi-Device' }}
{{ locale === 'zh' ? '一个账号管理多台 Mac Studio' : 'Manage multiple Mac Studios' }}
{{ locale === 'zh' ? '系统要求' : 'Requirements' }}
| {{ locale === 'zh' ? '项目' : 'Item' }} | {{ locale === 'zh' ? '要求' : 'Requirement' }} |
|---|---|
| {{ locale === 'zh' ? '操作系统' : 'OS' }} | macOS 13+ |
| Python | 3.10+ |
| {{ locale === 'zh' ? '网络' : 'Network' }} | {{ locale === 'zh' ? '可访问外网（WSS 外连）' : 'Internet access (WSS outbound)' }} |
| {{ locale === 'zh' ? '推理引擎' : 'Inference Engine' }} | Ollama / llama-server / LiteLLM ({{ locale === 'zh' ? '至少一个' : 'at least one' }}) |
{{ locale === 'zh' ? '一键安装' : 'One-Click Install' }}
{{ locale === 'zh' ? '第 1 步：添加设备获取安装命令' : 'Step 1: Add device to get install command' }}
{{ locale === 'zh' ? '在「设备」页面点击「添加设备」，输入设备名称后系统会生成专属安装命令（包含 token 和 tenant-id）。' : 'Go to the "Devices" tab, click "Add Device", enter a name. The system generates an install command with your token and tenant-id.' }}
{{ locale === 'zh' ? '第 2 步：在目标机器上运行' : 'Step 2: Run on target machine' }}
# {{ locale === 'zh' ? '一键安装 Manager + Agent（在目标机器终端执行）' : 'One-click install Manager + Agent (run on target machine)' }}
curl -fsSL https://manage-inference.cn.cloudos.com/install.sh | bash -s -- \
--token <YOUR_TOKEN> \
--tenant-id <YOUR_TENANT_ID> \
--name "My Mac Studio"
curl -fsSL https://manage-inference.cn.cloudos.com/install.sh | bash -s -- \
--token <YOUR_TOKEN> \
--tenant-id <YOUR_TENANT_ID> \
--name "My Mac Studio"
{{ locale === 'zh' ? '安装过程' : 'What happens' }}
- 检查 Python 版本（需要 3.10+）
- 检查端口 8080 是否可用
- 下载安装包到
~/.mac-studio-agent/（含 Agent + Manager） - 创建 Python 虚拟环境并安装依赖
- 探测本地引擎（Ollama、LiteLLM、llama-server）
- 注册 2 个 launchd 服务（Manager + Agent，开机自启）
- 启动 Manager → 等待就绪 → 启动 Agent → 验证连接
- Checks Python version (3.10+ required)
- Checks port 8080 availability
- Downloads bundle to
~/.mac-studio-agent/(Agent + Manager) - Creates Python venv and installs dependencies
- Detects local engines (Ollama, LiteLLM, llama-server)
- Registers 2 launchd services (Manager + Agent, auto-start on boot)
- Starts Manager → waits ready → starts Agent → verifies connection
{{ locale === 'zh' ? '安装的服务' : 'Installed services' }}
| {{ locale === 'zh' ? '服务' : 'Service' }} | launchd Label | {{ locale === 'zh' ? '端口' : 'Port' }} | {{ locale === 'zh' ? '说明' : 'Description' }} |
|---|---|---|---|
| Manager | com.reai.mac-studio-manager | 8080 | {{ locale === 'zh' ? '管理 API 服务' : 'Management API service' }} |
| Agent | com.reai.mac-studio-agent | - | {{ locale === 'zh' ? 'WSS 隧道客户端' : 'WSS tunnel client' }} |
{{ locale === 'zh' ? '安装目录结构' : 'Installation directory structure' }}
~/.mac-studio-agent/
├── agent/ # WSS {{ locale === 'zh' ? '隧道客户端' : 'tunnel client' }}
├── manager/ # {{ locale === 'zh' ? '管理 API 服务' : 'Management API' }}
│ ├── app.py
│ ├── routers/
│ └── services/
├── relay/
│ └── protocol.py # {{ locale === 'zh' ? '通信协议' : 'Protocol' }}
├── venv/ # Python {{ locale === 'zh' ? '虚拟环境' : 'venv' }}
├── .env # {{ locale === 'zh' ? '配置审计副本' : 'Config audit copy' }}
└── litellm-config.yaml
├── agent/ # WSS {{ locale === 'zh' ? '隧道客户端' : 'tunnel client' }}
├── manager/ # {{ locale === 'zh' ? '管理 API 服务' : 'Management API' }}
│ ├── app.py
│ ├── routers/
│ └── services/
├── relay/
│ └── protocol.py # {{ locale === 'zh' ? '通信协议' : 'Protocol' }}
├── venv/ # Python {{ locale === 'zh' ? '虚拟环境' : 'venv' }}
├── .env # {{ locale === 'zh' ? '配置审计副本' : 'Config audit copy' }}
└── litellm-config.yaml
{{ locale === 'zh' ? '重复安装 = 升级' : 'Re-install = Upgrade' }}
{{ locale === 'zh' ? '重复执行安装命令会自动升级：停止旧服务 → 更新代码 → 保留配置和数据 → 重启。无需先卸载。' : 'Re-running the install command auto-upgrades: stop old services → update code → keep config and data → restart. No need to uninstall first.' }}
{{ locale === 'zh' ? '卸载' : 'Uninstall' }}
{{ locale === 'zh' ? '标准卸载（保留数据）' : 'Standard uninstall (keep data)' }}
curl -fsSL https://manage-inference.cn.cloudos.com/install.sh | bash -s -- --uninstall
{{ locale === 'zh' ? '停止 Manager + Agent 服务，删除程序文件，保留统计数据和日志。' : 'Stops Manager + Agent services, removes program files, keeps stats data and logs.' }}
{{ locale === 'zh' ? '彻底清理（含数据）' : 'Full purge (including data)' }}
curl -fsSL https://manage-inference.cn.cloudos.com/install.sh | bash -s -- --uninstall --purge-data
{{ locale === 'zh' ? '在标准卸载基础上，额外删除统计数据库、状态文件和日志。' : 'In addition to standard uninstall, also removes stats database, state files, and logs.' }}
{{ locale === 'zh' ? '手动卸载' : 'Manual uninstall' }}
# 1. {{ locale === 'zh' ? '停止服务' : 'Stop services' }}
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# 2. {{ locale === 'zh' ? '删除 plist 文件' : 'Remove plist files' }}
rm -f ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
rm -f ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# 3. {{ locale === 'zh' ? '删除安装目录' : 'Remove installation directory' }}
rm -rf ~/.mac-studio-agent
# 4. {{ locale === 'zh' ? '清理状态文件' : 'Clean up status file' }}
rm -f /tmp/wss-agent-status
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# 2. {{ locale === 'zh' ? '删除 plist 文件' : 'Remove plist files' }}
rm -f ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
rm -f ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# 3. {{ locale === 'zh' ? '删除安装目录' : 'Remove installation directory' }}
rm -rf ~/.mac-studio-agent
# 4. {{ locale === 'zh' ? '清理状态文件' : 'Clean up status file' }}
rm -f /tmp/wss-agent-status
{{ locale === 'zh' ? '卸载不影响' : 'Uninstall does NOT affect' }}
- {{ locale === 'zh' ? '本地推理引擎（Ollama、llama-server 等）' : 'Local inference engines (Ollama, llama-server, etc.)' }}
- {{ locale === 'zh' ? '已下载的模型文件' : 'Downloaded model files' }}
- {{ locale === 'zh' ? '云端账号和设备记录（需在面板中手动删除）' : 'Cloud account and device records (delete manually in panel)' }}
{{ locale === 'zh' ? '架构原理' : 'Architecture' }}
{{ locale === 'zh' ? '整体架构' : 'Overall Architecture' }}
{{ locale === 'zh'
? `浏览器 (任意网络)
│
▼ HTTPS
┌─────────────────────────────────┐
│ 云端 Nginx (cn.cloudos.com) │
│ ├─ /static/* → 前端静态文件 │
│ └─ /ws/* → WSS Relay │
└────────────┬────────────────────┘
│ WebSocket (wss://)
▼
┌─────────────────────────────────┐
│ WSS Relay Server (:14081) │
│ ├─ 认证 + 租户路由 │
│ ├─ HTTP → WS 消息转换 │
│ └─ 流式响应 (SSE) 透传 │
└────────────┬────────────────────┘
│ WebSocket (wss://)
│ (Agent 主动外连，穿透 NAT)
▼
┌─────────────────────────────────┐
│ WSS Agent (Mac Studio 本地) │
│ ├─ WS 消息 → HTTP 请求转换 │
│ ├─ 心跳保活 (15s) │
│ └─ 指数退避重连 (1s → 30s) │
└────────────┬────────────────────┘
│ HTTP (127.0.0.1)
▼
┌─────────────────────────────────┐
│ Manager API (:8080) │
│ └→ Ollama (:11435) │
│ └→ llama-server (:8000) │
│ └→ LiteLLM (:4000) │
└─────────────────────────────────┘`
: `Browser (any network)
│
▼ HTTPS
┌─────────────────────────────────┐
│ Cloud Nginx (cn.cloudos.com) │
│ ├─ /static/* → Frontend files │
│ └─ /ws/* → WSS Relay │
└────────────┬────────────────────┘
│ WebSocket (wss://)
▼
┌─────────────────────────────────┐
│ WSS Relay Server (:14081) │
│ ├─ Auth + tenant routing │
│ ├─ HTTP → WS message bridge │
│ └─ SSE streaming passthrough │
└────────────┬────────────────────┘
│ WebSocket (wss://)
│ (Agent connects out, NAT traversal)
▼
┌─────────────────────────────────┐
│ WSS Agent (Mac Studio local) │
│ ├─ WS message → HTTP bridge │
│ ├─ Heartbeat (15s) │
│ └─ Exponential backoff (1-30s) │
└────────────┬────────────────────┘
│ HTTP (127.0.0.1)
▼
┌─────────────────────────────────┐
│ Manager API (:8080) │
│ └→ Ollama (:11435) │
│ └→ llama-server (:8000) │
│ └→ LiteLLM (:4000) │
└─────────────────────────────────┘` }}
{{ locale === 'zh' ? '核心原理：WSS 中继' : 'Core: WSS Relay' }}
传统方案需要公网 IP 或 FRP 端口转发。WSS 中继采用了不同的方式：
Traditional setups require a public IP or FRP port forwarding. WSS relay takes a different approach:
- Agent 主动外连 — 本地 Agent 主动连接云端 Relay（就像浏览器访问网站一样），不需要任何入站端口
- WebSocket 长连接 — 连接建立后保持在线，Relay 把浏览器的 API 请求封装成 WS 消息发给 Agent
- Agent 本地转发 — Agent 把 WS 消息还原为 HTTP 请求，发给本地 Manager API (127.0.0.1:8080)
- 前端云端直出 — 管理面板的 HTML/JS 直接由云端提供，只有 API 请求才通过隧道
- Agent connects outbound — The local Agent initiates a connection to the cloud Relay (like a browser visiting a website), no inbound ports needed
- WebSocket persistent connection — Once connected, the Relay wraps browser API requests as WS messages and sends them to the Agent
- Agent local forwarding — The Agent converts WS messages back to HTTP requests and sends them to the local Manager API (127.0.0.1:8080)
- Frontend served from cloud — The management panel HTML/JS is served directly from the cloud; only API requests go through the tunnel
{{ locale === 'zh' ? '安全机制' : 'Security' }}
| {{ locale === 'zh' ? '层级' : 'Layer' }} | {{ locale === 'zh' ? '机制' : 'Mechanism' }} |
|---|---|
| {{ locale === 'zh' ? '传输层' : 'Transport' }} | TLS/HTTPS + WSS ({{ locale === 'zh' ? '全链路加密' : 'end-to-end encrypted' }}) |
| {{ locale === 'zh' ? '用户认证' : 'User Auth' }} | OAuth 2.0 PKCE ({{ locale === 'zh' ? '外脑登录' : 'external provider' }}) |
| {{ locale === 'zh' ? '设备认证' : 'Device Auth' }} | Per-tenant Agent Token ({{ locale === 'zh' ? '哈希存储' : 'hashed storage' }}) |
| API {{ locale === 'zh' ? '认证' : 'Auth' }} | API Key ({{ locale === 'zh' ? '绑定租户' : 'tenant-bound' }}) |
| {{ locale === 'zh' ? '本地引擎' : 'Local Engines' }} | {{ locale === 'zh' ? '仅监听 127.0.0.1，不暴露外网' : 'Listen on 127.0.0.1 only, not exposed' }} |
{{ locale === 'zh' ? '故障排查' : 'Troubleshooting' }}
{{ locale === 'zh' ? '安装失败：Python 版本不够' : 'Install failed: Python version too low' }}
{{ locale === 'zh' ? '错误信息' : 'Error' }}:
{{ locale === 'zh' ? '解决：' : 'Fix:' }}
{{ locale === 'zh' ? '或从 python.org 下载安装' : 'Or download from python.org' }}
{{ locale === 'zh' ? '需要 Python 3.10+，当前版本 3.9' : 'Requires Python 3.10+, current version 3.9' }}{{ locale === 'zh' ? '解决：' : 'Fix:' }}
brew install python@3.12{{ locale === 'zh' ? '或从 python.org 下载安装' : 'Or download from python.org' }}
{{ locale === 'zh' ? '安装失败：agent-bundle 下载 404' : 'Install failed: agent-bundle 404' }}
{{ locale === 'zh' ? '错误信息' : 'Error' }}:
{{ locale === 'zh' ? '原因：云端服务未部署 agent-bundle。联系管理员重新部署。' : 'Cause: Cloud server has not deployed the agent-bundle. Contact the administrator to redeploy.' }}
curl: (22) The requested URL returned error: 404{{ locale === 'zh' ? '原因：云端服务未部署 agent-bundle。联系管理员重新部署。' : 'Cause: Cloud server has not deployed the agent-bundle. Contact the administrator to redeploy.' }}
{{ locale === 'zh' ? 'Agent 状态一直是 unknown' : 'Agent status stays unknown' }}
{{ locale === 'zh' ? '查看日志：' : 'Check logs:' }}
{{ locale === 'zh' ? '常见原因：' : 'Common causes:' }}
{{ locale === 'zh' ? '• Token 错误 → 重新在面板生成安装命令' : '• Wrong token → Re-generate install command in panel' }}
{{ locale === 'zh' ? '• 网络不通 → 确认可以访问 manage-inference.cn.cloudos.com' : '• Network issue → Verify access to manage-inference.cn.cloudos.com' }}
{{ locale === 'zh' ? '• Python 依赖缺失 → 重新安装' : '• Missing Python deps → Reinstall' }}
tail -f ~/Library/Logs/mac-studio-agent.err.log{{ locale === 'zh' ? '常见原因：' : 'Common causes:' }}
{{ locale === 'zh' ? '• Token 错误 → 重新在面板生成安装命令' : '• Wrong token → Re-generate install command in panel' }}
{{ locale === 'zh' ? '• 网络不通 → 确认可以访问 manage-inference.cn.cloudos.com' : '• Network issue → Verify access to manage-inference.cn.cloudos.com' }}
{{ locale === 'zh' ? '• Python 依赖缺失 → 重新安装' : '• Missing Python deps → Reinstall' }}
{{ locale === 'zh' ? 'Agent 频繁断连重连' : 'Agent keeps disconnecting and reconnecting' }}
{{ locale === 'zh' ? 'Agent 自带指数退避重连（1s → 30s），断线后会自动恢复。如果频繁发生：' : 'Agent has built-in exponential backoff reconnection (1s → 30s). If it happens frequently:' }}
{{ locale === 'zh' ? '• 检查网络稳定性' : '• Check network stability' }}
{{ locale === 'zh' ? '• 确认云端 Relay 服务正常运行' : '• Verify cloud Relay service is running' }}
{{ locale === 'zh' ? '• 检查网络稳定性' : '• Check network stability' }}
{{ locale === 'zh' ? '• 确认云端 Relay 服务正常运行' : '• Verify cloud Relay service is running' }}
{{ locale === 'zh' ? '常用运维命令' : 'Useful maintenance commands' }}
# {{ locale === 'zh' ? '查看 Agent 状态' : 'Check Agent status' }}
cat /tmp/wss-agent-status
# {{ locale === 'zh' ? '查看 Manager 健康' : 'Check Manager health' }}
curl -s http://127.0.0.1:8080/health
# {{ locale === 'zh' ? '查看日志' : 'View logs' }}
tail -f ~/Library/Logs/mac-studio-manager.err.log
tail -f ~/Library/Logs/mac-studio-agent.err.log
# {{ locale === 'zh' ? '重启服务' : 'Restart services' }}
launchctl kickstart -k gui/$(id -u)/com.reai.mac-studio-manager
launchctl kickstart -k gui/$(id -u)/com.reai.mac-studio-agent
# {{ locale === 'zh' ? '停止服务' : 'Stop services' }}
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# {{ locale === 'zh' ? '启动服务' : 'Start services' }}
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
cat /tmp/wss-agent-status
# {{ locale === 'zh' ? '查看 Manager 健康' : 'Check Manager health' }}
curl -s http://127.0.0.1:8080/health
# {{ locale === 'zh' ? '查看日志' : 'View logs' }}
tail -f ~/Library/Logs/mac-studio-manager.err.log
tail -f ~/Library/Logs/mac-studio-agent.err.log
# {{ locale === 'zh' ? '重启服务' : 'Restart services' }}
launchctl kickstart -k gui/$(id -u)/com.reai.mac-studio-manager
launchctl kickstart -k gui/$(id -u)/com.reai.mac-studio-agent
# {{ locale === 'zh' ? '停止服务' : 'Stop services' }}
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
# {{ locale === 'zh' ? '启动服务' : 'Start services' }}
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-manager.plist
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.reai.mac-studio-agent.plist
API {{ locale === 'zh' ? '参考' : 'Reference' }}
{{ locale === 'zh' ? '认证方式' : 'Authentication' }}
{{ locale === 'zh' ? '推理 API 和 Audio API 共用同一套 Key，支持两种认证方式：' : 'Inference API and Audio API share the same key and support two auth methods:' }}
# {{ locale === 'zh' ? '方式一：X-API-Key 头' : 'Method 1: X-API-Key header' }}
curl -H "X-API-Key: YOUR_KEY" {{ publicBases.inference }}/v1/models
curl -H "X-API-Key: YOUR_KEY" {{ publicBases.audio }}/v1/files
# {{ locale === 'zh' ? '方式二：Bearer Token' : 'Method 2: Bearer Token' }}
curl -H "Authorization: Bearer YOUR_KEY" {{ publicBases.inference }}/v1/models
curl -H "Authorization: Bearer YOUR_KEY" {{ publicBases.audio }}/v1/files
curl -H "X-API-Key: YOUR_KEY" {{ publicBases.inference }}/v1/models
curl -H "X-API-Key: YOUR_KEY" {{ publicBases.audio }}/v1/files
# {{ locale === 'zh' ? '方式二：Bearer Token' : 'Method 2: Bearer Token' }}
curl -H "Authorization: Bearer YOUR_KEY" {{ publicBases.inference }}/v1/models
curl -H "Authorization: Bearer YOUR_KEY" {{ publicBases.audio }}/v1/files
{{ locale === 'zh' ? '推理 API（OpenAI + Anthropic 兼容）' : 'Inference API (OpenAI + Anthropic-compatible)' }}
| {{ locale === 'zh' ? '端点' : 'Endpoint' }} | {{ locale === 'zh' ? '说明' : 'Description' }} |
|---|---|
GET /v1/models | {{ locale === 'zh' ? '列出可用模型' : 'List available models' }} |
POST /v1/chat/completions | {{ locale === 'zh' ? '对话补全（支持流式）' : 'Chat completion (supports streaming)' }} |
POST /v1/messages | {{ locale === 'zh' ? 'Anthropic Messages / Claude Messaging 兼容对话（支持流式）' : 'Anthropic Messages / Claude Messaging-compatible chat (supports streaming)' }} |
{{ locale === 'zh' ? '可用模型以 GET /v1/models（使用你的 Key）实时返回为准：不同 API Key 绑定不同设备/租户，看到的列表不同；模型 id 大小写敏感（如 glm-5.2 为全小写）。下方示例中的模型名仅为演示，请替换为你的列表中的任一 id。' : 'Available models are returned live by GET /v1/models with YOUR key: different keys are bound to different devices/tenants and see different lists; model ids are case-sensitive (e.g. glm-5.2 is all lowercase). Model names in the examples below are illustrative only — replace with any id from your list.' }}
{{ locale === 'zh' ? '示例：OpenAI Chat 请求' : 'Example: OpenAI Chat request' }}
curl -X POST {{ publicBases.inference }}/v1/chat/completions \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-flash",
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-flash",
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'
{{ locale === 'zh' ? '示例：Anthropic Messages 请求' : 'Example: Anthropic Messages request' }}
curl -X POST {{ publicBases.inference }}/v1/messages \
-H "x-api-key: YOUR_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-flash",
"max_tokens": 256,
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'
-H "x-api-key: YOUR_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-flash",
"max_tokens": 256,
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'
{{ locale === 'zh' ? 'Audio API（OpenAI 兼容）' : 'Audio API (OpenAI-compatible)' }}
{{ locale === 'zh' ? 'Audio API 公网入口暴露 /v1/files 和 /v1/audio/transcriptions；ASR 模型列表在管理面板的 Audio API 路由中查看。POST /v1/audio/transcriptions 中 model 可选，省略时使用当前 active ASR。' : 'The public Audio API exposes /v1/files and /v1/audio/transcriptions. ASR model routes are listed in the Manager Audio API routes. In POST /v1/audio/transcriptions, model is optional and falls back to the active ASR.' }}
| {{ locale === 'zh' ? '端点' : 'Endpoint' }} | {{ locale === 'zh' ? '说明' : 'Description' }} |
|---|---|
GET /v1/files | {{ locale === 'zh' ? '列出已上传文件' : 'List uploaded files' }} |
POST /v1/files | {{ locale === 'zh' ? '上传音频文件，返回 file_id' : 'Upload an audio file and return file_id' }} |
POST /v1/audio/transcriptions | {{ locale === 'zh' ? '音频转文字；model 可省略，命中当前 active ASR' : 'Audio transcription; model is optional and uses the current active ASR' }} |
{{ locale === 'zh' ? '示例：语音转录请求（file_id 两阶段）' : 'Example: Audio transcription request (two-stage file_id flow)' }}
curl -X POST {{ publicBases.audio }}/v1/files \
-H "Authorization: Bearer YOUR_KEY" \
-F "purpose=user_data" \
-F "file=@sample.wav"
curl -X POST {{ publicBases.audio }}/v1/audio/transcriptions \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"file_id": "file-xxx",
"model": "sherpa-onnx/sense-voice-small",
"language": "auto",
"response_format": "json"
}'
-H "Authorization: Bearer YOUR_KEY" \
-F "purpose=user_data" \
-F "file=@sample.wav"
curl -X POST {{ publicBases.audio }}/v1/audio/transcriptions \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"file_id": "file-xxx",
"model": "sherpa-onnx/sense-voice-small",
"language": "auto",
"response_format": "json"
}'
{{ locale === 'zh'
? '调试建议：Models 页上半部分会显示当前已暴露的 Audio API 路由；点击“测试”会直接带着对应 model 跳到 Playground 的 Audio 模式。'
: 'Tip: the Models page now shows currently exposed Audio API routes in the top section. Click "Test" to open Playground in Audio mode with the matching model preselected.'
}}
{{ toast.message }}