Capture anything
A photo of your plate, a voice note about how you slept, a line of text — or one tap for water, pee, poo and weight. hale timestamps it on the spot.
A personal health journal
One honest record of your health.
hale keeps meals, drinks, symptoms, measurements, meds and bathroom events on one timeline. Snap a photo, say it out loud, or tap once — and look back at a record you actually kept, instead of five apps you gave up on.
What it does
Log it the moment it happens.
The screens above are the real app, running with a day of sample entries.
Works without signal
Captures go into an on-device queue first and upload when you're back online, so a dead zone never costs you an entry.
Structured for you
An AI model turns a photo or voice note into proposed entries — a lunch, a coffee, a headache with its severity. Quick taps skip the model entirely.
You confirm what counts
Proposals wait in the Inbox until you confirm, edit or reject them. Nothing the model guessed becomes part of your record on its own.
See the day, and the pattern
One timeline, newest first, with today's water, meals and weight at the top. Step through day by day, or open the table to scan by type.
Plain, kind words
Bristol scale, symptom severity and doses are there when you want them, in everyday labels — never a clinical form to fill in.
How it works
From a snapshot to an entry.
-
01 — CAPTURE
Snap, say or tap
Use the big button from anywhere in the app. hale installs to your home screen as an app.
PhotoVoiceTextQuick log -
02 — PARSE
The model proposes
Your capture is read by the AI model the server is configured with, and comes back as one or more typed entries with a confidence.
MealDrinkSymptomMeasurementMed -
03 — REVIEW
You keep the record
Confirm, fix or reject each proposal in the Inbox. Confirmed entries join the timeline at the time they happened.
ConfirmEditReject
Your data
A health record belongs to the person it describes.
hale runs on a server you control. No ads, no analytics, no data broker in the loop.
Where it lives
Entries in a SQLite (libSQL) database, photos and voice notes in S3-compatible storage — both on the server you run.
Which model reads it
Any OpenAI-compatible endpoint you point it at, including a local Ollama. Leave it unset and hale still works, without AI parsing.
Who gets in
Sign-in is invite-only: an allowlist of email addresses decides who has an account. Each person sees only their own entries.
Get started
Run it yourself today.
There's no public sign-up yet. With access to the repository, hale runs locally with Bun in four commands.
From a checkout
bun install cp .env.example .env # local sign-in bypass is on by default bun run db:migrate # create the SQLite schema bun run dev # app on http://localhost:52811
Add an OpenAI-compatible model in .env to turn on photo and voice parsing. docker compose up runs the full stack instead.