vision-agents-plugins-wizper 0.0.17__py3-none-any.whl

.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

PKG-INFO

@@ -0,0 +1,15 @@
+Metadata-Version: 2.4
+Name: vision-agents-plugins-wizper
+Version: 0.0.17
+Summary: Wizper plugin 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,agents,fal,turn detection,voice agents
+Requires-Python: >=3.10
+Requires-Dist: fal-client>=0.7.0
+Requires-Dist: vision-agents
+Description-Content-Type: text/markdown
+
+Wizper for vision agents

README.md

@@ -0,0 +1 @@
+Wizper for vision agents

pyproject.toml

@@ -0,0 +1,41 @@
+[build-system]
+requires = ["hatchling", "hatch-vcs"]
+build-backend = "hatchling.build"
+
+[project]
+name = "vision-agents-plugins-wizper"
+dynamic = ["version"]
+description = "Wizper plugin for Vision Agents"
+readme = "README.md"
+keywords = ["fal", "turn detection", "AI", "voice agents", "agents"]
+requires-python = ">=3.10"
+license = "MIT"
+dependencies = [
+    "vision-agents",
+    "fal-client>=0.7.0",
+]
+
+[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",
+    "numpy>=2.2.6,<2.3",
+]

vision_agents/plugins/wizper/__init__.py

@@ -0,0 +1,5 @@
+from .stt import STT
+
+# Re-export under the new namespace for convenience
+
+__all__ = ["STT"]

vision_agents/plugins/wizper/stt.py

@@ -0,0 +1,180 @@
+"""
+Fal Wizper STT Plugin for Stream
+
+Provides real-time audio transcription and translation using fal-ai/wizper (Whisper v3).
+This plugin integrates with Stream's audio processing pipeline to provide high-quality
+speech-to-text capabilities.
+
+Example usage:
+    from vision_agents.plugins import fal
+
+    # For transcription
+    stt = fal.STT(task="transcribe")
+
+    # For translation to Portuguese
+    stt = fal.STT(task="translate", target_language="pt")
+
+    @stt.on("transcript")
+    async def on_transcript(text: str, user: Any, metadata: dict):
+        print(f"Transcript: {text}")
+
+    @stt.on("error")
+    async def on_error(error: str):
+        print(f"Error: {error}")
+"""
+
+import io
+import os
+import tempfile
+import time
+import logging
+from pathlib import Path
+from typing import Any, Dict, Optional, List, Tuple, Union, TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from vision_agents.core.edge.types import Participant
+import wave
+
+import fal_client
+from getstream.video.rtc.track_util import PcmData
+from vision_agents.core import stt
+
+logger = logging.getLogger(__name__)
+
+
+class STT(stt.STT):
+    """
+    Audio transcription and translation using fal-ai/wizper (Whisper v3).
+
+    This plugin provides real-time speech-to-text capabilities using the fal-ai/wizper
+    service, which is based on OpenAI's Whisper v3 model. It supports both transcription
+    and translation tasks.
+
+    Attributes:
+        task: The task type - either "transcribe" or "translate"
+        target_language: Target language code for translation (e.g., "pt" for Portuguese)
+    """
+
+    def __init__(
+        self,
+        task: str = "transcribe",
+        target_language: str | None = None,
+        sample_rate: int = 48000,
+        client: Optional[fal_client.AsyncClient] = None,
+    ):
+        """
+        Initialize FalWizperSTT.
+
+        Args:
+            task: "transcribe" or "translate"
+            target_language: Target language code (e.g., "pt" for Portuguese)
+            sample_rate: Sample rate of the audio in Hz.
+        """
+        super().__init__(sample_rate=sample_rate)
+        self.task = task
+        self.target_language = target_language
+        self.last_activity_time = time.time()
+        self._is_closed = False
+        self._fal_client = client if client is not None else fal_client.AsyncClient()
+
+    def _pcm_to_wav_bytes(self, pcm_data: PcmData) -> bytes:
+        """
+        Convert PCM data to WAV format bytes.
+
+        Args:
+            pcm_data: PCM audio data from Stream's audio pipeline
+
+        Returns:
+            WAV format audio data as bytes
+        """
+        wav_buffer = io.BytesIO()
+
+        with wave.open(wav_buffer, "wb") as wav_file:
+            wav_file.setnchannels(1)  # Mono
+            wav_file.setsampwidth(2)  # 16-bit
+            wav_file.setframerate(self.sample_rate)
+            wav_file.writeframes(pcm_data.samples.tobytes())
+
+        wav_buffer.seek(0)
+        return wav_buffer.read()
+
+    async def _process_audio_impl(
+        self, pcm_data: PcmData, user_metadata: Optional[Union[Dict[str, Any], "Participant"]] = None
+    ) -> Optional[List[Tuple[bool, str, Dict[str, Any]]]]:
+        """
+        Process accumulated speech audio through fal-ai/wizper.
+
+        This method is typically called by VAD (Voice Activity Detection) systems
+        when speech segments are detected.
+
+        Args:
+            speech_audio: Accumulated speech audio as numpy array
+            user: User metadata from the Stream call
+        """
+        if self._is_closed:
+            logger.debug("connection is closed, ignoring audio")
+            return None
+
+        if pcm_data.samples.size == 0:
+            logger.debug("No audio data to process")
+            return None
+
+        try:
+            logger.debug(
+                "Sending speech audio to fal-ai/wizper",
+                extra={"audio_bytes": pcm_data.samples.nbytes},
+            )
+            # Convert PCM to WAV format for upload
+            wav_data = self._pcm_to_wav_bytes(pcm_data)
+
+            # Create temporary file for upload
+            with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file:
+                temp_file.write(wav_data)
+                temp_file.flush()
+                temp_file_path = temp_file.name
+
+            try:
+                input_params = {
+                    "task": "transcribe",  # TODO: make this dynamic, currently there's a bug in the fal-ai/wizper service where it only works with "transcribe"
+                    "chunk_level": "segment",
+                    "version": "3",
+                }
+                # Add language for translation
+                if self.target_language is not None:
+                    input_params["language"] = self.target_language
+
+                # Upload file and get URL
+                audio_url = await self._fal_client.upload_file(Path(temp_file_path))
+                input_params["audio_url"] = audio_url
+
+                # Use regular subscribe since streaming isn't supported
+                result = await self._fal_client.subscribe(
+                    "fal-ai/wizper", arguments=input_params
+                )
+                if "text" in result:
+                    text = result["text"].strip()
+                    if text:
+                        self._emit_transcript_event(
+                            text, user_metadata, {"chunks": result.get("chunks", [])}
+                        )
+            finally:
+                # Clean up temporary file
+                try:
+                    os.unlink(temp_file_path)
+                except OSError:
+                    pass
+
+            # Return None for asynchronous mode - events are emitted when they arrive
+            return None
+
+        except Exception as e:
+            logger.error(f"FalWizper processing error: {str(e)}")
+            self._emit_error_event(e, "FalWizper processing")
+            return None
+
+    async def close(self):
+        """Close the STT service and release any resources."""
+        if self._is_closed:
+            return
+        self._is_closed = True
+        logger.info("FalWizperSTT closed")

vision_agents_plugins_wizper-0.0.17.dist-info/METADATA

@@ -0,0 +1,15 @@
+Metadata-Version: 2.4
+Name: vision-agents-plugins-wizper
+Version: 0.0.17
+Summary: Wizper plugin 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,agents,fal,turn detection,voice agents
+Requires-Python: >=3.10
+Requires-Dist: fal-client>=0.7.0
+Requires-Dist: vision-agents
+Description-Content-Type: text/markdown
+
+Wizper for vision agents

vision_agents_plugins_wizper-0.0.17.dist-info/RECORD

@@ -0,0 +1,9 @@
+./.gitignore,sha256=LiiMFm4RUXduFZI42AL85GrllvISRUwTt4t3lrUDGUE,408
+./PKG-INFO,sha256=WruHzrEAIpbxPodXi66z2-1VN1D8dknBCufm10gnYnw,505
+./README.md,sha256=7MDH68Ywzj2WKm3QAFCUvupBHxrTdjtAL0WqrqaCHFc,24
+./pyproject.toml,sha256=eSC8A7YqeCWd9_VfDiwOc9z3KPGzSPbf9IIqk_aWES0,959
+./vision_agents/plugins/wizper/__init__.py,sha256=aRLgDFc3zq4tNj3G9kmM4zJzSpO7hYqGujz3zTTPsMk,93
+./vision_agents/plugins/wizper/stt.py,sha256=dRJV4klITwZ1SKbe-MPiC_s7idDJGBTDfmGQlh5E6ss,6112
+vision_agents_plugins_wizper-0.0.17.dist-info/METADATA,sha256=WruHzrEAIpbxPodXi66z2-1VN1D8dknBCufm10gnYnw,505
+vision_agents_plugins_wizper-0.0.17.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+vision_agents_plugins_wizper-0.0.17.dist-info/RECORD,,

vision_agents_plugins_wizper-0.0.17.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.27.0
+Root-Is-Purelib: true
+Tag: py3-none-any