Private health questions deserve care.
NutritiScan offers a public health conversation, a local Android journal, and a separate shared health workspace. Shared accounts and private report storage require a deployed health service; the dashboard displays its connection status. Public-chat attachments and shared report uploads have different data boundaries, described below.
NutritiScan is designed for people of all ages. Younger users should review this notice with a parent or guardian before sharing personal health information. A verified parent or guardian consent flow for children's health records is not yet available; do not upload or store a child's records in the shared account until that safeguard is provided.
What stays in your browser
Chat history, your optional health profile, and meal notes are saved in this browser so they can be available in later conversations. They are not synced to a NutritiScan account. Anyone with access to this browser profile may be able to view them. Use your browser's site-data controls to remove locally saved information.
The separate synthetic discharge demo saves its fictional workflow steps, review notes and draft messages in this browser so a page refresh can restore them. They are not sent to NutritiScan's server. Use fictional information only in that demo. Its Start over control clears the saved steps for the selected fictional case. Clear site data to remove every saved demo case.
Research conversations and account backups
Research history and saved source-review notes stay in this browser until you choose to save an account snapshot or download a backup file. Account snapshots contain the selected device research history, source records, drafts and saved notes. They are stored encrypted by the account service and accessed using your signed-in account. Health profiles, meal logs and health chats are not included. Research backup actions do not call an AI model.
You can preview and restore the account snapshot on another device, or explicitly delete it from Research conversation history. Deleting the snapshot does not erase device copies or downloaded files. Account deletion removes the stored research snapshot as well. Hosting backup copies follow the host’s retention policy. Research searches and model-assisted follow-ups still send their selected public research context to configured sources and AI providers; saved source-review notes are excluded from model context.
Mac desktop research preview
The separate Mac app keeps project files in the folder you choose and saved app history in local Application Support storage. App history is plain JSON with owner-only file permissions, not encrypted by NutritiScan. Anyone with access to your macOS account may be able to read it. Desktop work does not automatically sync with browser chats or account research backups.
Importing a paper saves an original copy and an extracted text snapshot inside the chosen project. PDF extraction, previews and manual passage checks run locally. Passages requested during agent chat are sent to the connected model provider; importing alone does not send the PDF to that provider. Fetching open full text sends the article identifier to the official PMC service and retains the returned licence and attribution.
The desktop agent sends messages, requested file content, retrieved evidence and project context to the model through your installed Codex runtime. Public literature queries go to PubMed and ClinicalTrials.gov. Codex keeps its own sign-in; NutritiScan does not copy its credentials. The terminal can launch the original Claude Code CLI with its own account and data handling. Use public or de-identified research data.
Archiving retains saved history. Removing archived work asks for confirmation, then removes that history from this app; project files, saved reports, exports and original Codex history remain separate. Restore and export a conversation before removing it if you need a copy.
What is sent when you chat
When you send a message, the chat sends recent conversation messages and the profile and meal context saved in this browser to NutritiScan's server. The server sanitizes this input and runs urgent-symptom checks before agent reasoning. Depending on deployment configuration, the question and context may be processed by a hosted AI provider. Provider handling is subject to that provider's terms and retention practices. Do not include information you are not comfortable sending.
Chat report attachments are read in your browser. The original file is not uploaded as a file, but up to 12,000 extracted text characters are added to the conversation, saved with chat history in this browser, and sent with your message to NutriScan's server and, when configured, a hosted AI provider. Remove the message or clear this site's browser data to remove that local copy.
Android local data and public chat
The Android app saves local profiles, meal entries, health history and conversations on your device using app storage. NutritiScan does not encrypt that local journal storage. Public AI chat asks for consent before sending chosen messages to the web server and configured AI provider; the local profile and meal journal are not silently attached. You can revoke public-chat consent and delete local entries in the app’s You screen. Barcode lookup sends a product code to Open Food Facts, not your health profile.
Shared account, reports and consent
When connected, the shared health service stores records and conversations under your account. Health content and original files are encrypted by the service. Native access tokens use secure device credential storage; web sessions use an HttpOnly cookie. Explicit health storage consent is required, and cloud AI processing is a separate optional choice. The service records access and change events without clinical payloads.
Uploaded shared reports go to private object storage. Extracted values stay as drafts until you compare them with the original and confirm the name, value, unit and collection date. Patient confirmation is not clinical validation. Structured records support FHIR export; the prototype is not a certified FHIR server. NutritiScan does not use patient records for model training. Any configured provider’s retention and processing terms must be reviewed before real patient use.
Voice, export and deletion
Voice transcription requires an enabled provider and cloud AI consent. Audio is sent for transcription after you choose to transcribe, is not added to health history by this service, and the transcript is reviewed before a question is sent. The provider may have separate retention terms. Shared records and conversation history can be exported from the dashboard; originals are downloaded separately. Account deletion requires your password or a sign-in from the last 15 minutes. It removes active records, documents and sessions. Infrastructure backups follow the host’s retention policy. Android local entries are deleted separately in the app. Revoking storage consent stops routine access and new storage but keeps export and deletion available.
Safety and limits
Automated checks can miss urgent symptoms or misunderstand language. NutritiScan is educational support, not a diagnosis, prescription, emergency service, or replacement for a qualified clinician.
Contact
For privacy questions, contact adarshbhardwaj9182@gmail.com. Do not send health information in public GitHub issues.
Back to chat