Lumen
A desktop app for language models that run on your own machine, with no accounts or cloud services. It brings everyday assistants and fiction writing into one place, with replies streamed from the local model. I made Lumen on my own from the first idea, directing AI coding agents through the whole build: the product and its features, the brand, the interface, and the code that runs the app and connects it to the model.
- Year
- 2026
- Role
- Sole maker, from idea to working app
- Type
- Personal project
- Platform
- Desktop app for local language models
- Built with
- AI coding agents, directed and reviewed by me
- Work
- Product · brand · interface · app code
All assistants, chats and texts shown are demo content.
Assistants
Each assistant has its own role and instructions. Replies stream from the local model as they are generated and can be regenerated or read aloud. Two assistants can share a chat and take turns.
Fiction writing
The writer adds the next part or gives a direction and lets the model continue. Narration and speech are set apart, and a rolling summary keeps long texts within the model’s context window.
Brand
Night navy and amber, set in Jost.
