🧠 Transcript AI: Deployment on Google Cloud Run

Transcript AI is a production-grade, persona-driven synthesis engine built specifically for Google Cloud Run. It leverages Gemini Native Audio Intelligence to analyze multiple audio recordings and text transcripts simultaneously, generating cohesive executive reports without the need for heavy local dependencies like Whisper or FFmpeg.

---

🚀 Key Features

• Native Audio Intelligence: Directly processes .m4a, .mp3, and .wav files via the Gemini Files API. The AI "hears" the audio, preserving tone and context.
• Multi-Source Synthesis: Upload multiple audio and text files (Zoom/Teams transcripts) together. The engine synthesizes them into a single integrated report.
• Executive Persona Engine: Selectable AI roles (Software Architect, Sales AE, Technical Recruiter, etc.) with specialized reporting templates.
• Professional UI: An executive-styled Streamlit interface featuring custom CSS for a high-end enterprise feel.

---

📂 Project Structure

gcp-transcript-engine/
├── .streamlit/
│   └── config.toml      # Production server & websocket settings
├── app.py               # Main application logic (Gemini Native)
├── personas.py          # AI role definitions and prompt templates
├── requirements.txt     # Lightweight Python dependencies
├── Dockerfile           # OWASP-compliant, non-root container
└── deploy.ps1           # Automated one-click deployment script

---

☁️ Deployment to Google Cloud Run

This project uses a "Zero Local Install" model. You do not need to install the Google Cloud SDK on your machine; everything is handled via a Dockerized CLI.

1. Prerequisites

• A Google Cloud Project ID.
• A Gemini API Key (Free) from Google AI Studio.
• Docker Desktop running on your Windows/Mac/Linux machine.

2. Automated One-Click Deploy

1. Navigate to the gcp-transcript-engine directory.
2. Open deploy.ps1 and update the $PROJECT_ID at the top of the script.
3. Execute the script in PowerShell:

   Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
   ./deploy.ps1

What the script does:

• Authenticates your session via a persistent Docker volume.
• Automatically enables required APIs (Cloud Run, Cloud Build, Artifact Registry).
• Builds the image on Google Cloud and deploys the service.

---

🛡️ Security & Compliance

This application is built following OWASP best practices:

• Non-Root Execution: The container runs as appuser to minimize the attack surface.
• Ephemeral Secrets: No API keys are stored on the server. The "Bring Your Own Key" (BYOK) model keeps keys in transient browser memory.
• Data Privacy: Audio files are stored in an encrypted /tmp RAM-disk and are deleted immediately after the upload to Gemini is finalized.
• Secure Networking: Streamlit is configured to run behind the Google Cloud HTTPS proxy with XSRF protection enabled via config.toml.

---

⚙️ Configuration & Model Selection

The application is pre-configured with the latest stable Gemini model families:

• Gemini 3.5 Flash: Recommended default for speed and intelligence.
• Gemini 3.1 Pro: Best for complex reasoning in long technical documents.
• Gemini 2.5 Flash: High-performance alternative for high-volume tasks.

---

📄 License

This project is licensed under the MIT License.

---

Disclaimer: This is an experimental AI project. Users are responsible for ensuring their API usage complies with Google’s Generative AI Prohibited Use Policy.