Sentinel is content analysis starter code to support pharmaceutical regulatory affairs and marketing teams. Sentinel starter code can be used as a starting point to construct more efficient workflow applications in the Pharma Industry. Leveraging the power of Google Gemini AI, Sentinel code can help build automation workflows for the review process for promotional advertisements. Its core function is to flag potential issues, verify citation integrity, and check adherence to established industry standards, thereby identifying and annotating areas for expert review.

Key Features and Functionality:

• Pharma Ad Video Checker: Sentinel code performs analysis of video advertisements. It is designed to identify and flag potential issues, providing timestamped markers to facilitate precise editing and review.
• Ad Infographic Checker: The code offers capabilities for reviewing advertisements, and infographics. It provides annotations that highlight potential issues hence speeding up the checking process.
• Interactive Professional User Interface (UI): Sentinel features a clean, efficient interface. This UI provides regulatory and marketing teams with real-time feedback and generates reports, enhancing workflow efficiency and transparency using the Sentinel code.

Target Users:

Sentinel code is designed for pharmaceutical regulatory affairs and marketing teams responsible for checking promotional materials.

Technology and Architecture:

Sentinel code is built with a FastAPI (Python) backend, an HTML/CSS/JavaScript frontend utilizing Google Material Design, and integrates with the Google Gemini AI API for its analytical capabilities. It is designed for deployment on Cloud Run.

Important Note:

Sentinel code is a content analysis starter code and is for administrative and operational support only. It is not intended for any medical purpose, including diagnosis, prevention, monitoring, treatment, or alleviation of disease, injury, or disability, nor for the investigation, replacement, or modification of anatomy or physiological processes, or for the control of conception. Its function is solely to assist regulatory and marketing teams in identifying potential issues in pharmaceutical content. All outputs from Sentinel code should be considered preliminary, require independent verification and further investigation through established company process and methodologies for determining regulatory compliance for marketing content. This code is not intended to be used without appropriate validation, adaptation and/or making meaningful modifications by developers for their specific workflows. This code is intended as a developer accelerator and proof-of-concept. It has not undergone software validation (CSV), penetration testing, or quality assurance required for production environments in regulated industries. Deploying this code into a production workflow without significant modification, security hardening, and appropriate validation is at the user's sole risk.

For technical details on how to configure these settings, please refer to the official Vertex AI Zero Data Retention Documentation.

• Python 3.12+
• uv package manager (required - faster than pip/venv)
• Node.js and npm (for frontend)
• Google Cloud Project (for Vertex AI - preferred) OR Google Gemini API key (for AI Studio)
• Google Cloud SDK (for Vertex AI authentication)

1. Clone the repository:

   git clone https://github.com/GoogleCloudPlatform/LifeSciences
   cd LifeSciences

2. Backend Setup:

   # Create and activate a virtual environment
   uv venv --python "python3.12" ".venv"
   source .venv/bin/activate
   
   # Install all dependencies
   uv sync --all-extras

3. Frontend Setup:

   cd frontend
   npm install

4. Configure Environment Variables:

   Copy .env.example to .env.

   • Vertex AI (Recommended):

     • Set GOOGLE_GENAI_USE_VERTEXAI=true.
     • Set GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION.
     • Set GCS_BUCKET_NAME for image features.
     • Authenticate: gcloud auth application-default login.

   • AI Studio:

     • Set GEMINI_API_KEY.

5. Run the Development Servers:

   • Backend:

     uv run python -m api.main

     The API will run at http://localhost:8000.

   • Frontend:

     cd frontend
     npm run dev

     The frontend will run at http://localhost:5173.

The project uses pytest for unit testing the API services and routes.

uv run pytest

The application is containerized using a multi-stage Docker build that serves the React frontend via the FastAPI backend.

1. Build the Image:

   docker build -t sentinel .

2. Run the Container:

   • Run with AI Studio (API Key)

     docker run -p 8080:8080 --env-file .env sentinel

   • Run with Vertex AI

     You need to provide your Google Cloud credentials to the container.

     # Authenticate locally first
     gcloud auth application-default login
     
     # Run container with mounted credentials and environment file
     # We use --user to ensure the container can read the mounted credentials
     docker run -p 8080:8080 \
     --user $(id -u):$(id -g) \
     -v ~/.config/gcloud/application_default_credentials.json:/app/gcp_creds.json \
     -e GOOGLE_APPLICATION_CREDENTIALS=/app/gcp_creds.json \
     --env-file .env \
     sentinel

sentinel/
├── api/                    # FastAPI backend
│   ├── routes/            # API route handlers (analysis, storage, health)
│   ├── services/          # Business logic (Gemini client, analyzer service)
│   ├── models/            # Pydantic schemas
│   ├── config.py          # Configuration management
│   └── main.py            # FastAPI application entry point
├── tests/                  # Unit test suite
├── frontend/              # React frontend
│   ├── src/               # TypeScript source files
│   ├── public/            # Static assets
│   └── index.html         # SPA entry point
├── Dockerfile              # Multi-stage Docker build
├── pyproject.toml         # Python dependencies and metadata
└── README.md               # This file

• GET /health - Health check
• POST /api/v1/analyze - Analyze video (URL) or image (URL)
• POST /api/v1/analyze/upload - Analyze uploaded image file
• GET /api/v1/storage/list - List files in GCS
• POST /api/v1/storage/upload - Upload file to GCS
• GET /api/v1/storage/file/{path} - Retrieve/stream file from GCS

1. Select "YouTube Video" from the content type dropdown
2. Paste the YouTube URL
3. Optionally adjust the frame rate (lower = fewer tokens used)
4. Click "Analyze"

1. Select "Image URL" or "Upload Image"
2. Provide the image URL or select a file
3. Click "Analyze"
4. Click on numbered markers to see issue details

Contributions are welcome! Please feel free to submit a Pull Request.