Healthcare
Anamnesis
Assembles a medical history from scattered PDFs and lab scans, then reasons across it on-device
Health records are about as private as data gets, which makes shipping them to a cloud LLM a non-starter for most people. Anamnesis assembles a medical history from the scattered lab scans, PDFs, and visit notes already on a disk, then lets a team of on-device agents reason across all of it - connecting a low hemoglobin, low ferritin, and a visit note into "iron-deficiency anemia," citing the dates it used.
Gather, then synthesize
An orchestrator (Qwen3-1.7B) calls tools to pull evidence from a patient's own records - semantic search over a local RAG store, lab trends from a structured database, medication history - before a clinical synthesizer (MedGemma-4B) reasons over what it gathered and explains it in plain language. Each model does what it's best at: a small model routes, a medical model reasons.
a scan / note
--image?--> OCR (QVAC) --text--> MedGemma normalizer (strict JSON schema)
--text-------------------------------┘
│
structured store (lab trends) RAG vector store (GTE-large)
│ │
user question --> ORCHESTRATOR (Qwen3, tool-calling) <───────┘
tools: search_records · lab_trend · list_analytes · list_medications
│
CLINICAL SYNTHESIZER (MedGemma 4B)
grounded answer · cites dates · plain language · disclaimerDemo video
Why it stays local by design
MORE FROM THE SHOWCASE
Build this
Ship your own local AI
One install in JavaScript or Python. 10+ AI tasks. Private, offline and free - on the hardware you already own.