41 — Libre 2 glucose to LibreView, read through Juggluco
Esta página aún no está disponible en tu idioma.
2026-11-15 · investigation report · follows 04 — LibreView ·
spike investigations/41-libreview-libre2
The owner’s LibreView had no glucose after 2 Nov 18:59. Their Libre 3 Plus ended then, and since 3 Nov they’ve worn a Libre 2 read through Juggluco. Sukrara uploaded only treatments (docs/04: glucose went up only from a Libre 3 Sukrara reads itself). The owner: “it should not happen to begin with”, “Libre 2 glucose should go up”. Questions:
- How do other apps get a Libre 2 into LibreView, and can Sukrara do it as the one uploader? docs/04 found that two uploaders fight over the account’s active device.
- Which other sensors can go up?
- How much of the gap since 2 Nov can be filled?
Result
Section titled “Result”| Question | Answer |
|---|---|
| How Juggluco does it | A second identity: the LibreLink app (FSLibreLink.Android, config FSLibreLink_Android_2.10_<country>_config.json, header Abbott-ADC-App-Platform: Android/<rel>/FSLL/2.10.1.10406, no API key), its own DeviceId, the sensor’s 15-minute history (averaged stream) + scans, sensorstart with productType 3, wearDuration 20160, PUT nisperson with the whole serial. When LibreView answers wrongDeviceForUser, it takes over by itself (SetDevice:true). That’s the fight. (Libreview.java postgetauth, libreview.cpp, putsensor.cpp) |
| Sukrara’s way | One device, one identity for every Libre: the Libre 3 app identity Sukrara already uses. A Libre 2’s readings go up as 5-minute history under its own serial and life count. Nothing else uploads, so nothing fights. Proved on the real account: LibreView registered the Libre 2 and stored its 13 values exactly (below). |
| Other sensors | Libre 1 / 2 / 2 Plus through Juggluco: yes. Dexcom, SmartGuide, AiDEX, Sibionics: no. They would show in the clinic’s LibreView as Abbott data, and Juggluco’s author refuses them for that reason (Juggluco #250: “I don’t want to create a mess in Libreview”). A Libre 3 through Juggluco: not recognisable from Juggluco’s serial, so not sent. |
| The gap | 2 Nov 18:59 → 3 Nov 14:39: nothing to fill. No sensor gave readings: the Libre 3 Plus had ended, and the Libre 2 started at 13:38 with a one-hour warm-up. 3 Nov 14:39 → now: all of it (Juggluco holds 16 952 minute readings, 3 438 five-minute values; 14 gaps of 6–30 min are lost in Juggluco too). Sukrara sends it once the person taps Send earlier glucose. |
What other apps and people do
Section titled “What other apps and people do”- Juggluco (GPL; read 2026-11-15 at
afe6539) keeps separate tokens and DeviceIds per identity (getlibreDeviceID(libre3)). It sends Libre 2 data as LibreLink, “the last 89 days” on a first upload, and offers Resend data from a date. Its help says: “Don’t send data covering the same period to the same LibreView account from more than one source” (help, FAQ). It never sends non-Abbott sensors (#250 above). The DeviceId “identifies an app installation/uploader”, and LibreView lists each one as a device that can be deleted with all its data (#148). - The real LibreLink app’s upload, captured in xDrip+ discussion #2118:
scheduled glucose every 15 min, with record numbers 15 apart (serial base | life count), and a
PUT nispersonregistration (“a new sensor is introduced to the system: acknowledged by the server”). xDrip+’s maintainer declined to upload to LibreView at all. - Missing data: Juggluco users see “Success” while LibreView lacks data (#271, #14: it waits 20 min between uploads).
- One device per account: LibreLink users report that signing in on another phone signs the first one out (diabetes.co.uk forum). Whether LibreView tracks that per app identity or per account isn’t documented anywhere (step 2 of the proof asks). Per app identity: 45.
- Abbott’s apps: one LibreView login for LibreLink, the Libre 3 app and the new Libre app (Abbott). The Libre app replaces the Libre 2 and Libre 3 apps (App Store), so one app for both families is where Abbott itself is going.
Design (offline, proved by tests)
Section titled “Design (offline, proved by tests)”- Where the readings come from. Juggluco’s web server names every reading
SERIAL#lifeCount(_idinentries.json,recordIdin the live stream; the life count is the sensor’s own minute: it ran 86 min behind the phone’s clock over 12 days, so the start time can’t stand in for it).JugglucoSourcepasses these on asJugglucoEntry;SourceManageremits them as aSensorGlucoseBatchwitheveryMinute. - What goes up. A Libre 1 / 2 / 2 Plus serial (11 characters,
0M…/3M…: the NFC uid’s fixed top bytes; Sibionics, Dexcom never match). One value per 5 minutes of the sensor’s life: the first reading in each slot, under the slot’s first minute. Record number = the serial’s base (last 9 characters, docs/04) | that life count. The same readings in any order give the same records, so re-sending is harmless.sensorstartas LibreLink sends a Libre 2 (productType 3,wearDuration 20160). The registration uses the whole serial. - Fixture:
lv2.py fixture→contracts/treatments/libreview-glucose-fixture.json(synthetic minute stream with holes, a repeated minute, readings in slot 0) →LibreViewPayloadTestreproduces the values and the bytes. - The gap. Juggluco’s first history after the app starts is now the current sensor’s whole life (
historyPath,find[date][$gte]=start; before: the last 1 600 minutes), and the upload holds 15 days instead of 3. The upload still starts at the switch-on by itself (another app may have sent the earlier time). Send earlier glucose sends the rest when the person asks.
Proof on the real account
Section titled “Proof on the real account”2026-11-15, each step approved by the owner, run one at a time from the main session with the owner’s direct OK,
stopping at the first unexpected answer. Steps and expected answers: (1) Libre 3 identity, spike DeviceId, no take-over →
wrongDeviceForUser; (2) LibreLink identity, no take-over → status 0 means the active device is tracked per app;
(3) take over → status 0; (4) register 3MH01GFXNU8 → status 0; (5) 3 Nov 15:00–16:00, 13 five-minute values
under device Sukrara-spike41; (6) LibreView’s CSV. Cleanup: remove the spike device(s) in LibreView › My devices;
Sukrara › LibreView › Take over; delete ~/.config/sukrara-spike41.
| Time (local) | Step | Answer |
|---|---|---|
| ≈ 14:55 | 1. Libre 3 identity, spike device, no take-over | status 20 wrongDeviceForUser, as expected: Sukrara is the account’s uploading device |
| 2. LibreLink identity, spike device, no take-over | status 20 wrongDeviceForUser: the uploading device is (most likely) one per account, not per app. A stale LibreLink device would answer the same, so it isn’t conclusive, but either way a second identity next to Sukrara’s would have to take over. 45 settled it: per app identity. The LibreLink identity had its own stale device, and taking it over leaves the Libre 3 identity’s device and session alone |
|
| 3. Libre 3 identity, take over | status 0 |
|
4. Register 3MH01GFXNU8 (a Libre 2, whole serial) under the Libre 3 identity |
status 0 |
|
| 14:56 | 5. 13 five-minute values, 3 Nov 15:00–16:00 (207 → 150 mg/dL) + sensorstart |
status 0, UploadId, SerialNumber = the spike DeviceId; every MeasurementCounts and ItemCount 0. That’s the answer the real LibreLink app gets too (xDrip+ #2118), so it says nothing: ingestion is asynchronous |
CSV check (the owner’s Download glucose data, 15:10): passed. 14 rows of the spike device (Device “FreeStyle
Libre 3”, serial = the spike DeviceId). There are 13 Historic Glucose rows on 3 Nov, 15:00 → 15:59, the values identical to
Juggluco’s (207, 205, 196, 195, 186, 185, 177, 168, 162, 157, 160, 151, 150). So LibreView stores a Libre 2’s glucose sent
as the Libre 3 app, under the Libre 2’s serial. The 15:04 → 15:11 step is right: Juggluco itself has no reading
between 15:08:39 and 15:11:40 (minutes 90–91), so slot 90’s first reading is 15:11. The 14th row, the sensorstart
note, was filed on 15 Nov 14:56, the upload time: Sukrara’s (and the Libre 3 app’s) timestamp was “now”. Fixed:
both times are now the sensor’s start, so a start sent days later (mid-sensor, a backfill) is still filed on the day the
sensor started. Sukrara sends it once per sensor (LvSensorState.startSent).
Cleanup: the owner removes the spike device in LibreView › My devices (its 14 rows go with it) and taps Take over in
Sukrara; ~/.config/sukrara-spike41 and the copy of Juggluco’s readings are deleted.
In Sukrara
Section titled “In Sukrara”| Added | Where |
|---|---|
| Juggluco’s readings with their sensor and life count; the first history is the current sensor’s whole life | :sensor JugglucoSource (parseHistory, parseLive, historyPath), SourceManager; JugglucoSourceTest |
| A Libre 1 / 2 read through Juggluco goes up as 5-minute values of its own serial and life count, as the one device; other families never | :integrations LibreSerial, LibreViewQueue.slots, LibreViewUploader.buffer, LvSensor.productType; LibreViewPayloadTest (contract fixture), LibreViewUploaderTest |
| Send earlier glucose: the current sensor’s glucose from before the switch-on, once asked (warns that another app’s upload would count twice) | LibreViewUploader.sendEarlier, LibreViewPage; LibreViewUploaderTest |
| The sensor’s whole life held (15 days, was 3) | LibreViewQueue.HISTORY_MIN |
| The stalled banner without a clinic: “Nothing since …” | :feature:home strings |
sensorstart filed at the sensor’s start, not the upload (both families) |
LibreViewPayload, contracts/treatments/SPEC.md §3, the fixture |
| Discarded | Why |
|---|---|
| A second identity (LibreLink) for the Libre 2, as Juggluco does | two tokens, two devices in LibreView, two registrations, and if LibreView tracks the active device per account, the take-over fight inside one app; Abbott’s own Libre app is one app for both families. (45: it is per identity, so no fight; the other reasons stand) |
| Turning on Juggluco’s Send to Libreview next to Sukrara’s treatments | Juggluco takes over by itself on wrongDeviceForUser; two uploaders on one account is what docs/04 found fighting |
| Dexcom, SmartGuide, AiDEX, Sibionics glucose | the clinic would read it as Abbott data; Juggluco’s author refuses the same (#250) |
| A Libre 3 read through Juggluco | its serial in Juggluco’s _id can’t be told apart from other families with certainty; a Libre 3 is read by Sukrara itself |
| Life counts from the sensor’s start time | the Libre 2’s own minute drifted 86 min from the clock in 12 days; Juggluco’s counter is what LibreLink uses |
| 15-minute history (LibreLink’s) | the Libre 3 identity’s history is 5-minute; LibreView builds its reports from it either way |
| Backfilling by itself from the sensor’s start | another app may have sent that time: LibreView would count it twice; only the person knows |
| A date and time choice for Send earlier glucose | the phone holds only the current sensor, so the choice is “from when Sukrara has it” or nothing; a picker adds a step without a case that needs it yet |
| Naming Sukrara’s device in LibreView (the website’s Device nickname) | the website sets it with POST user/updateNickname {nickname, deviceId} on its own API (api[-region].libreview.io, Bearer token from the website’s auth/login, header Account-Id = SHA-256 of the user id), where deviceId is LibreView’s own id from the devices list (glucoseHistory data.devices: Sukrara’s sn 006aa5d7… is id 01a1cade…), not the upload’s DeviceId. Neither the upload nor the app login has a nickname field. Tried with the LibreLinkUp login Sukrara already has: without Account-Id → 400 RequiredHeaderMissing, with it → 400 ProductMismatch. Only the website’s own login would do: a second sign-in (with its extra steps) for a label, so not done. The person can type it on the website |
| Juggluco’s calibrated values | Sukrara takes what Juggluco serves; Juggluco’s own LibreView upload sends uncalibrated values. Not handled: the owner doesn’t calibrate |