Files
Piik/docs/project-memory.md
T
AAA a9b7507488 feat(app): add camera sharing and Host microphone controls
## Problem and outcome / 问题与结果

Add camera sharing and Host microphone commentary to the existing Host workflow,
with shared source cards, device selection, input gain and grouped sharing settings.
Browser and native capture keep one outgoing audio track and the existing room/media
owners. Include optional password-free invitations, public-tunnel protocol fallback,
capture retirement and App-discovery feedback fixes. This is the complete capture
maintenance phase; interaction, x86 capture and 4K experiments are excluded.

## Release notes

### 中文

这次更新让分享多了两种表达方式:用摄像头展示现场,用麦克风边分享边讲解。
分享设置也重新整理,并修复了公网邀请、音视频切换和采集退出中的一些问题。

#### 新功能与使用改进

- **摄像头分享**:在原有选源器中选择摄像头,支持设备选择和画面预览;支持摄像头访问的手机浏览器也可使用。
- **房主麦克风**:浏览器分享和 App 原生采集均可混入房主声音,支持选择麦克风、静音和输入音量调节。观众仍使用原有音量控件收听。
- **更集中的分享设置**:画质、声音和详细连接选项集中在画面下方,常用操作保持直接可用,并调整了窄屏播控与等待提示的布局。
- **邀请链接可选择是否免密加入**:关闭后,链接不再附带免密邀请;接收者按当前站点和房间的准入设置加入。
- **更清晰的启动设置**:可选站点口令默认收起,已有站点地址提供 `demo.piik.tv` 建议,并澄清 NAT 增强选项的含义。

#### 修复

- 公网邀请在 UDP 受限但 TCP 可用的网络中,可使用隧道组件原有的协议回退,减少这类启动失败。
- 调整音频时保留健康的视频发送;修复旧采集事件干扰当前分享的情况。
- 停止或取消原生采集时,先正常释放采集会话,超时才强制退出;修复本地复现的强制终止后 Windows 黄框残留问题。
- 从 App 打开的页面无法连接 App 时,选源器保留明确提示,不再只剩浏览器和摄像头选项而缺少原因说明。

#### 使用说明

摄像头和音频能力取决于系统、浏览器与设备。Windows 仍可能因权限或其他采集要求显示边框。
使用新增的 App 麦克风功能,请更新 App;连接自部署站点时,也请更新站点以获得对应页面功能。

### English

This update adds camera sharing and Host microphone commentary, alongside a
clearer sharing-settings layout. It also fixes several public-invitation,
audio/video transition and capture-shutdown issues.

#### Features and usability

- **Camera sharing:** choose a camera in the existing source picker, with device selection and thumbnail previews. Phone browsers with camera access can share too.
- **Host microphone:** add commentary to Browser sharing or App-native capture, with device selection, mute and input gain. Viewers listen through their existing volume controls.
- **Grouped sharing settings:** picture quality, sound and detailed connection options sit below the picture, while frequent actions remain directly available. Narrow-screen playback and waiting layouts are refined.
- **Optional password-free invitations:** turn this off to copy a link without an invitation credential. Recipients then follow the site's and room's current admission settings.
- **Clearer startup settings:** optional site-password settings are collapsed, the existing-site field suggests `demo.piik.tv`, and the NAT enhancement option has a clearer description.

#### Fixes

- Public invitations can use the tunnel component's existing protocol fallback when UDP is blocked but TCP is available, reducing startup failures on those networks.
- Audio changes preserve healthy video sending. Retired capture events no longer affect an active share.
- Stopping or cancelling native capture releases the capture session before enforcing the shutdown deadline. This fixes a locally reproduced case where forced termination left a Windows capture border behind.
- Pages opened from the App keep a clear connection warning in the source picker when the App cannot be reached, instead of showing only Browser and Camera without an explanation.

#### Usage notes

Camera and audio support depend on the OS, browser and device. Windows may still
require a border because of permissions or another active capture.
Update the App to use its new microphone feature; update self-hosted sites as
well to receive the matching page controls.

## Verification / 验证

- npm run check: 74 files / 1054 tests passed, TypeScript and production Web build passed.
- npm run check:go: Go tests, vet, target builds and Windows native compile/regressions passed.
- npm run gate:app-native-host: actual Windows H264 Host/Viewer, live/paused profile changes,
  quiet-source recovery, source replacement/failure, subsequent share and cleanup passed.
- Windows 11 capture-border check: parent cancellation retires the actual WGC session and
  clears its border; subprocess tests verify normal, concurrent and timed-out retirement.
- Website build, scripts typecheck, documentation and repository hygiene passed.
- Native macOS/Linux packaging and Linux race checks run in CI before publication.
  Physical phone camera and real audio-level/echo/multiple-device coverage remain recorded
  limits; the owner authorized release with those limits. No universal NAT/border fix is claimed.

## Impact / 影响

Compatible feature release **v1.6.0**, explicitly approved by the owner. Browser/Server v23,
Native control v9 and capture v7 stay unchanged; new App microphone commands use capability
negotiation. No new dependency, port, media route, room lifecycle or storage migration.
Published Server cutover retains configuration and room storage and uses bounded health
checks with the previous executable available for rollback. Source, tests and owned
documentation ship together; exact artifact/deployment identities remain outside the repo.
2026-09-21 18:34:56 +08:00

6.9 KiB

Project Memory

Last updated: 2026-09-21

Piik is private, low-latency screen sharing for one Host and up to 20 authenticated friends. The current product surface is Web Host, Web Viewer, and Browser relay; desktop and mobile Browsers are Viewer targets, and camera-capable mobile Browsers can also host. It is not a public broadcast service. Hosted and App run one shared Go core (ADR-0012); Node and Vite build the Browser UI and are not a runtime. Source includes a cross-platform App that runs that same core in a self-contained Local deployment or opens one configured Site through the system Browser. Its single Go process presents Local, temporary public-link, and saved-Site choices on every launch, starts one process-level loopback capability service, owns the Local room authority in-process when selected, and owns the explicit native media path chosen in the same Web UI. Room source and media implementation are independent: a topology may mix Browser and Native peers without changing participant, signaling, capacity, or route identity. Local is LAN-first; explicit one-link mode exposes that same authority through a temporary public control tunnel while media remains P2P. Host authority belongs to a tab, with one separate persistent origin resume hint. The App admits two independent native control sessions under its one loopback service.

The current source embeds Binding-only STUN and optional SFU forwarding in the Hosted Go process. Native and SFU use one shared Pion/LiveKit media adapter; Native parents reuse suitable encoded outputs and derive a missing lower output only for direct-child demand. Source-owned groups retain the highest needed output and lower fallbacks, while each child receives its selected output. Native Host SFU publication reuses the encoded source directly; its loopback Browser edge owns preview only. Eligible Browser capture and relay now share independent local WebRTC producers for compatible direct children under ADR-0014; unsupported APIs and failed pooling retain ordinary senders. Windows, macOS and Linux producers share one encoded-frame boundary and Host quality controls. Source replacement retains route and connection identity; lower-output adaptation preserves the original and higher sibling outputs. Status owns the remaining physical and release acceptance boundary; runtime release descriptors own deployed identity. The Local launcher keeps an optional user-chosen site-access password; leaving it blank keeps the self-contained site open.

Product Map

  • Rooms and access owns room codes, invitations, code entry, credentials, and lightweight/SQLite persistence.
  • Routing and transport owns P2P-first distribution, endpoint capacity, SFU fallback, recovery, and privacy limits.
  • Capture and media quality owns Browser capture, profiles, codec selection, audio, and framework adaptation.
  • Presentation and lifecycle owns Host and Viewer workflows, playback authority, roster, state, and Browser lifecycle.

Non-obvious decisions live in ADRs; reproducible evidence and platform limits live in research. Source code and tests own self-evident implementation and routine UI detail.

Core Invariants

  • Rooms use random four-digit codes, exact Host ownership, room-scoped Viewer invitations and independent open | private code entry. Room authority ends through explicit replacement/deletion, with separate invitation rotation or revocation. Hosted defaults to SQLite, with explicit process-memory opt-out; App Local remains process-only. Site access has a separate 24-hour idle cookie lifetime. Production persists SQLite room authority; live participants, routes, and media remain process-only.
  • Media is automatic and distributed. Ordinary peers are STUN-only and prefer direct/peer UDP. A deployment may enable same-host auxiliary STUN for bounded, room-wide, connection-local NAT prediction; its Host switch defaults on when available, and anonymous provenance distinguishes ordinary from predicted selected paths. The configuration default remains stock ICE. The only application fallback is one bounded embedded SFU/UDP publication; Piik configures no TURN or TCP media route. A pre-share peer-only policy can exclude that SFU suffix for one share generation.
  • Every endpoint shares one steady outbound-copy cap, default 2 and limited to 1..3. One room controller owns one committed graph, one reconcile loop, and one serial child operation. First decoded frame commits availability work; accepted quality work adds native edge proof before commit in that operation.
  • Browser WebRTC and the shared Pion/LiveKit media adapter own congestion control, media adaptation, pacing and layer selection. Parent demand is local to direct children and never a room-wide minimum or topology rebalance. Current source defaults persistent native-edge local P2P convergence on, with a pre-share Host opt-out. SFU quality work is limited to multi-root Host fanout relief and each Viewer must prove its own candidate non-regression before commit. A persistently limited edge with no different Peer parent may use one same-parent connection regeneration through that same operation; its exact post-commit sender identity damps only another same-edge regeneration. No weighted score, all-pairs probe, general parent-wide prediction, or periodic rebalance is accepted.
  • Browser video uses motion; each share chooses H.264 with VP8 fallback through an actual sender probe unless the Host explicitly selects VP8 or H264. Screen audio uses 64/128/192 kbps ceilings with 128 default.
  • Viewer playback uses one native video element. Presentation and recovery are fenced by current route/media identity and current-frame proof; hidden-page time alone is not media-failure evidence or a keepalive guarantee.

Current Snapshot

The current contract tuple is owned by status. Embedded SFU SDP/ICE and demand travel over the authenticated room WebSocket, without an external room service or media token. Committed first-frame readiness, bounded candidate-relative progress, durable room authority and the H.264/VP8 sender gate remain. Status indexes source and deployment evidence; protocol rollback requires matching Browser and App artifacts, not only an application symlink.

Current work is owned by the TODO ledger. Open physical evidence is owned by verification status. Environment and initial services are owned by configuration and self-hosting; application release is owned by deployment, with exact deployed identity retained in the release descriptor and deployment record. Git history and pull requests own completed timelines.