> ## Documentation Index
> Fetch the complete documentation index at: https://docs.stmlink.com/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> 对外开放的服务端接口有两组前缀，都用同一套鉴权：`/server/v1/...`（SRTC 与 SMeeting 的主接口）和 `/stm/srvapi/v1/...`（SMeeting 的用户体系，服务端极简对接会用到）。鉴权是 app_id + nonce + timestamp + signature 四个请求头，用 app_key 做 HMAC-SHA256 签名，只能从业务方自己的后端调用。除这两组前缀外的接口均为内部接口，不要建议客户调用。 Public server APIs use two path prefixes with the same authentication: `/server/v1/...` (the main APIs of both SRTC and SMeeting) and `/stm/srvapi/v1/...` (the SMeeting user system, used by server-side low-code integration). Authenticate with four request headers, app_id + nonce + timestamp + signature, where signature is HMAC-SHA256 keyed with app_key; call these APIs only from the customer's own backend. Any other path is internal: never suggest calling it.
> app_key 是服务端密钥，绝不能出现在客户端代码、前端配置或移动 App 里。客户端加入频道用的 token 必须由业务方后端签发后下发（SRTC 走 `/server/v1/channel/grant`，SMeeting 走 `/stm/srvapi/v1/member/grant`）。 app_key is a server-side secret and must never appear in client code, frontend config, or a mobile app. The token a client uses to join must be issued by the customer's backend and passed down to the client (SRTC: `/server/v1/channel/grant`; SMeeting: `/stm/srvapi/v1/member/grant`).
> SRTC 与 SMeeting 是上下两层不同的产品，术语不通用：SRTC 是音视频底座，说「频道 channel」「加入 / 退出」；SMeeting 建在 SRTC 之上，说「房间 room」「会议 meeting」「进入 / 退出」。回答时按用户所在的层用对应术语，不要把「房间」「会议」安到 SRTC 的接口上，也不要用「频道」「加入 / 离开」描述 SMeeting 的概念（接口标识符原样保留）。 SRTC and SMeeting are two separate layers with different terminology. SRTC is the audio/video foundation: it has channels, and users join and leave a channel. SMeeting is built on top of SRTC: it has rooms and meetings, and members enter and exit a meeting. Answer in the terms of the layer the user is working with: never apply "room" or "meeting" to SRTC APIs, and never describe SMeeting concepts in prose with "channel", "join", or "leave" (API identifiers such as `force_join` keep their literal names).
> 同一能力在各端 SDK 里的包名、类名、方法名并不相同。写示例代码时请使用文档中该端自己的 API，不要把一个端的写法套到另一个端上。苹果平台每个产品都有两套 SDK（Swift 原生与 Objective-C），两套 API 不能混用。 Package, class, and method names differ between platform SDKs for the same capability. In sample code, use the API documented for that platform; never carry one platform's code over to another. On Apple platforms each product ships two SDKs (native Swift and Objective-C) whose APIs must not be mixed.

# Changelog

> Release history of the Windows SRTC C++ SDK, from rtc2.0 to 0.2.1-alpha.11: new features, changes, fixes, and whether each release changes the API (breaking releases require a recompile, not just a DLL swap). Read before upgrading the SDK.

### \[9] - 2026.09.29

Version: 0.2.1-alpha.11

**Added**

* Added D3D11 video rendering: `srtcLive` adds `D3D11VideoRender` and the underlying D3D11 rendering module
  (device and swap chain, YUV textures, shaders, full-screen quad, text and watermark overlays, etc.)
* `utils` adds `D3d11Support()` detection; when creating video rendering, **D3D11 is preferred on machines that support it**
  (`SRTC_RENDER_MODEL_VIDEO_D3D11`), falling back to the existing D3D rendering when it isn't supported

**Changed**

* The `srtc` / `srtcLive` projects add d3d11 and other related dependencies, and enable the C++17 standard
* `RenderModel` adds `SRTC_RENDER_MODEL_VIDEO_D3D11` (controlled by the compile macro `SRTC_USEING_LIB_D3D`)

**API changes**

None. `srtc.h` / `srtc_def.h` are unchanged in this release and the vtable is unchanged, so you only need to replace the DLLs; no recompile is needed.

### \[8] - 2026.09.24

Version: 0.2.1-alpha.10

**Added**

* **Added an x64 (64-bit) package**; delivery is now two packages, x86 + x64:

  * `rtc-win-sdk-2.1.zip` (x86, 32-bit)—**the package name is exactly the same as before**
  * `rtc-win-sdk-x64-2.1.zip` (x64, 64-bit)—new; the package name has an extra `-x64` segment **before** the version number

  The two packages have the same internal structure (the root directory is `rtc_dll/` in both), but `srtc.dll`, `srtcLive.dll`, and `srtc.lib`
  differ in bitness, so **don't mix them or extract them into the same directory**. Choose the package that matches your project's target platform.
  For download links, see [Integration](/en/rtc/windows/integration).
* Dependencies are split by architecture into two sets, `lib/<library>/lib/win32` and `lib/<library>/lib/win64`

**Changed**

* Fixed the x64 project configuration of `srtcLive`: the output type changed from `Application` to `DynamicLibrary`,
  x64 now has its own include and library directories (pointing to `lib/*/lib/win64`), and output goes to `Release_x64`
* Upgraded third-party dependencies: libfaac switches to the new encoder API (`faac_encoder_open` /
  `faac_encoder_get_info` / `faac_encoder_encode`), libswresample switches to
  `swr_alloc_set_opts2`, and the openh264 header path changes from `wels/` to `svc/`;
  removed the no-longer-used x264, webrtc\_aec3, rtpsdk, and pthread dependencies
* `utils::getPathFile` now uses `std::filesystem` to enumerate directories
* `operator<<(size_t)` of the log stream `SDebug` changed to `operator<<(unsigned int)`:
  on 64-bit, `size_t` is `unsigned __int64` and conflicts with the `uint64_t` overload.
  After the change, 32-bit unsigned values hit this overload and 64-bit values hit `uint64_t`, so both architectures behave the same
* Added the `RTC_ENABLE_RTP_STREAM` compile switch to custom/rtp stream code

**Fixed**

* Fixed 64-bit build errors: the virtual machine detection in `IsInsideVPC` / `IsInsideVMWare` relies on x86 inline assembly and
  x86-only `CONTEXT` fields (`Ebx` / `Eip`), and MSVC doesn't support inline assembly on x64—
  in 64-bit builds these two checks simply return false
* `FAACEncode`: closing the handle now uses `faac_encoder_close(&hEncoder)`; encoding now checks the returned status code
  and the number of bytes written, avoiding misjudgments caused by changed return value semantics in the new API

**API changes**

None. `srtc.h` / `srtc_def.h` are unchanged in this release and the vtable is unchanged, so you only need to replace the DLLs; no recompile is needed.

### \[7] - 2026.09.22

Version: 0.2.1-alpha.9

**Changed**

* Refactored webrtc2 reconnection and thread safety: `web_pccontrol2` protects `peer` / `_recv` / `_pc_stat` with locks,
  statistics and subscription state use value-copy snapshots, and state read and written across threads (`_use` / `_status`, etc.) uses atomic variables
* After reconnecting, publishing and subscriptions are restored from the subscription snapshot, and the send/receive packet statistics baseline is reset (so rates and packet loss rates don't come out negative after a reconnect)
* `SetRemoteDescription` failures are propagated upward, so a failed handshake is no longer marked as connected
* Unified the default encoding of source files (fixes C4819 warnings and potential garbled text from UTF-8 Chinese comments in a GBK environment)
* Cleaned up the stream initialization object (`SRTCChannel`)

**Fixed**

* Fixed CreateOffer / CreateAnswer waiting forever for the SDP callback, which caused the pubsub thread and
  `CloseRoom` to deadlock when leaving the channel: they now wait with a timeout

**API changes**

None. `srtc.h` / `srtc_def.h` are unchanged in this release and the vtable is unchanged, so you only need to replace the DLLs; no recompile is needed.

### \[6] - 2026.08.21

Version: 0.2.1-alpha.7

**Changed**

* After `join()` succeeds, `opt` is now parsed separately per media streaming type (`stream_vendor`):
  `link_id`, `session_key`, `upload.address`, and `upload.port` are required only when `stream_vendor` is `ook`,
  and `link_id` (the OOK room number) in the channel info is likewise required only under OOK
* Cleaned up the indentation of `UploadLog::appendLog2Map` (no functional change)

**Fixed**

* Fixed an issue where, with WebRTC media streaming (`stream_vendor` other than `ook`), the absence of `link_id` / `session_key` in `opt`
  caused an early return in `onJoinSucceed`, so no stream was set up after joining the channel

**API changes**

None. `srtc.h` / `srtc_def.h` are unchanged in this release and the vtable is unchanged, so you only need to replace the DLLs; no recompile is needed.

### \[5] - 2026.08.20

Version: 0.2.1-alpha.6

**Added**

* Added `IRTCChannel`: one object per channel, with all channel-level methods on it
* Added `IRTCChannelEvent`: channel-level callbacks are now a separate interface, registered per channel
* Added `IRTCEngine::createChannel(token, IRTCChannel**)`: creates only the channel object without joining the channel,
  so you can set up the configuration and event handler before `join()` (`onJoinChannel` fires before `join()` returns)
* Added `IRTCChannel::join()` / `getChannelId()` / `getSetting()` / `setEventHandler()`
* Added the `RTCEngineOptions` struct as the second parameter of `RTCEngine_Init`, for setting the log switch and log directory

**Changed**

* `IRTCEngine` keeps only engine-level capabilities: creating/leaving channels, device enumeration, network speed tests, log upload, and IM
* `IRTCEngineEvent` keeps only engine-level callbacks: device changes, network probing, and IM
* `IRTCSetting` is renamed `IRTCChannelSetting` and moved from the engine to the channel (one per channel;
  changes to `den_model` / `limit_speed` are applied to that channel immediately)
* Media streaming objects (`IStreamBase` / `StreamControl` / `WebrtcControl` / `WebrtcControl2`) changed from process-level singletons
  to one `new` instance per channel, released when the channel is destroyed; OOK's `any_meeting` is still a process-level singleton
* Local recording has only one recording object per channel (its key is the channelId); the `mid` parameter of `getLocalRecord` is kept only for interface compatibility
* Adjusted file encoding formats

**Fixed**

* Fixed OOK callbacks (rendering, room events, audio, custom frames) still pointing to a freed object after the media streaming object was released:
  `StreamControl` now unregisters these callbacks in its destructor
* Fixed the thread not being reclaimed while the object was destroyed when `OpenRoom` failed midway: the failure path now calls `CloseRoom()`
* Fixed the `StreamControl` thread flag not being initialized, which could make `CloseRoom()` wait forever if the stream info thread never started
* Fixed `web_meeting` and `web_meeting2` each registering AEC pre- and post-processing with the shared libwebrtc factory,
  overwriting each other and leaking: `web_base` now holds a single shared instance
* Fixed leaks where `web_meeting2` in `WebrtcControl2` and the device stream object in `WebrtcControl` were never released

**API changes (breaking; you must recompile)**

`IRTCEngine` / `IRTCEngineEvent` / `IRTCChannel` / `IRTCChannelEvent` / `IRTCChannelSetting`
are all pure virtual interfaces, and this release changes their vtables; `RTCEngine_Init` is an exported function and its signature has also changed.
**Replacing only the DLLs without recompiling crashes.**

* `RTCEngine_Init(IRTCEngine** rtc)` → `RTCEngine_Init(IRTCEngine** rtc, RTCEngineOptions* opt)`.
  Passing `nullptr` for `opt` means no SDK log is written
* `IRTCEngine::joinChannel(token, char**, int*)` is **removed**; use `createChannel(token, IRTCChannel**)` + `IRTCChannel::join()` instead
* `IRTCEngine::getSetting()` is **removed**; use `IRTCChannel::getSetting()` instead
* The 24 channel-level methods on `IRTCEngine` with a leading `channelId` parameter are **all removed** (no compatibility forwarding is kept)
  and moved to `IRTCChannel`, with the signature simply dropping `channelId`: `getMe`, `getChannel`, `getMembers`, `getMember`,
  `getMemberByLinkId`, `getLocalRecord`, `getCameraTrack`, `getScreenTrack`, `getAudioTrack`,
  `getRemoteAudioTrack`, `getRemoteVideoTrack`, `getMCUVideoTrack`, `getCustomVideoTrack`,
  `getCustomAudioTrack`, `setCustomRecvBack`, `subscribe`, `unsubscribe`, `publish`, `updatePublish`, `unpublish`
* `getMCUVideoTrack` now takes no parameters
* The 21 channel-level callbacks of `IRTCEngineEvent` are **all moved** to `IRTCChannelEvent`, and **their leading `channelId` parameter is removed**:
  `onJoinChannel`, `onChannelUpdate`, `onUserJoin`, `onUserUpdate`, `onUserLeave`, `onUserStreamAdd`,
  `onUserStreamUpdate`, `onUserStreamRemove`, `onCustomMessage2`, `onUpLevel`, `onDownLevel`, `onUpStat`,
  `onDownStat`, `onFrameTimeOut`, `onSpeakers`, `onDeviceStatusChange`, `onShareTargetNotFind`,
  `onDisconnected`, `onReconnected`, `onReconnecting`, `onRecordStatusChange`
* `IRTCSetting` → `IRTCChannelSetting`, and `sdk_log_path` / `enable_stream_log` are removed (moved to `RTCEngineOptions`)

**Known limitations**

* `stream_model` / `simple` / `mcu_track` / `enable_audio_record` in `IRTCChannelSetting`
  are read during `join()`; changing them after `join()` has no effect
* An `IRTCChannel*` becomes invalid after `leaveChannel` / `leaveAllChannel`, and you need to set it to null yourself;
  there's no reference counting, so don't hold it on another thread and use it across `leaveChannel`

### \[4] - 2026.08.19

Version: 0.2.1-alpha.5

**Added**

* Support for joining multiple channels at the same time; the engine manages multiple channels internally by channelId
* Added `leaveAllChannel()` to leave all joined channels at once
* Added `getChannelIds()` to get the list of currently joined channelIds
* Added a CMake build project and adjusted the directory structure

**Changed**

* Separated channel and client; each channel holds its own independent session / roomno / stream\_host / user\_id configuration
* Cleaned up audio sync for screen sharing; cleaned up screen recording; cleaned up synchronous pushing in updateSenderPublish
* Moved message callbacks to the recording task
* WebRTC local recording supports writing audio data to the file
* Removed garbled comments

**Fixed**

* Fixed the audio levels during recording
* Fixed a crash when ending SeaStart screen sharing with audio

**API changes (breaking; you must recompile)**

`IRTCEngine` / `IRTCEngineEvent` are pure virtual interfaces, and this release changes their vtables, so **replacing only the DLLs without recompiling crashes**.

* `joinChannel` changes to `joinChannel(const char* token, char** channelId, int* channelIdSize)`, returning the channelId through an output parameter
* `leaveChannel` changes to `leaveChannel(const char* channelId)`
* All channel-level methods add a leading `channelId` parameter: `getMe`, `getChannel`, `getMembers`, `getMember`, `getMemberByLinkId`, `getLocalRecord`, `getCameraTrack`, `getScreenTrack`, `getAudioTrack`, `getRemoteAudioTrack`, `getRemoteVideoTrack`, `getMCUVideoTrack`, `getCustomVideoTrack`, `getCustomAudioTrack`, `subscribe`, `unsubscribe`, `publish`, `updatePublish`, `unpublish`
* All channel-level callbacks add a leading `channelId` parameter: `onJoinChannel`, `onCustomMessage2`, `onUpLevel`, `onDownLevel`, `onUpStat`, `onDownStat`, `onFrameTimeOut`, `onSpeakers`, `onDeviceStatusChange`, `onShareTargetNotFind`, `onDisconnected`, `onReconnecting`, `onRecordStatusChange`
* The first parameter of `onJoinChannel` is now the plain channel ID; the original channel info JSON moves to the second parameter
* The first parameter of `onReconnected` is renamed from `channel` to `channelId`, with the same meaning
* The two `publish` overloads drop their default argument (the old `publish(tk)` was itself an ambiguous call and didn't compile), so you must pass `opt` explicitly
* Device change (`onDeviceChange`, `onDefDeviceChange`), network probing (`onProbeResult`), and the IM callbacks are engine-level, and their signatures are unchanged
* When `channelId` is `nullptr` or an empty string and only one channel has been joined, that channel is used automatically, to ease migrating old code

### \[3] - 2026.03.9

**Added**

* Added local recording
* Added cdn-webrtc stream receiving
* Added stream receiving over in-house WebRTC
* Added bandwidth limiting

**Changed**
Converted the documentation to Markdown

### \[2] - 2025.01.8

**Added**

* Added logging to the log system
* Set Opus as the default codec
* Added retrieval of the composite stream
* Changed random number generation

### \[1] - 2022.05.30

**Added**

* System playback and capture volume detection

**Changed**

* Merged the media stream success event and the join-channel event
* Changed some function names and return value contents
* Screen sharing supports sharing a window and enumerating window info

### \[0] - 2022.05.15

**Added**

* Initial release of rtc2.0