Integración domótica LibreLinkUp, gratuita y de código abierto
Glucose of the FreeStyle Libre sensors shared with LibreLinkUp, read only. Not a medical device.
Publicado por LM1LC3N7
- Dispositivos
- Versión 0.1.1
- Requiere Gladys >=5.1.0
- ★ 0
La documentación de esta integración la escribe su autor y, por ahora, solo está disponible en inglés.
The glucose of the FreeStyle Libre sensors shared with your LibreLinkUp account, in Gladys, through the cloud of Abbott, with the account of the LibreLinkUp app (the follower app: the patient shares their data with you from their own app).
Not a medical device: never use it for a treatment decision; the alerts may be late or missing.
Gladys reads the cloud at the refresh you choose, through the internet, the phone of the patient and the servers of Abbott: any of them can be late, down or wrong, and Gladys says nothing when it cannot read. Keep the alarms of the FreeStyle Libre / LibreLinkUp apps and your care team's advice; this integration only adds home automation on top (a lamp that turns orange, a notification), never a medical alarm.
- each patient shared with the account becomes a Gladys device: glucose (mg/dL or mmol/L), trend, measurement age (minutes) and the low and high targets of the account (when the API gives them);
- three scene triggers: Glucose low, Glucose high and Measurement too old.
Read only: Gladys never changes anything at LibreLinkUp (alarms, sharing, targets stay in the apps).
This integration has never been tested with a real LibreLinkUp account. This first version was written from the pylibrelinkup library (MIT), ported to JavaScript, and tested against a fake LibreLinkUp API only. Tell the Gladys forum what works and what does not.
Unofficial integration, not affiliated with Abbott.
Configuration
- E-mail address and Password: those of your LibreLinkUp account (the follower app). The password is kept by Gladys in its database, never shown again, never written to a log, and sent only to the LibreLinkUp API of Abbott (over https, certificate checked). The e-mail address is never written to a log either. Abbott may lock an account after a few wrong passwords: see "Connection problems".
- Refresh: every minute to every 15 minutes (2 minutes by default). One request per reading (a login now and then, when the token of the session expires). The sensor sends a measurement about every minute; Abbott may slow down an account that reads too often.
- Glucose unit: mg/dL (by default) or mmol/L: the unit of the glucose
and target features that Discovery offers, and of the glucose handed to
the scene triggers. The core of Gladys has no glucose unit: the unit is
written in the name of the feature (
Glucose (mg/dL)) and in its key. Changing it adds features in the new unit (update the device from Discovery); every feature a device has keeps receiving its values, each in its own unit (you may delete the ones you no longer want). mmol/L values are converted from mg/dL (÷ 18.0156, one decimal): they may differ by 0.1 from what the app shows.
Then open the Discovery tab and create the patients to follow. Only the created patients receive their values and fire the triggers.
The region of the account (Europe, United States...) is found by the API
itself at the first login: Gladys follows it only to the official
LibreLinkUp hosts it knows (api-eu.libreview.io, api-us.libreview.io,
etc.), never to a host given by an answer.
What each feature shows
- Glucose (mg/dL or mmol/L): the latest measurement of the sensor.
- Trend (0 to 5): 1 falling fast, 2 falling, 3 stable, 4 rising, 5 rising fast (the arrow of the app); 0 when the API gives no trend with the measurement.
- Measurement age (minutes): how old the latest measurement is, from the clock of the sensor, at each reading. When an answer of LibreLinkUp has no measurement for a patient, the age keeps counting from the last measurement received (glucose and trend keep their last value): "Measurement too old" can then fire, never "low" nor "high".
- A measurement dated in the future (more than 5 minutes) adds a warning to the status of the integration: check the clock of Gladys.
- Target low / Target high: the target range of the account (the API's
targetLow/targetHigh, read as mg/dL: to be confirmed on a real account). The features appear in Discovery only once the API gave them for this patient (then update the created device to add them). - A value the API does not give is not published. Gladys accepts 300 states per minute from an integration: this one sends 250 at most.
- No history by default for every feature (health data): Gladys keeps only the last value. You may turn the history on per feature, knowing that Gladys then keeps these health data in its database and backups.
Scene triggers
Glucose low and Glucose high: choose the patient (or any) and the threshold (required): the target of the account, or 54 / 70 / 80 mg/dL (low), 180 / 250 / 300 mg/dL (high). The scene starts when a fresh measurement (15 minutes old at most) is below (above) it, at the next reading. A measurement counts as fresh only if the integration first saw it 15 minutes ago at most. The "target" threshold never fires for a patient whose account gives no target. Once per crossing: again only after a measurement 5 mg/dL back above (below) the threshold, or after at least one reading missed: when the patient was not read, or two of its measurements are apart, for longer than max(15 minutes, refresh + 2 minutes). A low or high already there when the integration starts is told at the first reading (a missing alert is worse than a repeated one): a restart of the integration may tell it once more. Never for an older measurement, nor at a reading where the clock of Gladys jumped. Variables: name of the patient, glucose, unit, trend, time of the measurement (UTC).
Measurement too old: choose the patient (or any) and the delay (required: 15 minutes to 4 hours). The scene starts when no new measurement came for this delay (sensor ended or removed, phone of the patient without network or Bluetooth...). Once per wait: a new measurement arms it again. Never for a delay already reached when the integration starts, nor at a reading after a clock jump. Nothing is told when LibreLinkUp itself cannot be read (no internet, Abbott down, account stopped): watch the connection status of the integration.
30 events at most per reading; the rest goes at the next reading.
Connection problems
- Credentials refused: a wrong password tried again and again locks
the account. Gladys therefore stops for 1 hour, then 6 hours,
then 24 hours after each new refusal, never in a loop. The stop is
kept in
/data: a restart, a reconnection of Gladys and the test button wait too. An attempt still in flight when the integration stops counts as a refusal. Saving the configuration with the same credentials makes one attempt at once per series of refusals; other credentials are tried at once. A reading that works forgets the refusals. - New terms of use, privacy policy or e-mail to verify: when the API asks for one of them, Gladys never accepts it for you. It stops with a message: open the LibreLinkUp app with this account, accept there, then save the configuration of the integration (or wait for the next attempt: 1 hour, then 6 hours, then 24 hours).
- LibreLinkUp overloaded (HTTP 429, or a login locked for too many
attempts) or server error (HTTP 5xx): every request is paused, for
the time LibreLinkUp asks between 1 minute and 6 hours, never less than
the refresh. The pause is kept in
/data, even across a restart, a reconnection of Gladys or a saved configuration. - App version refused: the API refuses old versions of its app. The integration sends a fixed version (see TODO.md); if LibreLinkUp refuses it, the status says so: update the integration.
- Unexpected answer, LibreLinkUp unreachable (network, 30 s timeout,
certificate, region outside the list): retried after the refresh, then
the delay doubles up to 1 hour, even after a restart (the wait is kept in
/data). The integration never gives up. - Test the connection reads again and says how many patients answer; it never shortens a stop nor a pause.
Privacy
Glucose values, trends, dates of measurements, the names and ids of the
patients, the e-mail address, the password and the token of the session are
never written to a log nor to the /data volume of the integration. /data
only keeps a short digest of the credentials (4 characters, used only to
see that they changed) with the end of a stop, of a pause or of a wait. The
token stays in memory. Gladys keeps the last value of each feature in its
database (and its backups), and the history only if you turn it on. The
scene triggers hand the glucose to your scenes (to write a notification):
it is then where your scenes send it. The external id of a device is a
non-secret digest of the LibreLinkUp id of the patient (never the id
itself). The name of a device is the name of the patient as LibreLinkUp
gives it: rename it in Gladys if you prefer (the scene triggers use the
name of the device in Gladys). The name and the glucose go wherever the
names and values of devices go: scenes, notifications, Gladys Plus, and
the assistants and AI connected to Gladys (Google Home, Alexa, MCP...).
Limits
- One account per integration, 20 patients at most (the patients created in Gladys first).
- Nothing is commanded: neither the alarms of the app, nor the sharing.
- No history of the sensor (the graph of the last 12 hours) nor logbook read: only the latest measurement.
- Never tested with a real account: see TODO.md in the repository.
Parámetros de configuración
Estos son los parámetros que LibreLinkUp te pide en su pantalla de configuración en Gladys.
| Parámetro | Tipo | Obligatorio | Descripción |
|---|---|---|---|
| How it works | section | No | NOT A MEDICAL DEVICE: never use it for a treatment decision; its values and alerts may be late or missing. Gladys reads, through the cloud of Abbott, the patients shared with your LibreLinkUp account (the follower app of FreeStyle Libre). Each patient becomes a Gladys device (Discovery tab): glucose, trend, age of the measurement and the targets of the account. Read only. Health data: no history by default, never in the logs. Never tested with a real account. Unofficial integration, not affiliated with Abbott. |
| E-mail address | string | Sí | The e-mail address of your LibreLinkUp account (the follower app, not the app of the patient). Never logged. |
| Password | secret | Sí | The password of your LibreLinkUp account. Kept by Gladys in its database, never shown again, never logged, sent only to the LibreLinkUp API of Abbott. |
| Refresh | select | Sí | How often Gladys reads LibreLinkUp (one request). The sensor sends a measurement about every minute; Abbott may slow down an account that reads too often. |
| Glucose unit | select | Sí | The unit of the glucose and target features that Discovery offers, and of the glucose of the scene triggers. Changing it adds features in the new unit (update the device from Discovery); every feature a device has keeps receiving its values, each in its own unit. |
Cómo instalar LibreLinkUp en Gladys
- En Gladys, abre Integraciones: LibreLinkUp aparece en el catálogo junto a las integraciones nativas, con una insignia de comunidad.
- Haz clic en Instalar. Gladys descarga la imagen Docker (
ghcr.io/lm1lc3n7/gladys-librelinkup:0.1.1), la ejecuta en un entorno aislado del núcleo y genera la interfaz de la integración (dispositivos, descubrimiento y configuración). - Abre la pantalla de Configuración de la integración, rellena los parámetros y guarda.
- También puedes instalarla directamente desde la URL de su repositorio: https://github.com/LM1LC3N7/gladys-librelinkup.
LibreLinkUp requiere Gladys >=5.1.0. El catálogo de Gladys se actualiza cada hora, así que una nueva versión está disponible como máximo una hora después de publicarse.
¿Todavía no usas Gladys? Es gratuito y de código abierto: sigue la guía de instalación para empezar.
Sobre las integraciones externas
LibreLinkUp es una integración externa: una integración comunitaria empaquetada como contenedor Docker y publicada en GitHub. Gladys la instala con un clic y la ejecuta en un entorno aislado del núcleo. La publica y mantiene LM1LC3N7, no el equipo principal de Gladys.
- Explorar todas las integraciones, nativas y de la comunidad
- Crear y publicar tu propia integración externa
- Código fuente en GitHub — fuente de esta documentación