Project case study
ContentOS
Air-gapped AI content platform. Text generation via Ollama, image generation via Stable Diffusion. Nothing leaves the machine.
Problem
- Teams wanted AI content tools but could not send proprietary data to cloud APIs
- Existing options were either cloud-only or assumed expensive GPU infrastructure
Constraints
- Zero cloud calls, air-gapped by design
- Both text and image generation
- Has to run on consumer-grade GPUs
- Human review before anything is published
Architecture
- FastAPI coordinates local text and image models
- React frontend runs a draft, review, and publish pipeline
- All data stays on the machine
Tradeoffs
- Ollama over hosting raw models. Operational simplicity matters when you are also the ops person
- Smaller local models over frontier APIs. Privacy mattered more than peak output quality for this use case
- Mandatory review step over auto-publish. Cost of a bad publish is much higher than the cost of a human click
Outcome
- A working local AI content platform that meets the zero-cloud constraint
- No data leaves the box
- Runs fine on a single consumer GPU