Server-Side Whisper & Security Upgrade

I’m refactoring an existing open-source app that currently runs Whisper speech-to-text and a lightweight LLM in the browser. My objective is to shift both engines to the backend so the heavy lifting happens on the server, giving users noticeably faster response times and freeing their devices from large model downloads. Here’s the work that needs to happen: • Move Whisper and the current LLM out of the front-end bundle and deploy them on the server (Python/FastAPI or Node/Express—feel free to suggest what best matches the original codebase). • Expose clean, authenticated endpoints so the front end simply posts audio or text and receives the model’s response. • Wrap all traffic in a solid SSL/TLS layer that covers data encryption, identity verification, and fully secured communication channels. I’ll provide the certificate or we can generate one with Let’s Encrypt as long as the chain installs correctly. • Migrate the existing SQLite store to PostgreSQL and switch it to full encryption at rest (pgcrypto or another approach you recommend). • Patch any dependencies that open known CVEs discovered during the move; a brief report on what you updated is enough. • Update the README so anyone can pull, run docker-compose up, and have the whole stack—models, database, and certificates—come online. Acceptance criteria – Whisper and the LLM run exclusively server-side and answer within the current latency budget. – All endpoints pass SSL Labs A rating. – Database files are encrypted on disk and can be decrypted only by the running service. – Clean install and test instructions in the repo. If you’ve already containerised similar AI workloads or hardened a Postgres deployment, you’ll be able to move quickly. Let me know the stack you’d prefer and any prior work that shows you’ve done this sort of migration before.

Projekt-ID: 40089331