Zero-Retention Architecture
The pipeline is deliberately built so that document text has nowhere to land: it exists only in the memory of the request that produced it.
Request lifecycle
The browser sends text to a server function, the server function forwards it to the language model, the rewritten result returns to the browser, and the request ends. No intermediate storage step exists.
Metadata only
The usage tables record a timestamp, the detected document type, the word count and a status value. There is no column capable of holding document text.
Client-side verification
Integrity checking of codes, dosages and laboratory values happens in your browser on text already in memory, so verification adds no additional transmission.