Voicecan Device Platform is a self-hosted platform for connecting Voicecan-compatible devices, managing recordings, and giving applications secure access to device data. It can run at the edge, on a private server, or as a containerized service without depending on Voicecan accounts, family services, model services, or the existing business API.
- Connect devices — provision, claim, monitor, and control devices through a browser or WebSocket connection.
- Manage recordings — discover recordings, synchronize files reliably, resume interrupted transfers, and keep immutable file history.
- Control access — organize users and devices into Groups, issue scoped application credentials, and review an audit trail of administrative activity.
- Build applications — use REST, TypeScript, Python, MCP, Webhooks, or the reusable Device Connect Web to connect your own products.
- Run device operations — manage local firmware packages, perform OTA updates, inspect device status, and operate through Admin or the CLI.
- Choose storage — use local files for an edge installation or connect an object-storage-backed deployment for shared workloads.
- Operate with confidence — use backups, restore tools, quotas, delivery retries, dead-letter inspection, health checks, and metrics as part of daily operations.
| Surface | Use it for |
|---|---|
| Admin | Setup, users, Groups, applications, devices, recordings, firmware, Webhooks, and operational tools |
| Device Connect Web | Public browser-based device provisioning and connection flows |
| REST API | Integrate devices, recordings, applications, permissions, and Webhooks into your product |
| TypeScript / Python clients | Build server-side and automation workflows with typed client libraries |
| MCP | Connect an AI assistant or automation tool to service status, applications, and device workflows |
| CLI | Install, configure, diagnose, back up, restore, and operate a self-hosted instance |
| Option | Best for |
|---|---|
npm + onboard |
A quick local Edge installation with a user-level service |
| Docker + Compose | A repeatable server or private-network deployment |
| Native Node runtime | A host installation without Docker |
| Source checkout | Development and customization |
Choose one installation method for each data directory. See installation and background services for the detailed setup guides.
Requirements: Node.js >=24.15.0 <25, npm, and a private persistent data directory.
npm install --global @voicecan/[email protected]
voicecan-device onboardThe onboarding flow creates the local Profile, prepares the database, starts the service, waits for readiness, and opens Admin. Use these commands to inspect the service afterwards:
voicecan-device service status --output json
voicecan-device doctor --output jsonInstall curl, Git, Docker Engine, and Docker Compose v2, then run:
curl -fsSL https://raw.githubusercontent.com/voicecan/device-platform/main/install.sh | bashThe installer prepares the service and opens the local Admin setup flow. For an existing installation, follow the version and migration guide.
Copy the prompt below into your AI coding or automation assistant to install the platform and connect an application through its public interfaces:
You are integrating Voicecan Device Platform from https://github.com/voicecan/device-platform into the current environment.
Use the current user request as the integration goal. Before acting, read and follow the repository Skills that match the task:
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-install/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-configure/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-bind-device/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-create-application/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-connect-mcp/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-operate/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/integrate-voicecan-device/SKILL.md
Inspect the environment and any existing installation first. Use only the public installation flow and public REST, SDK, Webhook, and MCP interfaces. Preserve existing data and configuration. Follow the selected Skills for dry runs, approvals, credentials, device actions, verification, and reporting.
Never expose or read secrets, passwords, setup tokens, Wi-Fi credentials, temporary URLs, production recordings, or private protocol sources. Ask before any destructive operation, credential creation, DNS or TLS change, cloud-storage change, or real device action. At the end, report the service URL, commands, checks, manual steps, and rollback plan without secrets.
Install the packages you need from the @voicecan scope:
npm install @voicecan/contracts @voicecan/server-client @voicecan/connector-runtime@voicecan/contracts— public data types, event names, and constants.@voicecan/server-client— REST access, event cursors, secure Recording downloads, Webhook verification, and media helpers.@voicecan/connector-runtime— durable Webhook delivery, event handling, and Recording reconciliation.
Install @voicecan/device-platform when you are deploying the self-hosted Server, Admin, and CLI.
- Open Platform — applications, permissions, credentials, REST, and administration.
- Quickstart — installation, setup, device connection, and recording synchronization.
- Device Connect Web — browser-based provisioning and connection.
- Local firmware repository and OTA — firmware packages and device updates.
- Recording Download Grants — secure application downloads.
- MCP server — connect MCP clients and AI tools.
- Operations runbook — deployment, backup, restore, and monitoring.
- Security model — authentication, authorization, storage, and Webhook security.
Voicecan Device Platform 是一个自托管设备平台，用于连接 Voicecan 兼容设备、管理录音，并为应用提供安全的数据访问能力。它可以运行在边缘设备、私有服务器或容器化环境中，不依赖 Voicecan 账户、家庭服务、模型服务或现有业务 API。
- 连接设备：通过浏览器或 WebSocket 完成配网、认领、状态查看和设备控制。
- 管理录音：发现录音、可靠同步文件、恢复中断传输，并保留不可变文件历史。
- 管理权限：使用 Group 组织用户和设备，创建有范围的应用凭据，并查看管理审计记录。
- 构建应用：通过 REST、TypeScript、Python、MCP、Webhook 或可复用的 Device Connect Web 接入自己的产品。
- 执行设备运维：管理本地固件包、执行 OTA、查看设备状态，并通过 Admin 或 CLI 操作平台。
- 选择存储：边缘安装使用本地文件，共享部署可连接对象存储。
- 持续运营：使用备份、恢复、配额、投递重试、Dead Letter 检查、健康检查和指标完成日常运维。
| 入口 | 用途 |
|---|---|
| Admin | Setup、用户、Group、Application、设备、录音、固件、Webhook 和运维工具 |
| Device Connect Web | 公开浏览器设备配网和连接流程 |
| REST API | 将设备、录音、Application、权限和 Webhook 接入自己的产品 |
| TypeScript / Python Client | 构建服务端集成和自动化流程 |
| MCP | 将 AI 助手或自动化工具连接到服务状态、Application 和设备流程 |
| CLI | 安装、配置、诊断、备份、恢复和操作自托管实例 |
| 方式 | 适合场景 |
|---|---|
npm + onboard |
快速安装带用户级后台服务的本地 Edge 实例 |
| Docker + Compose | 可重复的服务器或私有网络部署 |
| 专用 Node Runtime | 不使用 Docker 的主机安装 |
| 源码检出 | 开发和定制 |
每个数据目录只选择一种安装方式。详细说明见安装与后台持续运行。
要求 Node.js >=24.15.0 <25、npm，以及一个私有持久化数据目录。
npm install --global @voicecan/[email protected]
voicecan-device onboardOnboarding 会创建本地 Profile、准备数据库、启动服务、等待就绪并打开 Admin。之后可使用以下命令查看服务：
voicecan-device service status --output json
voicecan-device doctor --output json先安装 curl、Git、Docker Engine 和 Docker Compose v2，再运行：
curl -fsSL https://raw.githubusercontent.com/voicecan/device-platform/main/install.sh | bash安装器会准备服务并打开本地 Admin Setup 页面。已有实例升级请参考版本与迁移。
将下面的 Prompt 复制给 ChatGPT、Codex、Claude 或其他 AI 编程/自动化助手，即可让它按公开接口完成平台安装和应用接入：
你正在把 https://github.com/voicecan/device-platform 接入当前环境。
把用户当前请求作为接入目标。执行前先读取并遵守与任务相关的仓库 Skills：
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-install/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-configure/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-bind-device/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-create-application/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-connect-mcp/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/voicecan-operate/SKILL.md
https://github.com/voicecan/device-platform/blob/main/skills/integrate-voicecan-device/SKILL.md
先检查环境和已有安装。只使用公开安装流程以及公开 REST、SDK、Webhook 和 MCP 接口。保留已有数据和配置，并按照对应 Skill 执行 dry-run、确认、凭据、设备操作、验证和结果报告。
不得读取或暴露 Secret、密码、Setup Token、Wi-Fi 凭据、临时 URL、生产录音或私有协议源码。破坏性操作、创建凭据、DNS/TLS 变更、云存储变更或真实设备操作前必须先询问确认。完成后报告服务 URL、命令、检查结果、人工步骤和回滚方案，报告中不得包含 Secret。
按需安装 @voicecan Scope 下的应用包：
npm install @voicecan/contracts @voicecan/server-client @voicecan/connector-runtime@voicecan/contracts：公开数据类型、事件名称和常量。@voicecan/server-client：REST 访问、事件游标、安全 Recording 下载、Webhook 校验和媒体辅助工具。@voicecan/connector-runtime：持久化 Webhook 投递、事件处理和 Recording 对账。
部署自托管 Server、Admin 和 CLI 时安装 @voicecan/device-platform。
- Open Platform：Application、权限、凭据、REST 和管理。
- 快速开始：安装、Setup、设备连接和录音同步。
- Device Connect Web：浏览器设备配网和连接。
- 本地固件仓库与 OTA：固件包和设备更新。
- Recording Download Grant：安全的应用下载。
- MCP Server：连接 MCP Client 和 AI 工具。
- 运维手册：部署、备份、恢复和监控。
- 安全模型：认证、授权、存储和 Webhook 安全。