Integrate Google Camb-Vertex AI with crewAI

crewai_tools/__init__.py

@@ -67,6 +67,7 @@
     SpiderTool,
     StagehandTool,
     TXTSearchTool,
+    VertexCambAITool,
     VisionTool,
     WeaviateVectorSearchTool,
     WebsiteSearchTool,

crewai_tools/tools/__init__.py

@@ -81,6 +81,7 @@
 from .spider_tool.spider_tool import SpiderTool
 from .stagehand_tool.stagehand_tool import StagehandTool
 from .txt_search_tool.txt_search_tool import TXTSearchTool
+from .vertex_cambai_tool.vertex_cambai_tool import VertexCambAITool
 from .vision_tool.vision_tool import VisionTool
 from .weaviate_tool.vector_search import WeaviateVectorSearchTool
 from .website_search.website_search_tool import WebsiteSearchTool

crewai_tools/tools/vertex_cambai_tool/README.md

@@ -0,0 +1,243 @@
+# Vertex CambAI Tool
+
+The Vertex CambAI Tool provides voice cloning capabilities using CambAI's MARS7 model deployed on Google Cloud Vertex AI. This tool requires reference audio and transcription for all voice synthesis.
+
+## Features
+
+- **Voice Cloning with MARS7**: High-quality voice cloning using CambAI's advanced model
+- **Multilingual Support**: English (en-us) and Spanish (es-es)
+- **FLAC Audio Output**: Professional-grade audio format
+- **Google Cloud Integration**: Deployed on Vertex AI infrastructure
+
+## Installation
+
+Install the required dependencies:
+
+```bash
+pip install google-cloud-aiplatform soundfile
+```
+
+Or install with optional dependencies:
+
+```bash
+pip install crewai-tools[vertex-cambai]
+```
+
+## Setup
+
+### Required Environment Variables
+
+Set these environment variables before using the tool:
+
+```bash
+export GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account-key.json"
+export PROJECT_ID="your-google-cloud-project-id"
+export LOCATION="us-central1"
+export ENDPOINT_ID="your-vertex-ai-endpoint-id"
+```
+
+### Google Cloud Setup
+
+1. Create a Google Cloud service account with Vertex AI access
+2. Download the service account key file
+3. Deploy the MARS7 model on Vertex AI (contact CambAI for access)
+4. Get your endpoint ID from the Vertex AI console
+
+## Usage
+
+### Basic Voice Cloning
+
+```python
+from crewai_tools import VertexCambAITool
+
+# Initialize tool (requires environment variables)
+tts_tool = VertexCambAITool()
+
+# Synthesize speech with voice cloning
+result = tts_tool.run(
+    text="Hello, this will sound like my voice!",
+    audio_ref_path="./my_voice_reference.wav",
+    ref_text="This is the transcription of my reference audio",
+    language="en-us",
+    output_file="cloned_speech.flac"
+)
+
+print(result)
+```
+
+### Multilingual Voice Cloning
+
+```python
+from crewai_tools import VertexCambAITool
+
+tts_tool = VertexCambAITool()
+
+# English voice cloning
+english_result = tts_tool.run(
+    text="Hello, how are you today?",
+    audio_ref_path="./english_ref.wav",
+    ref_text="Hello, this is my English voice",
+    language="en-us",
+    output_file="english_output.flac"
+)
+
+# Spanish voice cloning
+spanish_result = tts_tool.run(
+    text="Hola, ¿cómo estás hoy?",
+    audio_ref_path="./spanish_ref.wav", 
+    ref_text="Hola, esta es mi voz en español",
+    language="es-es",
+    output_file="spanish_output.flac"
+)
+```
+
+## Parameters
+
+### Required Parameters
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `text` | str | Text to synthesize |
+| `audio_ref_path` | str | Path to reference audio file |
+| `ref_text` | str | Transcription of reference audio |
+
+### Optional Parameters
+
+| Parameter | Type | Default | Description |
+|-----------|------|---------|-------------|
+| `language` | str | `"en-us"` | Language code (en-us, es-es) |
+| `output_file` | str | `"output.flac"` | Output audio file path |
+
+## Response Format
+
+```json
+{
+  "status": "success",
+  "message": "Audio saved to output.flac",
+  "audio_file_path": "/absolute/path/to/output.flac",
+  "text": "Synthesized text",
+  "language": "en-us"
+}
+```
+
+## Error Handling
+
+```json
+{
+  "status": "error",
+  "message": "Error description",
+  "audio_file_path": null
+}
+```
+
+## Supported Audio Formats
+
+- **Input**: WAV, FLAC, MP3, M4A, OGG, WEBM
+- **Output**: FLAC (default)
+
+## Testing
+
+Run the test script to verify your setup:
+
+```bash
+python test_vertex_cambai_example.py
+```
+
+Make sure all environment variables are set before testing.
+
+## Integration with CrewAI
+
+```python
+from crewai import Agent, Task, Crew
+from crewai_tools import VertexCambAITool
+
+# Initialize tool
+voice_tool = VertexCambAITool()
+
+# Create agent
+voice_agent = Agent(
+    role="Voice Cloning Specialist",
+    goal="Convert text to cloned speech",
+    backstory="Expert in voice cloning with MARS7",
+    tools=[voice_tool]
+)
+
+# Create task
+task = Task(
+    description="Clone voice saying: 'Welcome to CrewAI voice cloning!'",
+    agent=voice_agent
+)
+
+# Run
+crew = Crew(agents=[voice_agent], tasks=[task])
+result = crew.kickoff()
+```
+
+## Troubleshooting
+
+### Common Issues
+
+**Missing Environment Variables**
+```bash
+ValueError: Missing required environment variables: PROJECT_ID, ENDPOINT_ID
+```
+Solution: Set all required environment variables
+
+**Import Error**
+```bash
+ImportError: Please install google-cloud-aiplatform
+```
+Solution: `pip install google-cloud-aiplatform soundfile`
+
+**File Not Found**
+```bash
+FileNotFoundError: [Errno 2] No such file or directory: './reference.wav'
+```
+Solution: Provide correct path to reference audio file
+
+**Authentication Error**
+```bash
+DefaultCredentialsError: Could not automatically determine credentials
+```
+Solution: Check GOOGLE_APPLICATION_CREDENTIALS path
+
+## Example Script
+
+```python
+#!/usr/bin/env python3
+import os
+import json
+from crewai_tools import VertexCambAITool
+
+# Set environment variables
+os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "/path/to/key.json"
+os.environ["PROJECT_ID"] = "your-project"
+os.environ["LOCATION"] = "us-central1"
+os.environ["ENDPOINT_ID"] = "your-endpoint"
+
+# Initialize and use tool
+tool = VertexCambAITool()
+result = tool.run(
+    text="This is a test of voice cloning",
+    audio_ref_path="./reference.wav",
+    ref_text="Hello, this is my reference voice",
+    language="en-us"
+)
+
+# Parse result
+data = json.loads(result)
+if data["status"] == "success":
+    print(f"✅ Audio saved: {data['audio_file_path']}")
+else:
+    print(f"❌ Error: {data['message']}")
+```
+
+## License
+
+This tool is part of the CrewAI Tools package.
+
+## Support
+
+- **Tool Issues**: CrewAI Tools GitHub repository
+- **MARS7 Model**: Contact CambAI support
+- **Google Cloud**: Google Cloud documentation

crewai_tools/tools/vertex_cambai_tool/vertex_cambai_tool.py

@@ -0,0 +1,111 @@
+"""Vertex CambAI Tool for text-to-speech synthesis using Google Cloud Vertex AI."""
+
+import base64
+import json
+import os
+from typing import Type
+
+from crewai.tools import BaseTool
+from pydantic import BaseModel
+from google.cloud import aiplatform
+
+class VertexCambAISchema(BaseModel):
+    """Input schema for Vertex CambAI Tool."""
+
+    text: str
+    audio_ref_path: str
+    ref_text: str
+    language: str = "en-us"
+    output_file: str = "output.flac"
+
+
+class VertexCambAITool(BaseTool):
+    """Tool for text-to-speech synthesis using Google Cloud Vertex AI with CambAI's MARS7 model.
+
+    This tool provides voice cloning capabilities using the MARS7 model.
+    Requires Google Cloud credentials and Vertex AI endpoint access.
+
+    Environment Variables Required:
+        GOOGLE_APPLICATION_CREDENTIALS: Path to service account key file
+        PROJECT_ID: Google Cloud project ID
+        LOCATION: Google Cloud location (e.g., us-central1)
+        ENDPOINT_ID: Vertex AI endpoint ID for MARS7 model
+    """
+
+    name: str = "Vertex CambAI Tool"
+    description: str = (
+        "Converts text to speech using CambAI's MARS7 model on Google Cloud Vertex AI. "
+        "Requires reference audio and transcription for voice cloning."
+    )
+    args_schema: Type[BaseModel] = VertexCambAISchema
+    package_dependencies: list = ["google-cloud-aiplatform", "soundfile"]
+
+    def __init__(self, **kwargs):
+        """Initialize the Vertex CambAI Tool."""
+        super().__init__(**kwargs)
+
+        # Check required environment variables
+        required_vars = ["GOOGLE_APPLICATION_CREDENTIALS", "PROJECT_ID", "LOCATION", "ENDPOINT_ID"]
+        missing_vars = [var for var in required_vars if not os.getenv(var)]
+
+        if missing_vars:
+            raise ValueError(f"Missing required environment variables: {', '.join(missing_vars)}")
+
+        # Initialize Vertex AI
+        try:
+            aiplatform.init(
+                project=os.getenv("PROJECT_ID"),
+                location=os.getenv("LOCATION")
+            )
+        except ImportError:
+            raise ImportError("Please install google-cloud-aiplatform: pip install google-cloud-aiplatform")
+
+    def _run(self, **kwargs) -> str:
+        """Execute the text-to-speech synthesis."""
+        try:
+            # Validate inputs
+            schema = VertexCambAISchema(**kwargs)
+
+            # Encode reference audio
+            with open(schema.audio_ref_path, "rb") as f:
+                audio_ref_base64 = base64.b64encode(f.read()).decode("utf-8")
+
+            # Build prediction instances
+            instances = {
+                "text": schema.text,
+                "audio_ref": audio_ref_base64,
+                "ref_text": schema.ref_text,
+                "language": schema.language
+            }
+
+            # Make prediction
+            endpoint = aiplatform.Endpoint(endpoint_name=os.getenv("ENDPOINT_ID"))
+            response = endpoint.raw_predict(
+                body=json.dumps({"instances": [instances]}).encode("utf-8"),
+                headers={"Content-Type": "application/json"}
+            )
+
+            # Save audio response
+            predictions = json.loads(response.content)["predictions"]
+            audio_bytes = base64.b64decode(predictions[0])
+
+            with open(schema.output_file, "wb") as f:
+                f.write(audio_bytes)
+
+            result = {
+                "status": "success",
+                "message": f"Audio saved to {schema.output_file}",
+                "audio_file_path": os.path.abspath(schema.output_file),
+                "text": schema.text,
+                "language": schema.language
+            }
+
+            return json.dumps(result, indent=2)
+
+        except Exception as e:
+            error_result = {
+                "status": "error",
+                "message": str(e),
+                "audio_file_path": None
+            }
+            return json.dumps(error_result, indent=2)

pyproject.toml

@@ -115,6 +115,12 @@ xml = [
 oxylabs = [
     "oxylabs==2.0.0"
 ]
+vertex-cambai = [
+    "google-cloud-aiplatform>=1.98.0",
+    "soundfile>=0.13.1",
+    "python-dotenv>=1.1.0",
+    "numpy>=1.21.0",
+]
 
 [tool.pytest.ini_options]
 pythonpath = ["."]