Vibe Coding Discoverimajev
View on GitHubOpen Jev-style typed-decision model that also takes images: photo + app state + typed questions in, calibrated probabilities out, locally.
Open 2B, 4B, and 9B multimodal models for typed business decisions from images, records, and text. Returns calibrated probabilities and an explicit unknown option; supports local inference and a Jev-style API.
Use Cases
Check product listings against photosCompare returned items with reference imagesInspect parts on a production lineRoute and classify support ticketsVerify records against business policiesFlag uncertain cases for human review
Built With
- Language
- Python
- Frameworks
- Transformers · PEFT · PyTorch · MLX-VLM · FastAPI
Tags
multimodal · vision-language · structured-output · typed-decisions · calibrated-probabilities · abstention · local-inference · open-weights · classification