vision-agents-plugins-kokoro 0.0.17__tar.gz

vision_agents_plugins_kokoro-0.0.17/.gitignore

@@ -0,0 +1,32 @@
+*/__pycache__
+*/chat/__pycache__
+*/video/__pycache__
+*/chat/sync/__pycache__
+*/chat/async_/__pycache__
+*/sync/__pycache__
+*/async_/__pycache__
+*/video/sync/__pycache__
+*/model/__pycache__/
+*/cli/__pycache__
+*/cli/__pycache__
+.env
+.venv
+.vscode/settings.json
+*.pyc
+dist/*
+dist/*
+*.log
+.python-version
+pyvenv.cfg
+.idea*
+bin/*
+lib/*
+shell.nix
+pyrightconfig.json
+.DS_Store
+
+*.egg-info/
+*.egg
+*.pt
+*.kef
+.env.bak

vision_agents_plugins_kokoro-0.0.17/PKG-INFO

@@ -0,0 +1,54 @@
+Metadata-Version: 2.4
+Name: vision-agents-plugins-kokoro
+Version: 0.0.17
+Summary: Kokoro TTS integration 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: AI,TTS,agents,kokoro,text-to-speech,voice agents
+Requires-Python: >=3.10
+Requires-Dist: kokoro>=0.9.4
+Requires-Dist: misaki[en]>=0.9.4
+Requires-Dist: numpy<2.3,>=2.2.6
+Requires-Dist: vision-agents
+Description-Content-Type: text/markdown
+
+# GetStream Kokoro Plugin
+
+This package integrates the open-weight [Kokoro-82M TTS model](https://github.com/hexgrad/kokoro) with the GetStream audio/video SDK.
+
+It provides a drop-in `KokoroTTS` class that implements the common `getstream_common.tts.TTS` interface, allowing you to stream PCM audio generated by Kokoro directly into a WebRTC `AudioStreamTrack`.
+
+```py
+from getstream.plugins.kokoro import KokoroTTS
+from getstream.video.rtc.audio_track import AudioStreamTrack
+
+track = AudioStreamTrack(framerate=24_000)
+
+tts = KokoroTTS(lang_code="a", voice="af_heart")
+tts.set_output_track(track)
+
+await tts.send("Hello from Kokoro!")
+```
+
+## Installation
+
+```bash
+pip install getstream-plugins-kokoro
+```
+
+This will pull in the required `kokoro`, `numpy` and `getstream[webrtc"]` dependencies.  You also need `espeak-ng` **at runtime** for pronunciation fallback.  On macOS you can install it with Homebrew:
+
+```bash
+brew install espeak-ng
+```
+
+## Configuration options
+
+| Parameter | Default | Description |
+|-----------|---------|-------------|
+| `lang_code` | `"a"` | Language group passed to `KPipeline` (`"a"` = American English, etc.) |
+| `voice` | `"af_heart"` | Kokoro voice preset.  See the [model card](https://huggingface.co/NeuML/kokoro-int8-onnx#speaker-reference) for available options. |
+| `speed` | `1.0` | Playback speed multiplier. |
+| `sample_rate` | `24000` | Output sample-rate (fixed by Kokoro).  **The attached `AudioStreamTrack` must use the same value.** |

vision_agents_plugins_kokoro-0.0.17/README.md

@@ -0,0 +1,38 @@
+# GetStream Kokoro Plugin
+
+This package integrates the open-weight [Kokoro-82M TTS model](https://github.com/hexgrad/kokoro) with the GetStream audio/video SDK.
+
+It provides a drop-in `KokoroTTS` class that implements the common `getstream_common.tts.TTS` interface, allowing you to stream PCM audio generated by Kokoro directly into a WebRTC `AudioStreamTrack`.
+
+```py
+from getstream.plugins.kokoro import KokoroTTS
+from getstream.video.rtc.audio_track import AudioStreamTrack
+
+track = AudioStreamTrack(framerate=24_000)
+
+tts = KokoroTTS(lang_code="a", voice="af_heart")
+tts.set_output_track(track)
+
+await tts.send("Hello from Kokoro!")
+```
+
+## Installation
+
+```bash
+pip install getstream-plugins-kokoro
+```
+
+This will pull in the required `kokoro`, `numpy` and `getstream[webrtc"]` dependencies.  You also need `espeak-ng` **at runtime** for pronunciation fallback.  On macOS you can install it with Homebrew:
+
+```bash
+brew install espeak-ng
+```
+
+## Configuration options
+
+| Parameter | Default | Description |
+|-----------|---------|-------------|
+| `lang_code` | `"a"` | Language group passed to `KPipeline` (`"a"` = American English, etc.) |
+| `voice` | `"af_heart"` | Kokoro voice preset.  See the [model card](https://huggingface.co/NeuML/kokoro-int8-onnx#speaker-reference) for available options. |
+| `speed` | `1.0` | Playback speed multiplier. |
+| `sample_rate` | `24000` | Output sample-rate (fixed by Kokoro).  **The attached `AudioStreamTrack` must use the same value.** |

vision_agents_plugins_kokoro-0.0.17/pyproject.toml

@@ -0,0 +1,42 @@
+[build-system]
+requires = ["hatchling", "hatch-vcs"]
+build-backend = "hatchling.build"
+
+[project]
+name = "vision-agents-plugins-kokoro"
+dynamic = ["version"]
+description = "Kokoro TTS integration for Vision Agents"
+readme = "README.md"
+keywords = ["kokoro", "TTS", "text-to-speech", "AI", "voice agents", "agents"]
+requires-python = ">=3.10"
+license = "MIT"
+dependencies = [
+    "vision-agents",
+    "kokoro>=0.9.4",
+    "misaki[en]>=0.9.4",
+    "numpy>=2.2.6,<2.3",
+]
+
+[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 = ["."]
+
+[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_kokoro-0.0.17/vision_agents/plugins/kokoro/__init__.py

@@ -0,0 +1,6 @@
+from .tts import TTS
+
+# Re-export under the new namespace for convenience
+__path__ = __import__("pkgutil").extend_path(__path__, __name__)
+
+__all__ = ["TTS"]

vision_agents_plugins_kokoro-0.0.17/vision_agents/plugins/kokoro/tts.py

@@ -0,0 +1,92 @@
+from __future__ import annotations
+
+import asyncio
+import logging
+
+import numpy as np
+from typing import AsyncIterator, List, Optional
+
+from vision_agents.core import tts
+from getstream.video.rtc.audio_track import AudioStreamTrack
+
+try:
+    from kokoro import KPipeline  # type: ignore
+except ModuleNotFoundError:  # pragma: no cover – mocked during CI
+    KPipeline = None  # type: ignore  # noqa: N816
+
+
+class TTS(tts.TTS):
+    """Text-to-Speech plugin backed by the Kokoro-82M model."""
+
+    def __init__(
+        self,
+        lang_code: str = "a",  # American English
+        voice: str = "af_heart",
+        speed: float = 1.0,
+        sample_rate: int = 24_000,
+        device: Optional[str] = None,
+        client: Optional[KPipeline] = None,
+    ) -> None:
+        super().__init__()
+
+        if KPipeline is None:
+            raise ImportError(
+                "The 'kokoro' package is not installed. ``pip install kokoro`` first."
+            )
+
+        self._pipeline = (
+            KPipeline(lang_code=lang_code)
+            if device is None
+            else KPipeline(lang_code=lang_code, device=device)
+        )
+        self.voice = voice
+        self.speed = speed
+        self.sample_rate = sample_rate
+        self.client = client if client is not None else self._pipeline
+
+    def get_required_framerate(self) -> int:
+        """Get the required framerate for Kokoro TTS."""
+        return self.sample_rate
+
+    def get_required_stereo(self) -> bool:
+        """Get whether Kokoro TTS requires stereo audio."""
+        return False  # Kokoro returns mono audio
+
+    def set_output_track(self, track: AudioStreamTrack) -> None:  # noqa: D401
+        if track.framerate != self.sample_rate:
+            raise TypeError(
+                f"Invalid framerate {track.framerate}, Kokoro requires {self.sample_rate} Hz"
+            )
+        super().set_output_track(track)
+
+    async def stream_audio(self, text: str, *_, **__) -> AsyncIterator[bytes]:  # noqa: D401
+        loop = asyncio.get_event_loop()
+        chunks: List[bytes] = await loop.run_in_executor(
+            None, lambda: list(self._generate_chunks(text))
+        )
+
+        async def _aiter():
+            for chunk in chunks:
+                yield chunk
+
+        return _aiter()
+
+    async def stop_audio(self) -> None:
+        """
+        Clears the queue and stops playing audio.
+
+        """
+        try:
+            await self.track.flush()
+            return
+        except Exception as e:
+            logging.error(f"Error flushing audio track: {e}")
+
+    def _generate_chunks(self, text: str):
+        for _gs, _ps, audio in self._pipeline(
+            text, voice=self.voice, speed=self.speed, split_pattern=r"\n+"
+        ):
+            if not isinstance(audio, np.ndarray):
+                audio = np.asarray(audio)
+            pcm16 = (np.clip(audio, -1.0, 1.0) * 32767.0).astype("<i2")
+            yield pcm16.tobytes()