vision-agents-plugins-tencent 0.6.0__tar.gz

vision_agents_plugins_tencent-0.6.0/.gitignore

@@ -0,0 +1,102 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.cursor/*
+# Distribution / packaging
+.Python
+build/
+dist/
+downloads/
+develop-eggs/
+eggs/
+.eggs/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+pip-wheel-metadata/
+MANIFEST
+*.egg-info/
+*.egg
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+coverage.xml
+nosetests.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Type checker / lint caches
+.mypy_cache/
+.dmypy.json
+dmypy.json
+.pytype/
+.pyre/
+.ruff_cache/
+
+# Environments
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+.env
+.env.local
+.env.*.local
+.env.bak
+pyvenv.cfg
+.python-version
+
+# Editors / IDEs
+.vscode/
+.idea/
+
+# Jupyter Notebook
+.ipynb_checkpoints/
+
+# OS / Misc
+.DS_Store
+*.log
+
+# Tooling & repo-specific
+pyrightconfig.json
+shell.nix
+bin/*
+lib/*
+stream-py/
+
+# Example lock files (regenerated by uv sync)
+examples/*/uv.lock
+plugins/*/example/uv.lock
+
+# Artifacts / assets
+*.pt
+*.kef
+*.onnx
+profile.html
+
+/opencode.json
+.ralph-tui/
+.claude/*
+!.claude/skills/
+
+.uv-cache/
+
+# pytest json report
+.report.json

vision_agents_plugins_tencent-0.6.0/PKG-INFO

@@ -0,0 +1,96 @@
+Metadata-Version: 2.4
+Name: vision-agents-plugins-tencent
+Version: 0.6.0
+Summary: Tencent TRTC edge transport for Vision Agents
+Project-URL: Documentation, https://visionagents.ai/
+Project-URL: Website, https://visionagents.ai/
+Project-URL: Source, https://github.com/GetStream/Vision-Agents
+License-Expression: MIT
+Keywords: edge,rtc,tencent,trtc,vision agents,voice agents
+Classifier: Operating System :: POSIX :: Linux
+Requires-Python: >=3.10
+Requires-Dist: liteav>=13.2; sys_platform == 'linux'
+Requires-Dist: tls-sig-api-v2
+Requires-Dist: vision-agents
+Description-Content-Type: text/markdown
+
+# Vision Agents Plugin: Tencent TRTC Edge
+
+Tencent TRTC (Real-Time Communication) edge transport for Vision Agents. Lets an agent join a Tencent TRTC room and exchange audio/video with participants using the [Tencent LiteAV SDK](https://cloud.tencent.com/document/product/647) Python bindings.
+
+## Quickstart
+
+Talk to an agent in 5 minutes. You'll need Docker, an `.env` with `TENCENT_SDK_APP_ID`, `TENCENT_SDK_SECRET_KEY`, `OPENAI_API_KEY`, and `ELEVEN_API_KEY` at the repo root, and a working microphone in a Chromium-based browser. The example pairs Tencent TRTC with OpenAI (LLM) and ElevenLabs (STT + TTS) — chosen because Gemini and Cartesia aren't reachable from mainland China, which is the natural deployment target for this edge.
+
+1. **Open Tencent's hosted TRTC Web SDK quick demo:**
+   <https://web.sdk.qcloud.com/trtc/webrtc/v5/demo/quick-demo-js/index.html>
+
+   - Paste your `TENCENT_SDK_APP_ID` into **SDKAppID** and `TENCENT_SDK_SECRET_KEY` into **SDKSecretKey** — the page generates `UserSig` client-side.
+   - Leave the auto-generated **UserID** and **RoomID(String)** as is.
+   - Click **Enter Room** → demo log should print `🟩 [user_***] enterRoom.`
+   - Click **Start Local Video** — this also publishes the mic.
+
+2. **Copy the `RoomID(String)` from the demo form** and launch the agent with it:
+
+   ```bash
+   cd plugins/tencent
+   TENCENT_TEST_ROOM_ID=<paste-room-id-here> docker compose run --rm tencent-agent
+   ```
+
+   On first run this builds the image and resolves the workspace; subsequent runs are fast. When the agent joins you'll see `Tencent TRTC OnRemoteUserEnterRoom: <userId>` matching the **UserID** shown in the demo.
+
+3. **Talk.** The flow is `browser → Tencent → STT (ElevenLabs) → LLM (OpenAI) → TTS (ElevenLabs) → Tencent → browser`. Confirm each leg in the agent log:
+   - `🎤 [Transcript Complete]: …` — STT got your speech.
+   - `🤖 [LLM response final]: …` — LLM produced a reply.
+   - Reply plays back in the browser.
+
+## Usage in your own code
+
+```python
+from vision_agents.core import Agent, User
+from vision_agents.plugins import tencent
+
+agent_user = User(id="agent-1", name="Agent")
+edge = tencent.Edge(
+    sdk_app_id=YOUR_SDK_APP_ID,
+    user_sig=USER_SIG,  # or key=KEY to generate per join
+)
+
+agent = Agent(
+    edge=edge,
+    agent_user=agent_user,
+    llm=...,
+    # stt, tts, etc.
+)
+
+call = await edge.create_call(call_id=str(ROOM_ID), room_id=ROOM_ID)
+await agent.join(call)
+await agent.finish()
+```
+
+## Install
+
+```bash
+uv add vision-agents-plugins-tencent
+```
+
+On Linux this pulls `liteav` from PyPI. On macOS the `liteav` dependency is skipped via a platform marker, so the package installs but `tencent.Edge()` raises at runtime — use the Docker setup from the Quickstart.
+
+## Configuration
+
+`tencent.Edge(...)` parameters:
+
+- **sdk_app_id** (int): Tencent TRTC SDK App ID. Falls back to `TENCENT_SDK_APP_ID` env var.
+- **user_sig** (str): User signature for the agent user.
+- **key** (str): Optional. If set and `user_sig` is not, the plugin generates `user_sig` via TLSSigAPIv2. Falls back to `TENCENT_SDK_SECRET_KEY` env var.
+- **video_fps** (int): Outgoing video frame rate.
+
+Environment variables:
+
+- `TENCENT_SDK_APP_ID`, `TENCENT_SDK_SECRET_KEY` — credentials.
+- `TENCENT_TRTC_SCENE` — one of `auto` (default), `videocall`, `call`, `record`.
+- `TENCENT_TEST_ROOM_ID` — interpolated by `docker-compose.yml` into the example runner's `--call-id` flag.
+
+## Platform support
+
+The underlying [`liteav`](https://pypi.org/project/liteav/) package ships only manylinux wheels — **Linux x86_64 / aarch64 only**. macOS and Windows must run the agent inside a Linux container (see Quickstart). User sigs are needed for room entry; either generate them with [TLSSigAPIv2](https://cloud.tencent.com/document/product/647/34399) or pass `key=` and let the plugin sign per join.

vision_agents_plugins_tencent-0.6.0/README.md

@@ -0,0 +1,80 @@
+# Vision Agents Plugin: Tencent TRTC Edge
+
+Tencent TRTC (Real-Time Communication) edge transport for Vision Agents. Lets an agent join a Tencent TRTC room and exchange audio/video with participants using the [Tencent LiteAV SDK](https://cloud.tencent.com/document/product/647) Python bindings.
+
+## Quickstart
+
+Talk to an agent in 5 minutes. You'll need Docker, an `.env` with `TENCENT_SDK_APP_ID`, `TENCENT_SDK_SECRET_KEY`, `OPENAI_API_KEY`, and `ELEVEN_API_KEY` at the repo root, and a working microphone in a Chromium-based browser. The example pairs Tencent TRTC with OpenAI (LLM) and ElevenLabs (STT + TTS) — chosen because Gemini and Cartesia aren't reachable from mainland China, which is the natural deployment target for this edge.
+
+1. **Open Tencent's hosted TRTC Web SDK quick demo:**
+   <https://web.sdk.qcloud.com/trtc/webrtc/v5/demo/quick-demo-js/index.html>
+
+   - Paste your `TENCENT_SDK_APP_ID` into **SDKAppID** and `TENCENT_SDK_SECRET_KEY` into **SDKSecretKey** — the page generates `UserSig` client-side.
+   - Leave the auto-generated **UserID** and **RoomID(String)** as is.
+   - Click **Enter Room** → demo log should print `🟩 [user_***] enterRoom.`
+   - Click **Start Local Video** — this also publishes the mic.
+
+2. **Copy the `RoomID(String)` from the demo form** and launch the agent with it:
+
+   ```bash
+   cd plugins/tencent
+   TENCENT_TEST_ROOM_ID=<paste-room-id-here> docker compose run --rm tencent-agent
+   ```
+
+   On first run this builds the image and resolves the workspace; subsequent runs are fast. When the agent joins you'll see `Tencent TRTC OnRemoteUserEnterRoom: <userId>` matching the **UserID** shown in the demo.
+
+3. **Talk.** The flow is `browser → Tencent → STT (ElevenLabs) → LLM (OpenAI) → TTS (ElevenLabs) → Tencent → browser`. Confirm each leg in the agent log:
+   - `🎤 [Transcript Complete]: …` — STT got your speech.
+   - `🤖 [LLM response final]: …` — LLM produced a reply.
+   - Reply plays back in the browser.
+
+## Usage in your own code
+
+```python
+from vision_agents.core import Agent, User
+from vision_agents.plugins import tencent
+
+agent_user = User(id="agent-1", name="Agent")
+edge = tencent.Edge(
+    sdk_app_id=YOUR_SDK_APP_ID,
+    user_sig=USER_SIG,  # or key=KEY to generate per join
+)
+
+agent = Agent(
+    edge=edge,
+    agent_user=agent_user,
+    llm=...,
+    # stt, tts, etc.
+)
+
+call = await edge.create_call(call_id=str(ROOM_ID), room_id=ROOM_ID)
+await agent.join(call)
+await agent.finish()
+```
+
+## Install
+
+```bash
+uv add vision-agents-plugins-tencent
+```
+
+On Linux this pulls `liteav` from PyPI. On macOS the `liteav` dependency is skipped via a platform marker, so the package installs but `tencent.Edge()` raises at runtime — use the Docker setup from the Quickstart.
+
+## Configuration
+
+`tencent.Edge(...)` parameters:
+
+- **sdk_app_id** (int): Tencent TRTC SDK App ID. Falls back to `TENCENT_SDK_APP_ID` env var.
+- **user_sig** (str): User signature for the agent user.
+- **key** (str): Optional. If set and `user_sig` is not, the plugin generates `user_sig` via TLSSigAPIv2. Falls back to `TENCENT_SDK_SECRET_KEY` env var.
+- **video_fps** (int): Outgoing video frame rate.
+
+Environment variables:
+
+- `TENCENT_SDK_APP_ID`, `TENCENT_SDK_SECRET_KEY` — credentials.
+- `TENCENT_TRTC_SCENE` — one of `auto` (default), `videocall`, `call`, `record`.
+- `TENCENT_TEST_ROOM_ID` — interpolated by `docker-compose.yml` into the example runner's `--call-id` flag.
+
+## Platform support
+
+The underlying [`liteav`](https://pypi.org/project/liteav/) package ships only manylinux wheels — **Linux x86_64 / aarch64 only**. macOS and Windows must run the agent inside a Linux container (see Quickstart). User sigs are needed for room entry; either generate them with [TLSSigAPIv2](https://cloud.tencent.com/document/product/647/34399) or pass `key=` and let the plugin sign per join.

vision_agents_plugins_tencent-0.6.0/pyproject.toml

@@ -0,0 +1,42 @@
+[build-system]
+requires = ["hatchling", "hatch-vcs"]
+build-backend = "hatchling.build"
+
+[project]
+name = "vision-agents-plugins-tencent"
+dynamic = ["version"]
+description = "Tencent TRTC edge transport for Vision Agents"
+readme = "README.md"
+keywords = ["tencent", "trtc", "rtc", "edge", "vision agents", "voice agents"]
+requires-python = ">=3.10"
+license = "MIT"
+classifiers = ["Operating System :: POSIX :: Linux"]
+dependencies = [
+    "vision-agents",
+    "liteav>=13.2; sys_platform == 'linux'",
+    "tls-sig-api-v2",
+]
+
+[project.urls]
+Documentation = "https://visionagents.ai/"
+Website = "https://visionagents.ai/"
+Source = "https://github.com/GetStream/Vision-Agents"
+
+[tool.hatch.version]
+source = "vcs"
+raw-options = { root = "..", search_parent_directories = true, fallback_version = "0.0.0" }
+
+[tool.hatch.build.targets.wheel]
+packages = [".", "vision_agents"]
+
+[tool.hatch.build.targets.sdist]
+include = ["/vision_agents"]
+
+[tool.uv.sources]
+vision-agents = { workspace = true }
+
+[dependency-groups]
+dev = [
+    "pytest>=8.4.1",
+    "pytest-asyncio>=1.0.0",
+]

vision_agents_plugins_tencent-0.6.0/vision_agents/plugins/tencent/__init__.py

@@ -0,0 +1,5 @@
+"""Tencent TRTC edge transport for Vision Agents."""
+
+from .tencent_edge import TencentEdge as Edge
+
+__all__ = ["Edge"]

vision_agents_plugins_tencent-0.6.0/vision_agents/plugins/tencent/bindings.py

@@ -0,0 +1,125 @@
+"""Conditional `liteav` imports for the Tencent plugin.
+
+`liteav` ships only manylinux wheels on PyPI, so on macOS/Windows the
+import will fail. This module centralises the conditional import noise:
+all other modules in the plugin re-import the names they need from
+here. ``require_liteav()`` raises a friendly RuntimeError at call time
+if liteav couldn't be loaded.
+"""
+
+import logging
+import os
+from typing import Optional
+
+logger = logging.getLogger(__name__)
+
+__all__ = [
+    "AUDIO_CODEC_TYPE_PCM",
+    "AUDIO_OBTAIN_METHOD_CALLBACK",
+    "AudioEncodeParams",
+    "AudioFrame",
+    "CreateTRTCCloud",
+    "DestroyTRTCCloud",
+    "EnterRoomParams",
+    "LITEAV_IMPORT_ERROR",
+    "PixelFrame",
+    "STREAM_TYPE_VIDEO_HIGH",
+    "TRTCCloudDelegate",
+    "TRTC_ROLE_ANCHOR",
+    "TRTC_SCENE_CALL",
+    "TRTC_SCENE_RECORD",
+    "TRTC_SCENE_VIDEOCALL",
+    "TrtcString",
+    "VIDEO_PIXEL_FORMAT_YUV420p",
+    "VIDEO_ROTATION_0",
+    "VideoEncodeParams",
+    "require_liteav",
+]
+
+LITEAV_IMPORT_ERROR: Optional[ImportError] = None
+
+try:
+    # liteav's typed stubs don't expose these top-level module constants
+    # and convenience constructors even though they exist at runtime, so
+    # we silence the attr-defined / call-arg complaints in this block.
+    from liteav import (  # type: ignore[attr-defined]
+        AUDIO_CODEC_TYPE_PCM,
+        AUDIO_OBTAIN_METHOD_CALLBACK,
+        STREAM_TYPE_VIDEO_HIGH,
+        VIDEO_PIXEL_FORMAT_YUV420p,
+        VIDEO_ROTATION_0,
+        AudioEncodeParams,
+        AudioFrame,
+        CreateTRTCCloud,
+        DestroyTRTCCloud,
+        EnterRoomParams,
+        PixelFrame,
+        TrtcString,
+        TRTCCloudDelegate,
+        TRTC_ROLE_ANCHOR,
+        TRTC_SCENE_RECORD,
+        VideoEncodeParams,
+    )
+except ImportError as e:
+    LITEAV_IMPORT_ERROR = e
+    # All names need module-level fallbacks so downstream `from bindings
+    # import ...` succeeds on non-linux. TencentEdge.__init__ calls
+    # require_liteav() which raises before any of these get touched.
+    AUDIO_CODEC_TYPE_PCM = None  # type: ignore[assignment]
+    AUDIO_OBTAIN_METHOD_CALLBACK = None  # type: ignore[assignment]
+    STREAM_TYPE_VIDEO_HIGH = None  # type: ignore[assignment]
+    VIDEO_PIXEL_FORMAT_YUV420p = None  # type: ignore[assignment]
+    VIDEO_ROTATION_0 = None  # type: ignore[assignment]
+    AudioEncodeParams = None  # type: ignore[misc, assignment]
+    AudioFrame = None  # type: ignore[misc, assignment]
+    CreateTRTCCloud = None
+    DestroyTRTCCloud = None
+    EnterRoomParams = None  # type: ignore[misc, assignment]
+    PixelFrame = None  # type: ignore[misc, assignment]
+    TrtcString = None  # type: ignore[misc, assignment]
+    TRTCCloudDelegate = None  # type: ignore[misc, assignment]
+    TRTC_ROLE_ANCHOR = None  # type: ignore[assignment]
+    TRTC_SCENE_RECORD = None  # type: ignore[assignment]
+    VideoEncodeParams = None  # type: ignore[misc, assignment]
+
+try:
+    from liteav import TRTC_SCENE_CALL, TRTC_SCENE_VIDEOCALL  # type: ignore[attr-defined]
+except ImportError:
+    TRTC_SCENE_VIDEOCALL = None  # type: ignore[assignment]
+    TRTC_SCENE_CALL = None  # type: ignore[assignment]
+
+
+# liteav's C++ side writes verbose `[I][...]` / `[E][...]` lines straight
+# to stdout. Calling DisableConsoleLog + SetLogLevel(kNone) silences the
+# periodic noise (thread_watchdog, audio_track_health_monitor, etc.) but
+# the SDK still writes a chunk of connection-setup logs from
+# CreateTRTCCloud → EnterRoom that we can't suppress through its public
+# API. Set TENCENT_LITEAV_DEBUG=1 to bring those periodic logs back for
+# diagnostics. Note on the level enum (non-standard): 0=kAll, 1=kInfo,
+# 2=kWarning, 3=kError, 4=kFatal, 5=kNone (>=6 is UN_DEF, no-op).
+_TRUTHY = {"1", "true", "yes", "on"}
+_LITEAV_LOG_LEVEL_NONE = 5
+
+if (
+    LITEAV_IMPORT_ERROR is None
+    and os.environ.get("TENCENT_LITEAV_DEBUG", "").strip().lower() not in _TRUTHY
+):
+    try:
+        import liteav as _liteav
+
+        _liteav.DisableConsoleLog()  # type: ignore[attr-defined]
+        _liteav.SetLogLevel(_LITEAV_LOG_LEVEL_NONE)  # type: ignore[attr-defined]
+    except AttributeError as e:
+        # Older liteav builds may not expose these helpers.
+        logger.debug("Could not silence liteav console log: %s", e)
+
+
+def require_liteav() -> None:
+    """Raise a friendly error if liteav couldn't be imported on this platform."""
+    if LITEAV_IMPORT_ERROR is not None:
+        raise RuntimeError(
+            "Tencent TRTC edge requires the `liteav` package, which ships only "
+            "manylinux wheels and cannot be imported on this platform. Install "
+            "on Linux (x86_64 or aarch64) with `pip install liteav` or run the "
+            "agent inside a Linux container (see plugins/tencent/README.md)."
+        ) from LITEAV_IMPORT_ERROR

vision_agents_plugins_tencent-0.6.0/vision_agents/plugins/tencent/delegate.py

@@ -0,0 +1,242 @@
+"""TRTCCloudDelegate subclass that forwards SDK callbacks to TencentEdge.
+
+The class is defined only when liteav was importable; on non-linux
+platforms ``TencentDelegate`` resolves to ``None`` and TencentEdge
+fails earlier via ``require_liteav()``.
+
+All callbacks are wrapped in try/except because any unhandled Python
+exception in a SWIG director callback triggers
+Swig::DirectorMethodException on the C++ side, which calls
+std::terminate and kills the process.
+"""
+
+import asyncio
+import logging
+from typing import TYPE_CHECKING, Any
+
+from vision_agents.plugins.tencent.bindings import (
+    STREAM_TYPE_VIDEO_HIGH,
+    AudioEncodeParams,
+    TRTCCloudDelegate,
+    VideoEncodeParams,
+)
+from vision_agents.plugins.tencent.tracks import CHANNELS, SAMPLE_RATE
+
+if TYPE_CHECKING:
+    from vision_agents.plugins.tencent.tencent_edge import (
+        TencentConnection,
+        TencentEdge,
+    )
+
+logger = logging.getLogger(__name__)
+
+
+def _extract_pcm(frame: Any) -> bytes:
+    if frame.size <= 0:
+        return b""
+    return frame.data
+
+
+TencentDelegate: Any = None
+if TRTCCloudDelegate is not None:
+
+    class _TencentDelegateCls(TRTCCloudDelegate):
+        """Forwards TRTC SDK callbacks to TencentEdge and TencentConnection."""
+
+        def __init__(
+            self,
+            edge: "TencentEdge",
+            connection: "TencentConnection",
+            loop: asyncio.AbstractEventLoop,
+        ):
+            # Intentionally not super().__init__(): TRTCCloudDelegate is a SWIG
+            # director — its __init__ binds the C++ side via `self.this` and
+            # registers Python callback dispatch. Tencent's bundled
+            # python_x86_64 sample uses exactly this explicit form, and that
+            # is the only invocation contract liteav documents. We follow it
+            # verbatim rather than relying on cooperative MRO that the C++
+            # binding never promised to honour.
+            TRTCCloudDelegate.__init__(self)
+            self._edge = edge
+            self._connection = connection
+            self._agent_user_id = edge._agent_user_id or ""
+
+        def OnError(self, error: int) -> None:  # type: ignore[override]
+            try:
+                logger.error("Tencent TRTC OnError: %s", error)
+                self._edge._emit_call_ended()
+            except BaseException:
+                logger.exception("OnError callback failed")
+
+        def OnEnterRoom(self) -> None:
+            try:
+                logger.info("Tencent TRTC OnEnterRoom")
+                cloud = self._edge._cloud
+                if cloud is None:
+                    return
+                param = AudioEncodeParams()
+                param.sample_rate = SAMPLE_RATE
+                param.channels = CHANNELS
+                param.bitrate_bps = 54000
+                cloud.CreateLocalAudioChannel(param)
+
+                if self._edge._outgoing_video_track is not None:
+                    cloud.CreateLocalVideoChannel(STREAM_TYPE_VIDEO_HIGH)
+            except BaseException:
+                logger.exception("OnEnterRoom callback failed")
+
+        def OnExitRoom(self) -> None:
+            try:
+                logger.info("Tencent TRTC OnExitRoom")
+                self._edge._emit_call_ended()
+            except BaseException:
+                logger.exception("OnExitRoom callback failed")
+
+        def OnLocalAudioChannelCreated(self) -> None:
+            try:
+                if self._edge._audio_track and self._edge._cloud:
+                    self._edge._audio_track.set_cloud(self._edge._cloud)
+            except BaseException:
+                logger.exception("OnLocalAudioChannelCreated callback failed")
+
+        def OnConnectionStateChanged(self, old_state: int, new_state: int) -> None:  # type: ignore[override]
+            logger.debug(
+                "Tencent TRTC connection state: %s -> %s", old_state, new_state
+            )
+
+        def OnLocalAudioChannelDestroyed(self) -> None:
+            pass
+
+        def OnLocalVideoChannelCreated(self, stream_type: int) -> None:  # type: ignore[override]
+            try:
+                logger.info(
+                    "Tencent TRTC OnLocalVideoChannelCreated: stream_type=%s",
+                    stream_type,
+                )
+                edge = self._edge
+                if edge._outgoing_video_track and edge._cloud and edge._loop:
+                    vp = VideoEncodeParams()
+                    vp.frame_rate = edge._video_fps
+                    vp.bitrate_bps = 1_000_000
+                    vp.gop_in_seconds = 3
+                    edge._cloud.SetVideoEncodeParam(STREAM_TYPE_VIDEO_HIGH, vp)
+                    edge._outgoing_video_track.set_cloud(edge._cloud, edge._loop)
+            except BaseException:
+                logger.exception("OnLocalVideoChannelCreated callback failed")
+
+        def OnLocalVideoChannelDestroyed(self, stream_type: int) -> None:  # type: ignore[override]
+            pass
+
+        def OnRequestChangeVideoEncodeBitrate(  # type: ignore[override]
+            self, stream_type: int, bitrate_bps: int
+        ) -> None:
+            pass
+
+        def OnRequestKeyFrame(self, stream_type: int) -> None:  # type: ignore[override]
+            logger.debug("Tencent TRTC OnRequestKeyFrame: stream_type=%s", stream_type)
+
+        def OnRemoteAudioAvailable(self, user_id: str, available: bool) -> None:
+            try:
+                if user_id and user_id != self._agent_user_id:
+                    if available:
+                        self._edge._emit_track_added(user_id)
+                    else:
+                        self._edge._emit_track_removed(user_id)
+            except BaseException:
+                logger.exception("OnRemoteAudioAvailable callback failed")
+
+        def OnRemoteVideoAvailable(  # type: ignore[override]
+            self, user_id: str, available: bool, stream_type: int
+        ) -> None:
+            try:
+                if not user_id or user_id == self._agent_user_id:
+                    return
+                if available:
+                    logger.info(
+                        "Tencent TRTC video available from %s (stream_type=%s)",
+                        user_id,
+                        stream_type,
+                    )
+                    self._edge._emit_video_track_added(user_id)
+                else:
+                    logger.info("Tencent TRTC video unavailable from %s", user_id)
+                    self._edge._emit_video_track_removed(user_id)
+            except BaseException:
+                logger.exception("OnRemoteVideoAvailable callback failed")
+
+        def OnRemoteVideoFrameReceived(  # type: ignore[override]
+            self, user_id: str, stream_type: int, frame: Any
+        ) -> None:
+            pass
+
+        def OnRemotePixelFrameReceived(  # type: ignore[override]
+            self, user_id: str, stream_type: int, frame: Any
+        ) -> None:
+            try:
+                if not user_id or user_id == self._agent_user_id:
+                    return
+                width = frame.width
+                height = frame.height
+                if width <= 0 or height <= 0 or frame.size <= 0:
+                    return
+                self._edge._push_video_frame(
+                    user_id, frame.data, width, height, frame.pts
+                )
+            except BaseException:
+                logger.exception("OnRemotePixelFrameReceived callback failed")
+
+        def OnSeiMessageReceived(  # type: ignore[override]
+            self, user_id: str, stream_type: int, message_type: int, message: Any
+        ) -> None:
+            pass
+
+        def OnReceiveCustomCmdMsg(
+            self, user_id: str, cmd_id: int, seq: int, message: Any
+        ) -> None:
+            pass
+
+        def OnMissCustomCmdMsg(
+            self, user_id: str, cmd_id: int, error_code: int, missed: int
+        ) -> None:
+            pass
+
+        def OnNetworkQuality(self, local_quality: Any, remote_qualities: Any) -> None:
+            pass
+
+        def OnRemoteUserEnterRoom(self, info: Any) -> None:
+            try:
+                user_id = info.user_id.GetValue() if info and info.user_id else ""
+                if user_id and user_id != self._agent_user_id:
+                    self._connection._on_remote_entered()
+                logger.info("Tencent TRTC OnRemoteUserEnterRoom: %s", user_id)
+            except BaseException:
+                logger.exception("OnRemoteUserEnterRoom callback failed")
+
+        def OnRemoteUserExitRoom(self, info: Any, reason: int) -> None:
+            try:
+                user_id = info.user_id.GetValue() if info and info.user_id else ""
+                if user_id and user_id != self._agent_user_id:
+                    self._connection._on_remote_left()
+                logger.info(
+                    "Tencent TRTC OnRemoteUserExitRoom: %s reason=%s", user_id, reason
+                )
+            except BaseException:
+                logger.exception("OnRemoteUserExitRoom callback failed")
+
+        def OnRemoteAudioReceived(self, user_id: str, frame: Any) -> None:
+            try:
+                if not user_id or user_id == self._agent_user_id:
+                    return
+                data = _extract_pcm(frame)
+                if data:
+                    self._edge._emit_audio_received(user_id, data)
+            except BaseException:
+                logger.exception("OnRemoteAudioReceived failed")
+
+        def OnRemoteMixedAudioReceived(self, frame: Any) -> None:
+            # Skip mixed audio — per-user callbacks already deliver individual
+            # streams and processing both doubles GIL / buffer pressure with
+            # no benefit for single-speaker scenarios.
+            pass
+
+    TencentDelegate = _TencentDelegateCls