OfflineNotebook & SettleLah: Full-Stack Project Archive

Building custom utilities is about solving exact problems with zero friction. Here is a look at my two most recent development environments.

SettleLah: AI-Assisted Receipt Splitting

Designed to remove the manual data entry from group dining, SettleLah is a cross-platform mobile and web application. The core logic relies on an AI-assisted OCR pipeline that automatically parses transaction costs, quantities, and service fees directly from a photo of a receipt.

• Architecture: Built with a glassmorphic UI theme.
• Data Handling: Features completely localized data persistence.

OfflineNotebook

A standalone, local desktop application built to interact with LLMs without relying on external cloud APIs.

• Frontend: Built using Electron.
• Backend: A containerized FastAPI Uvicorn application.
• Engine: Integrated with an offline Ollama inference engine, utilizing asynchronous non-blocking model cloning hooks and specifically pinned NumPy C-extensions.

Check out the GitHub Repo for the full source code and Docker compose files.