How Formatted Chinese Runs

Back to FormattedChinese.com

FormattedChinese.com is powered by a small local setup. The website runs on a Raspberry Pi, while translation and OCR requests are handled by local Ollama models on our own GPU machine.

Local GPU machine with MSI GeForce RTX GPUs and Samsung SSD
Two RTX 5060 GPUs
Our local LLMs run here with 16GB total VRAM.
Samsung SSD
The model weights are stored here.

The local LLM box

This machine runs the local models used for translation and image text extraction. The model files live on the Samsung SSD inside the same machine. Today this beta runs on two RTX 5060 GPUs with 16GB total VRAM. If FormattedChinese.com gains traction, we plan to upgrade the hardware so we can run stronger local models.

Raspberry Pi 5 hosting the FormattedChinese.com web app
Raspberry Pi 5
The website runs here.

The website server

The Raspberry Pi serves the web app, receives your request, and sends work to the local LLM machine only when needed.

Your data is not stored here. Your data is yours only.

Text and images are processed transiently for the task you requested. They are not sold, not used for training, and not stored in a database. Completed outputs are removed from server memory after delivery, and abandoned jobs expire automatically.