10 Commits
Author SHA1 Message Date
willemjan e08f269855 test jenkins
willemjan/glucose-dashboard/pipeline/head Something is wrong with the build of this commit
2026-08-10 23:19:27 +02:00
willemjan b8e01fa8c9 Show estimated HbA1c in mmol/mol (IFCC) alongside the % value
Dutch lab results conventionally report HbA1c in mmol/mol rather than
DCCT/NGSP %. Adds gmi_mmol_per_mol() using the standard IFCC conversion
(Hoelzel et al. 2004) and shows it next to the existing % value on the
Samenvatting page and in the PDF summary.

Closes gitea #25.
2026-08-01 23:12:43 +02:00
willemjan 374ead70e5 Replace monthly GMI bar chart with daily scatter + rolling avg + linear trend/CI
Adds daily_gmi() and gmi_trend(): per-day GMI as scatter points, a 7-day
rolling average line, and a linear regression trend line with a 95%
confidence band (normal approximation, no scipy dependency, consistent
with the existing preference for a lean container image).

Recomputed from the already-fetched period data on each page load; the
existing 1-hour Streamlit cache on load_range_data() already avoids
repeated Nightscout calls for the same period, so no separate storage
layer is needed at the current max. 90-day report window.

Closes gitea #19.
2026-08-01 22:57:54 +02:00
willemjan 9e11dd8d73 Fix bolus/basal insulin categorization: split SMB from manual corrections, add basal calc
AndroidAPS records an automatic SMB micro-bolus as eventType "Correction
Bolus" with type "SMB", not as its own eventType. bolus_stats() was
grouping by eventType alone, so every SMB was silently counted as a manual
correction bolus. Verified against production data: 100% of "Correction
Bolus" treatments over 90 days were actually type=SMB.

Also adds basal_stats(), computing delivered basal directly from Temp
Basal treatments (rate x duration) without needing the pump profile.

Closes gitea #20.
2026-08-01 22:52:54 +02:00
willemjan 441bb84212 Add TIR bar/threshold lines, Dagoverzicht charts, and hypo/hyper Level-1 charts to PDF export
Closes gitea #14 (PDF part), #15 (PDF part), #16, #17, #18.
2026-08-01 22:33:45 +02:00
willemjan 00fb81b283 Add clinical threshold lines to AGP/Dagoverzicht charts, TIR stacked bar to Samenvatting
Closes gitea #14, #15.
2026-08-01 22:26:26 +02:00
willemjan 699848b479 fix: treatments endpoint needs a created_at cursor, not date
Investigated the "only ~4 days of treatments" report via direct
MongoDB access (issue #13): the collection has 123k+ documents, no
capping, no TTL index — the full year of history is in the database.

The actual cause: /api/v1/treatments.json filters/sorts on the string
field `created_at`, not the numeric `date` field /api/v1/entries uses.
Paginating with find[date][$lte] (or no filter at all) silently caps
treatments to an internal ~4-day window even though every document has
a valid `date` field; find[created_at][$lte] correctly walks back
through the full history.

_paginated_get() now takes an explicit cursor_field per endpoint and
always sends the cursor from the first request (never omitted), so
that hidden default window can't trigger. Verified against the real
instance: a 92-day treatments fetch now returns 39,403 rows (was ~700).
2026-08-01 21:43:08 +02:00
willemjan 55883480c0 docs: record FR1 findings and add profile-tuning design spec
requirements-profile-tuning.md: document the LXC 102/107 investigation
outcome (LXC 102's houthacker/nightscout-autotune image already solves
both the analysis and NS-upload directions) and resolve the two open
questions this depended on.

design-profile-tuning.md: full design spec for the separate
glucose-profile-tuning app/service, reviewed and approved.
2026-07-31 14:21:46 +02:00
willemjan 29d20ca441 fix: Nightscout API silently truncated multi-day fetches to ~3-4 days
_entries_since/_treatments_since requested a single page via `count`,
but Nightscout's /api/v1/entries and /api/v1/treatments ignore a large
`count` and cap the response around ~1150/~700 documents regardless of
how far back the requested period goes. Every report page relying on
fetch_entries_range/fetch_treatments_range (7/14/30/90-day selectors)
was silently only getting the most recent ~3-4 days, no matter the
selected period.

Fixed by paginating with `find[date][$lte]` cursors in a new
_paginated_get(), walking back until the requested `since` boundary is
reached or the server returns a short/empty page (= genuinely no more
data, confirmed by manually paginating against the real instance:
entries go back 50+ days, treatments only ~4 days on this install).

Also added reporting.completeness_warning(), shared by the interactive
pages (via ui_common.render_completeness_warning) and the PDF export,
that distinguishes "a structural gap at the start of the period" (data
wasn't fetched/available) from "scattered gaps below the clinical AGP
coverage threshold" (Danne et al. 2017) — the previous single generic
"X% sensor coverage" message read as a CGM/sensor problem even when
the real cause was this fetch bug.
2026-07-31 14:21:35 +02:00
willemjanandClaude Sonnet 5 fe6a471f17 add PDF export (FR10)
pdf_export.py builds a PDF report (summary stats, AGP chart, hypo/hyper
event summaries) via fpdf2 + matplotlib rather than plotly's kaleido, to
avoid bundling a Chromium build into the image just for this. The Export
page lets the user pick which sections to include and a date range, then
download the result to send to the hospital team.

This completes all ten functional requirements in
docs/requirements-reporting.md.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
2026-07-26 21:10:46 +02:00
19 changed files with 1823 additions and 99 deletions
Vendored
+11
View File
@@ -0,0 +1,11 @@
pipeline {
agent any
stages {
stage('Do nothing') {
steps {
sh '/bin/true'
}
}
}
}
+328
View File
@@ -0,0 +1,328 @@
> **Moved.** Implementation now lives in the separate `glucose-profile-tuning`
> repo (per this doc's own §2 isolation decision); the canonical, actively
> maintained copy of this file is at `glucose-profile-tuning/docs/design-profile-tuning.md`.
> This copy is kept for history and is no longer updated.
# Design specification: oref0-autotune profile tuning
Status: **final — user reviewed**
Requirements: [requirements-profile-tuning.md](requirements-profile-tuning.md) (FR1–FR9, NFR1–NFR5)
Related: [design-reporting.md](design-reporting.md) (same conventions, separate app)
This is a design spec, not code. Nothing here has been implemented yet —
implementation starts after this doc is reviewed/approved, one phase at a
time, each shipped through its own CI/CD pipeline (§8).
## 1. FR1 recap — infra decision (already executed)
FR1 (boot + inspect LXC 102/107) is done; full findings are recorded directly
in [requirements-profile-tuning.md §4 FR1](requirements-profile-tuning.md).
Summary: **LXC 102 (`autotune`, 192.168.2.13)** already has a working, pinned
Docker image — `ghcr.io/houthacker/nightscout-autotune:latest` — that bundles
the *entire* oref0 npm toolkit, not just the autotune algorithm:
| Tool (already installed in the image) | Role in this design |
|---|---|
| `nightscout-autotune` (image's default CMD) | Fetches NS entries/treatments/profile, runs oref0-autotune, writes results to `$OPENAPS_WORKDIR` |
| `oref0-upload-profile profile.json NSURL secret [--preview] [--switch]` | Converts an oref0-format profile back to Nightscout's profile-store format **and** uploads it, optionally issuing a Profile Switch treatment |
| `convert-ns-profile` / `mm-format-ns-profile` | The NS-format ⇄ oref0-format conversion primitives, used internally by the two tools above |
This changes the shape of the whole feature for the better: **we do not need
to write any Nightscout-profile-format conversion code ourselves.** Both
directions (NS → oref0 for analysis, oref0 → NS for staging) are already
solved, tested, and previously proven against the user's real Nightscout
instance (bash history on LXC 102 shows a full run against
`https://zb-ns.duckdns.org` completing 9 months ago). This is exactly the
"had to manually parse parts of the profile" pain point called out in FR2 —
resolved by wrapping existing tools instead of reimplementing them.
LXC 107 is not used; it's a messier, native (non-Docker) install of the same
underlying toolkit with no functional advantage over 102. It stays stopped.
## 2. Architecture overview
Two new pieces, one existing piece untouched:
```
┌─────────────────────────────┐ ┌──────────────────────────────────┐
│ LXC 120 (glucose-app) │ │ LXC 102 (autotune) │
│ │ │ │
│ ┌──────────────────────────┐ │ HTTPS │ ┌────────────────────────────────┐│
│ │ glucose-dashboard │ │ Bearer│ │ autotune-service (new) ││
│ │ (existing, untouched) │ │ token │ │ Node/Express, image FROM ││
│ └──────────────────────────┘ │◄───────┼─┤ ghcr.io/houthacker/ ││
│ │ │ │ nightscout-autotune ││
│ ┌──────────────────────────┐ │ │ │ - owns the NS write token ││
│ │ autotune-app (new) │─┼────────┼─▶ - owns the SQLite audit DB ││
│ │ Streamlit, own auth gate │ │ │ │ - runs analyses (spawn) ││
│ └──────────────────────────┘ │ │ │ - runs oref0-upload-profile ││
│ │ │ │ - nightly scheduler (04:00) ││
└──────────────────────────────┘ │ └────────────────────────────────┘│
│ │ │
└────────────┼─────────────────────┘
│ HTTPS (existing NS
│ read/write token)
▼
Nightscout (zb-ns.duckdns.org)
```
Key property driving this split: **the frontend never talks to Nightscout
directly, and never holds a Nightscout credential.** It only holds a shared
bearer token to call `autotune-service`. All Nightscout I/O — reading
entries/treatments/profile for analysis, and writing the staged profile /
Profile Switch — happens on LXC 102, inside `autotune-service`, using the one
write-scoped NS token (NFR2). This keeps the dosing-adjacent credential in
exactly one place, matches architecture decision #3 in the requirements doc
(isolating write-capable code from the read-only dashboard), and means the
Streamlit app can be redeployed or rebuilt at any time without touching
secrets or state — it's a pure, stateless UI.
Repo layout: one new repo, **`glucose-profile-tuning`**, with two
subdirectories and their own Dockerfiles:
- `service/` — the Node microservice (deployed to LXC 102)
- `app/` — the Streamlit frontend (deployed to LXC 120, alongside
`glucose-dashboard` as a second `docker-compose` service — that LXC has
~800MB RAM headroom today, plenty for a second, mostly-idle Streamlit UI)
One repo keeps this a single reviewable unit (matching how the user framed
FR1–FR9 as one feature); the two Dockerfiles/deploy targets keep the runtime
isolation NFR1 asks for.
## 3. `autotune-service` (LXC 102)
### 3.1 Why Node, not Python
The base image is Alpine + Node 24 with the entire oref0 toolkit already
installed as Node/shell scripts under `/usr/local/bin`. Adding Python +
FastAPI on top would mean two language runtimes doing the same job one
already does. Building the HTTP layer in Node (Express) lets it `spawn()`
`nightscout-autotune` and `oref0-upload-profile` as sibling processes in the
same container/filesystem — no Docker socket, no Docker-in-Docker. (Mounting
the host's `docker.sock` to let a container launch sibling containers was
considered and rejected — unnecessary privilege escalation for a homelab
service that already has direct in-image access to the tools it needs.)
### 3.2 Job model
A single SQLite table, `analysis_runs`:
| column | type | notes |
|---|---|---|
| `id` | TEXT (uuid) | PK |
| `trigger` | TEXT | `manual` \| `nightly` |
| `requested_at` / `started_at` / `finished_at` | TEXT (ISO 8601) | |
| `status` | TEXT | `pending` \| `running` \| `done` \| `failed` |
| `days`, `profile_name`, `autosens_min`, `autosens_max`, `insulin_type`, `min_5m_carb_impact`, `uam_as_basal` | job params as submitted | `profile_name` nullable — omitted means "NS's current default", per the tool's own documented behavior |
| `error_message` | TEXT, nullable | |
| `current_profile_json` / `proposed_profile_json` | TEXT (JSON), nullable until done | oref0 pump-profile shape (`basalprofile[]`, `isf`, `carbratio`, `dia`, `autosens_min/max`) — this is the *input* `profile.json` and *output* `newprofile.json` the tool already produces |
| `recommendations_log` | TEXT, nullable | verbatim `autotune_recommendations.log` — human-readable Pump/Autotune/Days-missing table, used for FR4's estimation detail |
| `staged_at` | TEXT, nullable | set once FR6 staging succeeds |
| `staged_profile_json` | TEXT (JSON), nullable | the final merged profile actually sent to NS |
| `switched` | INTEGER (bool), nullable | whether `--switch` was used |
| `staging_decisions_json` | TEXT (JSON), nullable | per-slot accept/reject/custom record, for FR7 audit trail |
Only one job may be `pending`/`running` at a time (LXC 102 is 1 core / 1GB —
not built for concurrency); a second trigger while one is in flight gets
`409 Conflict`.
Job execution: `POST /v1/analyses` inserts a `pending` row, immediately
returns `{id, status: "pending"}`, then asynchronously:
1. mark `running`, set `started_at`
2. `spawn("nightscout-autotune", [], {env: {...NS_HOST, AUTOTUNE_DAYS, UAM_AS_BASAL, NS_TOKEN (sha1'd), NS_PROFILE, AUTOSENS_MIN/MAX, INSULIN_TYPE, MIN_5MIN_CARBIMPACT, OPENAPS_WORKDIR: "/data/jobs/<id>"}})`
3. on exit code 0: read `profile.json`, `newprofile.*.json`, and
`autotune_recommendations.log` out of the job's workdir, populate the row,
mark `done`
4. on non-zero exit: capture stderr/stdout tail as `error_message`, mark
`failed`
### 3.3 HTTP API
All endpoints (except `/healthz`) require `Authorization: Bearer
<AUTOTUNE_SERVICE_API_KEY>`.
- `POST /v1/analyses` — body `{days, profile_name?, autosens_min?, autosens_max?, insulin_type?, min_5m_carb_impact?, uam_as_basal, trigger: "manual"}` → `202 {id, status}` or `409` if one's already running.
- `GET /v1/analyses/{id}` — full row as JSON (shape above).
- `GET /v1/analyses?trigger=&limit=` — history list, newest first (backs FR7 and FR9's nightly-runs page).
- `POST /v1/analyses/{id}/stage` — body `{profile: <merged oref0-shape JSON>, switch: bool, decisions: [...]}` → runs `oref0-upload-profile <tmpfile> $NS_HOST <secret> [--switch]`, updates the row, returns `200` or `502` with the CLI's stderr on failure. **To verify during implementation:** whether `oref0-upload-profile`'s positional `api-secret` argument accepts a token interchangeably (the analyze side supports both `NS_API_SECRET` and `NS_TOKEN`; the upload side's `--help` only documents `api-secret` positionally) — confirm against the installed script's source before wiring this up, rather than assuming.
- `GET /v1/profiles` — thin passthrough listing NS profile-store names, so the frontend can offer a dropdown instead of a free-text field for FR3's "select a current profile."
- `GET /healthz` — liveness, no auth required (used by the frontend's connectivity check and Docker's own healthcheck).
### 3.4 Nightly scheduler (FR9)
`node-cron` inside the same process, firing daily at 04:00 (container
timezone set to `Europe/Amsterdam`). Calls the same code path as
`POST /v1/analyses` with `trigger: "nightly"`, `profile_name` omitted (always
tunes from whatever is currently active — correct behavior for an unattended
run), and `days` from `NIGHTLY_AUTOTUNE_DAYS` (env var, default **7** — a
reasonable middle ground between enough data for stable estimates and not
diluting recent pattern changes; the doc's own wording — "whatever gives
usually the most accurate results" — doesn't have one universally-correct
answer, so this is exposed as a tunable env var rather than a hardcoded
guess, and the user can adjust it after watching a few nights of real
output). No result is ever auto-staged; a nightly run populates a `done`/
`failed` row exactly like a manual one, staging is always a separate FR6
action.
### 3.5 Logging (NFR "production grade")
Structured JSON lines to stdout via `pino`, one line per job lifecycle event
(`job.started`, `job.finished`, `job.failed` with `id`/`duration_ms`), `WARN`
for FR8 large-jump flags, `ERROR` for NS/upload failures — **never** logging
the raw token/secret. Captured by Docker's `json-file` driver with
`max-size`/`max-file` set in `docker-compose.yml` (no external log
aggregator exists in the homelab today per inspection of the `monitoring`
LXC, so this is deliberately self-contained rather than assuming a
Loki/Elastic-style sink that isn't there).
## 4. `autotune-app` (LXC 120, Streamlit)
### 4.1 Framework decision (NFR5)
**Streamlit**, not Shiny for Python. Reasons: every functional requirement
here (form inputs, polling with a spinner, a comparison table, an editable
grid, history lists) is already something `glucose-dashboard` does or
Streamlit does natively — `st.data_editor` with
`column_config.NumberColumn(step=0.01)` directly implements FR5's "editable
numeric fields, 0.01 steps" requirement without extra code. Shiny's reactive
model would add a second UI paradigm for the user to maintain with no
functional gain here. The password-gate (`auth.py`), theming, and
`AppTest`-based testing conventions from `glucose-dashboard` carry over
directly (copied into this new repo, not imported cross-repo, per NFR1's
isolation).
### 4.2 Pages
1. **`app.py`** — landing page, `require_auth()`, reads
`AUTOTUNE_SERVICE_URL` + `AUTOTUNE_SERVICE_API_KEY` from env, shows a
connectivity check against `GET /healthz`.
2. **`pages/1_Nieuwe_Analyse.py`** (FR3) — form: days, starting profile
(dropdown from `GET /v1/profiles`), an expander for advanced overrides
(autosens min/max, insulin type, min 5-min carb impact). Submits
`POST /v1/analyses`, then polls `GET /v1/analyses/{id}` (via
`st.spinner` + a rerun loop) until `done`/`failed`; a clear NS-connection
error surfaces the service's `error_message` verbatim (FR3 acceptance:
fail loud, don't hang).
3. **`pages/2_Overzicht_Resultaat.py`** (FR4 + FR5 + FR6 + FR8) — given a job
id: summary cards (ISF/CR pump vs. autotune, `autosens_min`/`max` bounds
shown explicitly per FR8), an `st.data_editor` grid — columns *time-slot*
(hourly), *huidig profiel*, *voorstel*, *bron* (select: current/suggested/
custom), *definitieve waarde* (editable number, step 0.01, disabled unless
*bron* = custom) — plus a raw `autotune_recommendations.log` expander for
the underlying estimation detail. Rows whose suggested value differs from
current by more than a configurable threshold (default: 20% relative or
0.1 U/hr absolute, whichever is smaller magnitude) are visually flagged
(FR8's "unusually large jump"). A "Stage to Nightscout" section requires
an explicit confirmation checkbox before `POST /v1/analyses/{id}/stage`
fires (FR8: no silent staging), with a separate "ook direct activeren
(Profile Switch)" checkbox controlling `switch`.
4. **`pages/3_Nachtelijke_Runs.py`** (FR9) — `GET /v1/analyses?trigger=nightly`,
latest first; failed rows show `error_message` inline; clicking a row
opens it in the same view as page 2.
5. **`pages/4_Geschiedenis.py`** (FR7) — full audit list (manual + nightly),
filterable by status/trigger, showing staged/switched state — this *is*
the audit trail; no separate mechanism needed since every run (and every
staging decision) is already a row in `autotune-service`'s database.
### 4.3 Merge logic (the one piece of real business logic on this side)
`merge_profile(current, proposed, decisions) -> final_profile` — a pure
function, unit-tested exactly like `reporting.py`'s style in
`glucose-dashboard`: for each slot, `decisions[slot]` is
`"current"`/`"suggested"`/`"custom"` (+ a value when `"custom"`); rejecting a
slot must leave `current`'s value byte-for-byte in `final_profile`
(FR5 acceptance criterion). This function, plus the large-jump-flagging
function, are the only two pieces of logic in this app worth unit tests
beyond `AppTest` page smoke tests.
## 5. Nightscout token (NFR2)
A new Subject/Access Token created in Nightscout's Admin Tools, distinct from
`glucose-dashboard`'s existing read-only token. Role: **`careportal`** — the
conventional Nightscout role that grants treatment and profile writes on top
of read access (vs. `readable`, read-only, or `admin`, full control this
doesn't need). **To confirm during implementation**: exact role-string
behavior can vary slightly by Nightscout version — verify a `careportal`-role
token can actually write `/api/v1/profile` on this specific instance before
relying on it, rather than assuming the classic-NS role semantics hold
exactly. Only `autotune-service`'s environment ever holds this token; it is
never sent to or stored in the browser/frontend.
## 6. Data requirements — resolved vs. the original doc
The original requirements doc's §6 assumed `nightscout_client.py`-style
`GET`/`POST /api/v1/profile.json` functions would need to be written. With
the LXC 102 findings, that's no longer true for the write direction
(`oref0-upload-profile` already does it) or the analysis-time read direction
(`nightscout-autotune` already does it). The frontend's only remaining NS-ish
need is the profile-name dropdown, served via `autotune-service`'s
`GET /v1/profiles` passthrough — so **no Nightscout client code is written in
Python at all** for this feature. All NS-format handling stays inside the
existing, already-proven Node/oref0 tooling.
## 7. Testing strategy (NFR3)
- **`service/`**: unit tests mock `child_process.spawn` and the filesystem —
never invoke the real `nightscout-autotune`/`oref0-upload-profile` binaries
or a real Nightscout in CI. Cover: job state transitions, the
pending/running 409 guard, SQLite persistence (temp-file DB per test),
route handlers, and the large-jump/merge-adjacent logic that lives here
(none currently — kept in the frontend, see §4.3).
- **`app/`**: `merge_profile`/large-jump pure-function unit tests; `AppTest`
page tests mocking `requests` against `autotune-service` (same
`FakeResponse`/`monkeypatch` pattern as `glucose-dashboard`'s suite);
`auth.py` tests copied from the existing suite.
- **UAT** (post-implementation, against the real LXC 102 service and the
user's real Nightscout): trigger a manual analysis → review the
comparison/flagged-jump view → reject a slot and confirm it stays at the
current value → accept + custom-edit others → stage without switch →
manually confirm in Nightscout the new profile appears correctly and the
previously active one is untouched → separately test staging *with*
switch → confirm a nightly run appears on the nightly page the next
morning, including a deliberately-forced-failure case showing a clear
reason.
## 8. CI/CD (NFR1, NFR4)
New Gitea repo `glucose-profile-tuning`, pipeline modeled on
`glucose-dashboard/.gitea/workflows/pipeline.yml` with two differences:
- **Two build targets**: `service/Dockerfile` (deploy → LXC 102) and
`app/Dockerfile` (deploy → LXC 120, second compose service alongside
`glucose-dashboard`), each independently Trivy-scanned and pushed to
`gitea.zb-kw.duckdns.org`.
- **`pull_request` trigger added to the test/lint job** (NFR4) — unlike
`glucose-dashboard`'s current `push`-only pipeline, this one can end in a
write to a real medical-data store, so tests must gate merge, not just run
after the fact.
New secrets/variables to provision (mirroring the existing dashboard's
setup): SSH deploy credentials for both target hosts, the new NS
`careportal` token, and a freshly generated `AUTOTUNE_SERVICE_API_KEY`
shared between the two new containers.
## 9. Implementation phasing
Same incremental, ship-behind-CI approach used for the reporting feature —
each phase is a separate PR, merged only once green:
1. **Scaffold** — both Dockerfiles building, empty pipeline green end-to-end
(test → Trivy → deploy to both LXCs), before any real feature — this
phase alone would have caught the hostname/secret/token issues the
reporting feature's deployment hit, ahead of time.
2. **Microservice core** (FR2) — SQLite schema, job runner wrapping
`nightscout-autotune`, `POST`/`GET /v1/analyses` only, unit-tested with
mocked `spawn`. No frontend yet — verified via `curl` against the real
LXC 102 (matches FR2's own acceptance criterion).
3. **Frontend read path** (FR3 + FR4) — pages 1 + 2 up to the comparison
view, no staging yet.
4. **Staging** (FR5 + FR6 + FR8) — `st.data_editor` merge grid, `/stage`
endpoint, confirmation gating, large-jump flags.
5. **Nightly + audit** (FR9 + FR7) — scheduler, nightly-runs page, history
page.
## 10. Open items carried into implementation (not blocking doc approval)
- Verify `oref0-upload-profile`'s secret-vs-token argument behavior (§3.3).
- Verify the `careportal` role actually permits profile writes on this NS
version before building against it (§5).
- Confirm the default `NIGHTLY_AUTOTUNE_DAYS=7` after watching real output
for a week or two; it's a tunable, not a fixed clinical constant (§3.4).
+95 -31
View File
@@ -1,7 +1,12 @@
> **Moved.** Implementation now lives in the separate `glucose-profile-tuning`
> repo; the canonical, actively maintained copy of this file is at
> `glucose-profile-tuning/docs/requirements-profile-tuning.md`. This copy is
> kept for history and is no longer updated.
# Requirements: oref0-autotune profile tuning
Status: **draft — for review and editing**
Owner: willemjan
Status: **final**
Owner: willemjan
Related: [requirements-reporting.md](requirements-reporting.md) (separate feature, same Nightscout data source)
## 1. Purpose & background
@@ -12,16 +17,16 @@ on-device history — it doesn't use Nightscout at all, and the user is still on
AAPS 3.3 regardless. Previously, the user ran a hosted tool (inspired by
`MarkMpn/AutotuneWeb`, now archived/discontinued) that analyzed Nightscout data
with `oref0-autotune` and proposed a new pump profile, which was then manually
entered into AAPS — a process the user describes as "a lot of trial and error."
entered into AAPS — a process the user describes as "a lot of trial and error." Before that, the user ran the tool from `MarkMpn/AutotuneWeb` but that was discontinued and cost a lot of effort to make changes to the profile by manually entering the changes into AAPS.
This feature builds a guided, human-in-the-loop workflow: trigger an
oref0-autotune analysis from a dashboard, review the proposed profile against
oref0-autotune analysis from an internally self hosted application, review the proposed profile against
the current one with the underlying estimations visible, selectively
accept/adjust values, and stage the result in Nightscout for manual activation.
The user has two LXC containers on Proxmox from earlier experimentation —
`autotune` (VMID 102, 192.168.2.13) and `oref0` (VMID 107, 192.168.2.14) —
currently stopped, of unknown/half-remembered working state.
currently stopped, of unknown/half-remembered working state. The working container can be used for the profile calculations, but if another solutions surpasses in terms of integratability a new solutions can be set up.
## 2. Scope
@@ -29,20 +34,17 @@ currently stopped, of unknown/half-remembered working state.
- Booting and inspecting the two existing LXCs to determine (or establish) a
working oref0-autotune setup.
- A small HTTP microservice on that LXC that runs oref0-autotune against the
user's Nightscout data on demand.
- A **separate** Streamlit app/service (own deployment, own CI/CD pipeline) that
user's Nightscout data on demand. Either one of the two existing LXCs or a new one.
- A **separate** app/service (own deployment, own CI/CD pipeline) that
triggers the analysis, displays old-vs-new profile comparisons with
estimation detail, lets the user selectively adjust proposed values, and
stages the result in Nightscout.
stages the result in Nightscout. The user can then choose if the profile is automatically activated in AndroidAPS (creating a
`"Profile Switch"` treatment)
- This app could be based on the glucose-dashboard streamlit app, but streamlit is not a fixed requirement, another solution can be used. For example, shiny for python.
- A nightly scheduled autotune run of which the results are visible in the morning. The profile of each run should be stored in a database. The user can read those back from the database in the application and can send it to nightscout/androidAPS, but the nightly scheduled runs should not send profiles automatically to nightscout and/or AAPS, this should always be a user action.
**Out of scope (this iteration):**
- Automatically activating the new profile in AndroidAPS (creating a
`"Profile Switch"` treatment) — the dashboard stages the profile only; the
user activates it manually. This may become an opt-in "phase 2" once trust is
established (see §8).
- Any change to AAPS settings or on-device configuration.
- Continuous/scheduled autotune runs — this is a manually triggered, on-demand
analysis, not a nightly automated loop.
## 3. Architecture decisions (made during spec review)
@@ -51,13 +53,12 @@ currently stopped, of unknown/half-remembered working state.
run or bundling autotune into the dashboard's own container. The dashboard
calls it over HTTP and polls for job completion.
2. **Profile push safety**: **stage-only**. The service writes a new named
profile into the Nightscout profile store but does **not** create a
`"Profile Switch"` treatment — final activation stays a deliberate manual
step in AAPS or the Nightscout UI. This matters because AndroidAPS applies
profile into the Nightscout profile store and does create a
`"Profile Switch"` treatment when the user selected that option — final activation stays a manual
step in AAPS when the option is not selected. This matters because AndroidAPS applies
profiles via Profile Switch treatments (gated behind an explicit "accept
profile switch from Nightscout" AAPS setting), not by simply reading
`profile.json` — so "stage-only" is a real, meaningful safety boundary, not
just a formality.
`profile.json`.
3. **Deployment**: a separate app/service from `glucose-dashboard`, with its own
repo-or-directory, Dockerfile, and Gitea CI/CD pipeline — isolating
write-capable, dosing-adjacent code from the read-only dashboard.
@@ -73,6 +74,60 @@ Document findings (what's there, what's missing, what needs rebuilding).
*Acceptance:* a clear decision on which LXC hosts the microservice, and what
(if anything) needs to be installed/fixed there before FR2 can start.
**Status: done (2026-07-26).** Both LXCs were booted and inspected, then
stopped again (back to their original state — no state was left running).
- **LXC 102 (`autotune`, 192.168.2.13, 1 core / 1GB RAM / 16GB disk)** —
Docker installed, with the image `ghcr.io/houthacker/nightscout-autotune:latest`
(Alpine + Node 24, ~279MB) already pulled and previously run successfully
(9 months ago) against the real Nightscout instance
(`https://zb-ns.duckdns.org`). It bundles the full `oref0` npm toolkit
(`oref0-autotune`, `oref0-upload-profile`, `convert-ns-profile`,
`mm-format-ns-profile`, and more) under `/usr/local/bin`, not just the
autotune algorithm itself. Its documented interface is a single container
run with env vars: `NS_HOST`, `AUTOTUNE_DAYS`, `UAM_AS_BASAL` (mandatory),
plus optional `NS_TOKEN`/`NS_API_SECRET` (as sha1 hashes), `NS_PROFILE`,
`AUTOSENS_MIN`/`MAX`, `MIN_5MIN_CARBIMPACT`, `INSULIN_TYPE`,
`OPENAPS_WORKDIR`, `HTML_EXPORT`. A prior run's output is still on disk at
`/root/autotune-output/autotune/`: per-day `ns-entries`/`ns-treatments`
fetched from Nightscout, the oref0-format `profile.json` (input) and
`newprofile.YYYY-MM-DD.json` (recommendation), and a human-readable
`autotune_recommendations.log` (a `Pump | Autotune | Days Missing` table per
parameter/basal-slot — ideal source for FR4's estimation detail). Critically,
the image also has `oref0-upload-profile` (`profile.json NSURL api-secret
[--preview] [--switch]`) already installed — this is the piece that converts
an oref0-format profile back into Nightscout's profile-store format *and*
optionally issues the Profile Switch treatment, i.e. it already implements
what FR6 needs. `--preview` (dry run, no upload) is also useful for FR8's
confirmation step.
- **LXC 107 (`oref0`, 192.168.2.14, same specs)** — no Docker; a native
npm/`openaps` install of the raw `oref0` toolkit. Bash history shows a
messier path to a working run (locale errors needing `locale-gen`, several
retries with different flag spellings) before landing on a working
`oref0-autotune --dir=~/myopenaps --ns-host=... --start-date=...` invocation
and a manual `oref0-upload-profile ... --switch=true` call. It reaches the
same destination but through an unreproducible, hand-patched native
environment rather than a pinned, documented container image.
**Decision: LXC 102 is the base.** It already has a working, versioned,
well-documented container with both the analysis direction (Nightscout →
oref0 profile → recommendation) and the upload direction
(oref0 recommendation → Nightscout profile store, with optional Profile
Switch) solved, via tools already installed in the same image. No new
autotune image needs to be built or evaluated — the open question in §8 about
choosing between `p5nbTgip0r/autotune-docker` /
`houthacker42/nightscout-autotune` / a from-scratch `oref0` build is moot:
`houthacker/nightscout-autotune` is what's already there, already proven
against this Nightscout instance, and turns out to be the most complete of
the three. LXC 107 will be left stopped/decommissioned — its only asset (a
successful manual run) is already superseded by 102's cleaner path. FR2's
microservice will be built as a thin wrapper *around* the CLI tools already
baked into the `nightscout-autotune` image (see
[design-profile-tuning.md](design-profile-tuning.md)), rather than
reimplementing Nightscout-profile ⇄ oref0-profile conversion from scratch —
directly resolving the "had to manually parse parts of the profile" pain
point named in FR2.
### FR2 — Autotune HTTP microservice
A small HTTP API (exact framework TBD during design — e.g. Flask/FastAPI
wrapping a Docker-based oref0-autotune image) exposing at minimum:
@@ -81,15 +136,17 @@ wrapping a Docker-based oref0-autotune image) exposing at minimum:
- `GET /analyze/{job_id}` — status (`pending`/`running`/`done`/`failed`) and,
once done, the recommended profile JSON plus the estimation/recommendation
detail (per-hour deviations, the autotune log).
- In previous tries, the user had to manually parse parts of the profile as obtained from Nightscout to have it accpeted by the autotune service. This should be thoroughly investigated and subsequently automated
*Acceptance:* a manual `curl` against the running service, pointed at the
user's real Nightscout instance, returns a plausible recommended profile within
a reasonable time (minutes, not seconds — oref0-autotune is not instant).
### FR3 — Trigger analysis from the dashboard
A new Streamlit app page where the user picks a date range (and confirms which
current profile to tune from, fetched live from Nightscout), starts the
A new app where the user picks a date range and selects a
current profile to tune from (fetched live from Nightscout), starts the
analysis, and sees progress while the microservice job runs (via polling).
Nightscout connection should be made using a token with a proper role.
*Acceptance:* starting an analysis with bad/unreachable Nightscout credentials
fails with a clear error rather than hanging silently.
@@ -109,6 +166,7 @@ Per time-slot controls (accept/reject each proposed basal/ISF/CR change
individually) plus editable numeric fields so the user can fine-tune any
accepted value before staging — matching the user's explicit request to
"select parts of the proposed profile and still adjust the values a bit."
This should look like a table with the columns 'time-slot' (per hour), 'current profile', 'suggestion', a input to select which column to select (current vs suggestion) and on the last column the final selected value, but as input numeric fields which can be changed by 0.01 steps.
*Acceptance:* rejecting a slot keeps the current value unchanged in the
staged profile; editing an accepted value's number is reflected in what
@@ -117,7 +175,7 @@ actually gets sent to Nightscout.
### FR6 — Stage to Nightscout
Write the resulting merged profile (accepted + manually adjusted values) as a
new named profile in the Nightscout profile store (e.g. `"Autotune
YYYY-MM-DD"`) via a new POST to `/api/v1/profile`, without creating a Profile
YYYY-MM-DD"` or `"Autotune YYYY-MM-DD ID"`) via a new POST to `/api/v1/profile`, with optionally creating a Profile
Switch treatment. After staging, show clear instructions for manual activation
in AAPS (or the Nightscout UI).
@@ -140,6 +198,11 @@ separate store — **decide during design**.
- Require an explicit confirmation step (e.g. a review summary + confirmation
checkbox) before the FR6 write happens — no silent/implicit staging.
### FR9 — Automated nightly runs
Each night at 4:00 the analysis is triggered with a fixed amount of days, whatever gives usually the most accurate results, and stores the result in a database. The app should have a page where the latest run is shown and should have the same downstream options as FR3 has (FR4, FR5 and FR6).
*Acceptance:* The analysis appears on the nightly analysis page and gets pulled from the database with autotune results. When the analysis failed a reason for failure must be shown.
## 5. Non-functional requirements
- **NFR1 — Isolation**: separate repository-or-directory, Dockerfile, and Gitea
@@ -156,10 +219,11 @@ separate store — **decide during design**.
existing `monkeypatch`-on-`requests` pattern from `glucose-dashboard`'s test
suite.
- **NFR4 — CI/CD gating**: given this pipeline can end in a write to a real
medical data store, consider (during design) adding a `pull_request` trigger
medical data store, add a `pull_request` trigger
for the test/lint job so changes are validated *before* merge, not only
after — the current `glucose-dashboard` pipeline only triggers on `push` to
`main`.
- **NFR5 — Framework**: make a descision on which framework the solution can be based best. For example streamlit, shiny for python etc.
## 6. Data requirements
@@ -187,16 +251,16 @@ separate store — **decide during design**.
## 8. Open questions for the user's edit pass
1. Which LXC (102 or 107) ends up hosting the service, once FR1 is done?
2. Which autotune Docker image to standardize on —
`p5nbTgip0r/autotune-docker`, `houthacker42/nightscout-autotune`, or
building directly from the `oref0` npm package? (Design-phase decision,
informed by FR1's findings.)
1. ~~Which LXC (102 or 107 or new) ends up hosting the service, once FR1 is
done?~~ **Resolved (FR1 complete, see below): LXC 102 (`autotune`).**
2. ~~Which autotune Docker image to standardize on?~~ **Resolved: the image
already installed on LXC 102 —
`ghcr.io/houthacker/nightscout-autotune:latest`.** See findings below.
3. Should automatic Profile Switch activation become an opt-in "phase 2" later,
once you've built trust in the recommendations — or should manual
activation remain permanent policy regardless of trust?
activation remain permanent policy regardless of trust? Can already start automate it as i've added in the requirements
4. FR7 (audit/history) — is reusing Nightscout's own profile-document history
sufficient, or do you want a dedicated log outside Nightscout?
sufficient, or do you want a dedicated log outside Nightscout? Create a dedicated store of analysis and include proper production grade logging for the full application so debugging later will be easier.
5. Confirm the separate write-scoped Nightscout token is something you're
comfortable creating/managing (NFR2) — this is a manual one-time setup step
in the Nightscout admin UI, not something the app can do for you.
in the Nightscout admin UI, not something the app can do for you. Yes, i can create a new token, if you indicate which rights it needs.
+112 -29
View File
@@ -31,6 +31,38 @@ SENSOR_CHANGE_EVENT_TYPES = {"Sensor Change", "Sensor Start"}
_REQUEST_TIMEOUT = 10
# Nightscout's /api/v1/entries en /api/v1/treatments negeren een groot
# `count`-verzoek: het antwoord wordt hoe dan ook stilzwijgend afgeknipt rond
# een paar honderd tot ~1150 documenten, ongeacht hoe ver de opgevraagde
# periode teruggaat (waargenomen: count=100000 op entries gaf slechts de
# meest recente ~1150 metingen terug — nog geen 4 dagen — terwijl er wel
# degelijk weken oudere data in Nightscout beschikbaar bleek te zijn). Eén
# verzoek is dus niet genoeg voor meerdaagse rapportages; _paginated_get()
# loopt terug in de tijd met `find[...][$lte]`-cursors tot de gevraagde
# `since`-grens bereikt is of de server minder dan een volle pagina teruggeeft
# (= geen oudere data meer beschikbaar, wat verschilt van "afgeknipt").
#
# De twee endpoints filteren op een ANDER veld, en /treatments heeft nog een
# addertje: entries filtert/sorteert op het numerieke `date` (epoch ms) en
# geeft zonder filter gewoon de meest recente `count` terug. treatments
# filtert/sorteert op het string-veld `created_at` (ISO 8601) — filteren op
# `date` levert daar stil een lege pagina op zodra je voorbij een intern
# venster van ~4 dagen gaat, ook al bevatten de documenten prima een geldig
# `date`-veld (geverifieerd rechtstreeks in MongoDB: alle 123k+ treatments
# hebben een numeriek `date`, en zonder enig filter geeft het endpoint alsnog
# stilzwijgend niet meer dan ~4 dagen terug). Zodra je expliciet op
# `created_at` filtert — zelfs met "nu" als bovengrens op de eerste pagina —
# werkt paginering wél correct en is de volledige geschiedenis (maanden,
# geverifieerd in de database) gewoon opvraagbaar. Vandaar: de cursor wordt
# nooit weggelaten (ook niet op de allereerste pagina), en welk veld
# gebruikt wordt is per endpoint expliciet ("date" voor entries,
# "created_at" voor treatments) — zie ENTRIES_CURSOR_FIELD/TREATMENTS_CURSOR_FIELD.
_PAGE_SIZE = 1000
_MAX_PAGES = 150 # 150 000 documenten; ruim boven wat een 90-dagenrapport nodig heeft
ENTRIES_CURSOR_FIELD = "date"
TREATMENTS_CURSOR_FIELD = "created_at"
class NightscoutConfigError(RuntimeError):
"""NIGHTSCOUT_URL ontbreekt of is ongeldig."""
@@ -75,15 +107,62 @@ def _get(path: str, params: dict) -> list:
return response.json()
def _entries_since(since: datetime, count: int, until: datetime | None = None) -> pd.DataFrame:
"""Haal entries op sinds `since` (en optioneel tot `until`), lokaal gefilterd.
def _format_cursor(cursor_dt: datetime, cursor_field: str) -> int | str:
"""Zet een cursor-datetime om naar de vorm die het gekozen veld verwacht."""
if cursor_field == ENTRIES_CURSOR_FIELD:
return int(cursor_dt.timestamp() * 1000)
# created_at: milliseconde-precisie ISO 8601 met Z-suffix, exact het
# formaat dat Nightscout zelf in created_at opslaat.
return cursor_dt.strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3] + "Z"
Nightscout geeft resultaten terug in aflopende volgorde; vraag ruim genoeg
entries op (CGM meet doorgaans elke 5 min) en filter lokaal op tijd. Dit is
robuuster dan server-side find[]-query's, die niet op elke Nightscout-versie
identiek werken.
def _parse_item_time(item: dict, cursor_field: str) -> datetime | None:
"""Lees het tijdstip van een document uit het veld waar dit endpoint op filtert."""
if cursor_field not in item:
return None
if cursor_field == ENTRIES_CURSOR_FIELD:
return datetime.fromtimestamp(item[cursor_field] / 1000, tz=timezone.utc)
return datetime.fromisoformat(item[cursor_field].replace("Z", "+00:00"))
def _paginated_get(
path: str, since: datetime, until: datetime | None, cursor_field: str
) -> list:
"""Haal alle documenten op tussen `since` en `until` (of tot nu), pagina voor pagina.
Zie de toelichting bij _PAGE_SIZE hierboven voor waarom dit niet één
verzoek met een groot `count` kan zijn, en waarom `cursor_field` per
endpoint verschilt en nooit wordt weggelaten (BUG, zowel de originele
count-truncatie als treatments' verborgen ~4-dagenvenster zonder
expliciet find[created_at]-filter). Elke pagina wordt opgevraagd met een
cursor net onder de oudste timestamp van de vorige pagina, zodat elk
document precies één keer terugkomt. Stopt zodra de oudste timestamp in
een pagina op of voorbij `since` ligt, of zodra de server minder dan een
volle pagina teruggeeft (dan is er geen oudere data meer, punt).
"""
data = _get("/api/v1/entries/sgv.json", {"count": count})
cursor_dt = until if until is not None else datetime.now(timezone.utc)
results: list = []
for _ in range(_MAX_PAGES):
params = {"count": _PAGE_SIZE, f"find[{cursor_field}][$lte]": _format_cursor(cursor_dt, cursor_field)}
page = _get(path, params)
if not page:
break
results.extend(page)
item_times = [t for t in (_parse_item_time(item, cursor_field) for item in page) if t is not None]
if not item_times:
break
oldest_dt = min(item_times)
if oldest_dt <= since or len(page) < _PAGE_SIZE:
break
cursor_dt = oldest_dt - timedelta(milliseconds=1)
return results
def _entries_since(since: datetime, until: datetime | None = None) -> pd.DataFrame:
"""Haal entries op sinds `since` (en optioneel tot `until`), lokaal gefilterd."""
data = _paginated_get("/api/v1/entries/sgv.json", since, until, ENTRIES_CURSOR_FIELD)
if not data:
return pd.DataFrame(columns=["timestamp", "mmol"])
@@ -97,29 +176,39 @@ def _entries_since(since: datetime, count: int, until: datetime | None = None) -
return df[["timestamp", "mmol"]].sort_values("timestamp").reset_index(drop=True)
def _treatments_since(since: datetime, count: int, until: datetime | None = None) -> pd.DataFrame:
"""Haal treatments op sinds `since` (en optioneel tot `until`), lokaal gefilterd."""
data = _get("/api/v1/treatments.json", {"count": count})
_TREATMENT_COLUMNS = ["timestamp", "eventType", "insulin", "carbs", "type", "absolute", "duration"]
def _treatments_since(since: datetime, until: datetime | None = None) -> pd.DataFrame:
"""Haal treatments op sinds `since` (en optioneel tot `until`), lokaal gefilterd.
`type` en `absolute`/`duration` komen erbij naast de oorspronkelijke
kolommen: AndroidAPS registreert een SMB-microbolus niet als eigen
`eventType`, maar als `eventType == "Correction Bolus"` met
`type == "SMB"` (zie reporting.bolus_stats() / gitea #20) — en
`absolute` (E/uur) + `duration` (minuten) op Temp Basal-treatments maken
het mogelijk afgeleverde basaalinsuline te berekenen zonder het
pompprofiel te hoeven ophalen (zie reporting.basal_stats()).
"""
data = _paginated_get("/api/v1/treatments.json", since, until, TREATMENTS_CURSOR_FIELD)
if not data:
return pd.DataFrame(columns=["timestamp", "eventType", "insulin", "carbs"])
return pd.DataFrame(columns=_TREATMENT_COLUMNS)
df = pd.DataFrame(data)
df["timestamp"] = pd.to_datetime(df["created_at"], utc=True).dt.tz_localize(None)
for col in ("insulin", "carbs"):
for col in ("insulin", "carbs", "absolute", "duration"):
if col not in df.columns:
df[col] = pd.NA
df[col] = pd.to_numeric(df[col], errors="coerce")
if "eventType" not in df.columns:
df["eventType"] = ""
for col in ("eventType", "type"):
if col not in df.columns:
df[col] = ""
df[col] = df[col].fillna("")
df = df[df["timestamp"] >= since.replace(tzinfo=None)]
if until is not None:
df = df[df["timestamp"] <= until.replace(tzinfo=None)]
return (
df[["timestamp", "eventType", "insulin", "carbs"]]
.sort_values("timestamp")
.reset_index(drop=True)
)
return df[_TREATMENT_COLUMNS].sort_values("timestamp").reset_index(drop=True)
def fetch_entries(hours: int = 24) -> pd.DataFrame:
@@ -129,8 +218,7 @@ def fetch_entries(hours: int = 24) -> pd.DataFrame:
gesorteerd op tijd.
"""
since = datetime.now(timezone.utc) - timedelta(hours=hours)
count = max(300, hours * 15)
return _entries_since(since, count)
return _entries_since(since)
def fetch_treatments(hours: int = 24) -> pd.DataFrame:
@@ -141,8 +229,7 @@ def fetch_treatments(hours: int = 24) -> pd.DataFrame:
'insulin' en 'carbs' (de laatste twee NaN waar niet van toepassing).
"""
since = datetime.now(timezone.utc) - timedelta(hours=hours)
count = max(200, hours * 10)
return _treatments_since(since, count)
return _treatments_since(since)
def fetch_entries_range(start_date: date, end_date: date) -> pd.DataFrame:
@@ -154,9 +241,7 @@ def fetch_entries_range(start_date: date, end_date: date) -> pd.DataFrame:
"""
since = datetime.combine(start_date, time.min, tzinfo=timezone.utc)
until = datetime.combine(end_date, time.max, tzinfo=timezone.utc)
days = (end_date - start_date).days + 1
count = min(days * 300, 100_000) # ruime marge boven 5-min-interval
return _entries_since(since, count, until)
return _entries_since(since, until)
def fetch_treatments_range(start_date: date, end_date: date) -> pd.DataFrame:
@@ -166,6 +251,4 @@ def fetch_treatments_range(start_date: date, end_date: date) -> pd.DataFrame:
"""
since = datetime.combine(start_date, time.min, tzinfo=timezone.utc)
until = datetime.combine(end_date, time.max, tzinfo=timezone.utc)
days = (end_date - start_date).days + 1
count = min(days * 50, 50_000)
return _treatments_since(since, count, until)
return _treatments_since(since, until)
+17 -7
View File
@@ -4,12 +4,19 @@ import streamlit as st
from auth import require_auth
from nightscout_client import NightscoutConfigError
from reporting import (
gmi_mmol_per_mol,
gmi_percent,
percent_cv,
sensor_active_percent,
tir_breakdown_5band,
)
from ui_common import load_range_data, select_period_days
from ui_common import (
build_tir_stacked_bar_figure,
get_colors,
load_range_data,
render_completeness_warning,
select_period_days,
)
st.set_page_config(page_title="Glucose dashboard — Samenvatting", page_icon="📊", layout="wide")
require_auth()
@@ -32,15 +39,12 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
active_pct = sensor_active_percent(glucose, start, end)
if active_pct < 70:
st.warning(
f"Slechts {active_pct:.0f}% sensordekking in deze periode (streefwaarde ≥70%). "
"De onderstaande cijfers zijn mogelijk minder betrouwbaar door de vele hiaten."
)
mean_mmol = glucose["mmol"].mean()
gmi = gmi_percent(mean_mmol)
gmi_mmol_mol = gmi_mmol_per_mol(gmi)
cv = percent_cv(glucose)
tir = tir_breakdown_5band(glucose)
@@ -48,7 +52,11 @@ st.subheader("Kerncijfers")
col1, col2, col3, col4 = st.columns(4)
col1.metric("Sensordekking", f"{active_pct:.0f}%", help="Streefwaarde: ≥70%")
col2.metric("Gemiddelde glucose", f"{mean_mmol:.1f} mmol/L")
col3.metric("Geschat HbA1c (GMI)", f"{gmi:.1f}%", help="Bergenstal et al. 2018 — geen labwaarde")
col3.metric(
"Geschat HbA1c (GMI)",
f"{gmi:.1f}% ({gmi_mmol_mol:.0f} mmol/mol)",
help="Bergenstal et al. 2018 — geen labwaarde",
)
col4.metric("Variabiliteit (%CV)", f"{cv:.0f}%", help="Streefwaarde: ≤36%")
st.subheader("Tijd in bereik (5 klinische banden)")
@@ -70,6 +78,8 @@ cols = st.columns(5)
for col, band in zip(cols, band_labels, strict=True):
col.metric(band_labels[band], f"{tir[band]:.0f}%", help=f"Streefwaarde: {band_targets[band]}")
st.plotly_chart(build_tir_stacked_bar_figure(tir, get_colors()), width="stretch")
st.caption(
f"Periode: {start.strftime('%d-%m-%Y')} t/m {end.strftime('%d-%m-%Y')} "
f"({len(glucose)} metingen) · streefwaarden volgens Battelino et al. 2019"
+11 -1
View File
@@ -6,7 +6,14 @@ from auth import require_auth
from data_processing import TARGET_HIGH, TARGET_LOW
from nightscout_client import NightscoutConfigError
from reporting import agp_percentile_bands
from ui_common import get_colors, hex_to_rgba, load_range_data, select_period_days
from ui_common import (
add_threshold_lines,
get_colors,
hex_to_rgba,
load_range_data,
render_completeness_warning,
select_period_days,
)
st.set_page_config(page_title="Glucose dashboard — AGP profiel", page_icon="📊", layout="wide")
require_auth()
@@ -32,6 +39,8 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
bands = agp_percentile_bands(glucose, bin_minutes=30)
colors = get_colors()
@@ -94,6 +103,7 @@ fig.update_layout(
yaxis={"title": "mmol/L", "gridcolor": colors["grid"], "zeroline": False},
xaxis={"title": "Uur van de dag", "gridcolor": colors["grid"], "tickmode": "linear", "dtick": 2},
)
add_threshold_lines(fig, colors)
st.plotly_chart(fig, width="stretch")
st.caption(
f"Periode: {start.strftime('%d-%m-%Y')} t/m {end.strftime('%d-%m-%Y')} ({len(glucose)} metingen)"
+10 -1
View File
@@ -6,7 +6,13 @@ from plotly.subplots import make_subplots
from auth import require_auth
from data_processing import TARGET_HIGH, TARGET_LOW
from nightscout_client import NightscoutConfigError
from ui_common import get_colors, load_range_data, select_period_days
from ui_common import (
add_threshold_lines,
get_colors,
load_range_data,
render_completeness_warning,
select_period_days,
)
st.set_page_config(page_title="Glucose dashboard — Dagoverzicht", page_icon="📊", layout="wide")
require_auth()
@@ -34,6 +40,8 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
colors = get_colors()
glucose = glucose.copy()
glucose["date"] = glucose["timestamp"].dt.date
@@ -67,6 +75,7 @@ for i, day in enumerate(dates, start=1):
col=1,
)
fig.update_yaxes(title_text="mmol/L", gridcolor=colors["grid"], row=i, col=1)
add_threshold_lines(fig, colors, row=i)
fig.update_xaxes(title_text="Uur van de dag", range=[0, 24], dtick=4, gridcolor=colors["grid"])
fig.update_layout(
+3
View File
@@ -7,6 +7,7 @@ from reporting import LOW, VERY_LOW
from ui_common import (
get_colors,
load_range_data,
render_completeness_warning,
render_event_level,
select_period_days,
)
@@ -36,6 +37,8 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
colors = get_colors()
render_event_level(glucose, LOW, "below", f"Level 1 (<{LOW} mmol/L)", days, colors)
+3
View File
@@ -7,6 +7,7 @@ from reporting import HIGH, VERY_HIGH
from ui_common import (
get_colors,
load_range_data,
render_completeness_warning,
render_event_level,
select_period_days,
)
@@ -35,6 +36,8 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
colors = get_colors()
render_event_level(glucose, HIGH, "above", f"Level 1 (>{HIGH} mmol/L)", days, colors)
+81 -21
View File
@@ -6,13 +6,20 @@ from auth import require_auth
from data_processing import classify_treatments
from nightscout_client import NightscoutConfigError
from reporting import (
basal_stats,
bolus_stats,
day_of_week_pattern,
gmi_percent,
gmi_trend,
overnight_daytime_split,
site_interval_stats,
)
from ui_common import get_colors, load_range_data, select_period_days
from ui_common import (
get_colors,
hex_to_rgba,
load_range_data,
render_completeness_warning,
select_period_days,
)
st.set_page_config(page_title="Glucose dashboard — Trends", page_icon="📊", layout="wide")
require_auth()
@@ -35,29 +42,75 @@ if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
colors = get_colors()
events = classify_treatments(treatments)
# --- FR7: geschat HbA1c (GMI) per maand ------------------------------------
st.subheader("Geschat HbA1c (GMI) per maand")
monthly = glucose.copy()
monthly["month"] = monthly["timestamp"].dt.to_period("M").astype(str)
monthly_gmi = monthly.groupby("month")["mmol"].mean().apply(gmi_percent)
# --- FR7: geschat HbA1c (GMI) trend ----------------------------------------
st.subheader("Geschat HbA1c (GMI) trend")
trend = gmi_trend(glucose)
if len(monthly_gmi) >= 1:
fig = go.Figure(
go.Bar(x=monthly_gmi.index, y=monthly_gmi.to_numpy(), marker_color=colors["glucose"])
if len(trend) >= 2:
fig = go.Figure()
fig.add_trace(
go.Scatter(
x=trend["date"], y=trend["trend_upper"], mode="lines", line={"width": 0}, showlegend=False,
hoverinfo="skip",
)
)
fig.add_trace(
go.Scatter(
x=trend["date"],
y=trend["trend_lower"],
mode="lines",
line={"width": 0},
fill="tonexty",
fillcolor=hex_to_rgba(colors["muted"], 0.2),
name="95% betrouwbaarheidsband",
hoverinfo="skip",
)
)
fig.add_trace(
go.Scatter(
x=trend["date"],
y=trend["trend"],
mode="lines",
line={"color": colors["muted"], "width": 2, "dash": "dash"},
name="Trend (lineair)",
)
)
fig.add_trace(
go.Scatter(
x=trend["date"],
y=trend["rolling_avg"],
mode="lines",
line={"color": colors["glucose"], "width": 2},
name="Voortschrijdend gemiddelde (7d)",
)
)
fig.add_trace(
go.Scatter(
x=trend["date"],
y=trend["gmi"],
mode="markers",
marker={"color": colors["glucose"], "size": 5, "opacity": 0.5},
name="Dag-GMI",
)
)
fig.update_layout(
paper_bgcolor="rgba(0,0,0,0)",
plot_bgcolor="rgba(0,0,0,0)",
font_color=colors["text"],
height=280,
height=320,
margin={"l": 10, "r": 10, "t": 10, "b": 10},
yaxis={"title": "GMI (%)", "gridcolor": colors["grid"]},
xaxis={"title": "Maand", "gridcolor": colors["grid"]},
xaxis={"title": "Datum", "gridcolor": colors["grid"]},
legend={"orientation": "h", "yanchor": "bottom", "y": 1.02, "xanchor": "left", "x": 0},
)
st.plotly_chart(fig, width="stretch")
else:
st.info("Niet genoeg dagen met data voor een trendlijn (minstens 2 nodig).")
st.caption("GMI is een schatting (Bergenstal et al. 2018), geen labwaarde.")
# --- FR8: nacht vs. overdag --------------------------------------------------
@@ -86,18 +139,25 @@ if not weekday_plot.empty:
)
st.plotly_chart(fig, width="stretch")
# --- FR8: bolus-insuline --------------------------------------------------
st.subheader("Bolus-insuline")
# --- FR8: insuline (basaal + bolus-uitsplitsing) ---------------------------
st.subheader("Insuline")
st.caption(
"Alleen bolusinsuline (maaltijd/correctie/combo/SMB) — basaalinsuline vereist "
"profielgegevens en is hier niet in verwerkt (zie docs/design-reporting.md)."
"Basaal berekend uit Temp Basal-treatments (snelheid x duur), zonder pompprofiel "
"nodig — een ondergrens bij periodes zonder actieve temp basal. Boluses "
"opgesplitst naar voeding, handmatige correctie en automatische SMB-microbolus "
"(gitea #20)."
)
bolus = bolus_stats(treatments)
col1, col2 = st.columns(2)
col1.metric("Totaal bolus", f"{bolus['total_units']:.1f} E")
col2.metric("Aantal boluses", bolus["count"])
if bolus["by_type"]:
st.table({"type": list(bolus["by_type"].keys()), "eenheden": list(bolus["by_type"].values())})
basal = basal_stats(treatments, days)
total_units = bolus["total_units"] + basal["total_units"]
col1, col2, col3 = st.columns(3)
col1.metric("Totaal insuline", f"{total_units:.1f} E")
col2.metric("Gem. per dag", f"{total_units / days:.1f} E")
col3.metric("Aantal boluses", bolus["count"])
breakdown = {"Basaal": basal["total_units"], **bolus["by_type"]}
if any(breakdown.values()):
st.table({"Categorie": list(breakdown.keys()), "Eenheden": list(breakdown.values())})
# --- FR8: pompsite-leeftijd vs. glucose/insuline ---------------------------
st.subheader("Pompsite-leeftijd vs. glucose en insuline")
+54
View File
@@ -0,0 +1,54 @@
import requests
import streamlit as st
from auth import require_auth
from nightscout_client import NightscoutConfigError
from pdf_export import AVAILABLE_SECTIONS, SECTION_EVENTS, build_report_pdf
from reporting import HIGH, LOW, VERY_HIGH, VERY_LOW, detect_events, summarize_events
from ui_common import load_range_data, render_completeness_warning, select_period_days
st.set_page_config(page_title="Glucose dashboard — Export", page_icon="📊", layout="wide")
require_auth()
st.title("Export")
st.caption("Stel een PDF-rapport samen om naar het diabetesteam te sturen (FR10).")
start, end, days = select_period_days(default_label="90 dagen")
sections = st.multiselect("Onderdelen", AVAILABLE_SECTIONS, default=AVAILABLE_SECTIONS)
try:
glucose, _treatments = load_range_data(start, end)
except NightscoutConfigError as exc:
st.error(str(exc))
st.stop()
except (requests.RequestException, ValueError) as exc:
st.error(f"Kon geen data ophalen van Nightscout: {exc}")
st.stop()
if glucose.empty:
st.info("Geen glucosemetingen gevonden in deze periode.")
st.stop()
render_completeness_warning(glucose, start, end)
events_summary = {}
if SECTION_EVENTS in sections:
levels = {
f"Hypo Level 1 (<{LOW})": (LOW, "below"),
f"Hypo Level 2 (<{VERY_LOW})": (VERY_LOW, "below"),
f"Hyper Level 1 (>{HIGH})": (HIGH, "above"),
f"Hyper Level 2 (>{VERY_HIGH})": (VERY_HIGH, "above"),
}
for label, (threshold, direction) in levels.items():
events = detect_events(glucose, threshold, direction)
events_summary[label] = summarize_events(events, days)
pdf_bytes = build_report_pdf(sections, glucose, events_summary, start, end)
st.download_button(
"Download PDF",
data=pdf_bytes,
file_name=f"glucoserapport_{start.strftime('%Y%m%d')}_{end.strftime('%Y%m%d')}.pdf",
mime="application/pdf",
)
st.caption(f"Periode: {start.strftime('%d-%m-%Y')} t/m {end.strftime('%d-%m-%Y')}")
+286
View File
@@ -0,0 +1,286 @@
"""
PDF-rapportgeneratie voor geselecteerde rapportageonderdelen (FR10).
Gebruikt fpdf2 (puur Python, geen systeemdependencies) voor de PDF-opbouw en
matplotlib voor een eenvoudige statische AGP-grafiek specifiek voor de PDF —
bewust niet dezelfde Plotly-figuren als op de pagina's zelf (die zouden
kaleido vereisen, met een ingebakken Chromium-build: een groter image en meer
Trivy-oppervlak, voor dit doel niet nodig).
"""
from __future__ import annotations
import io
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import pandas as pd
from fpdf import FPDF
from fpdf.enums import XPos, YPos
from reporting import (
HIGH,
LOW,
VERY_HIGH,
VERY_LOW,
agp_percentile_bands,
completeness_warning,
detect_events,
gmi_mmol_per_mol,
gmi_percent,
percent_cv,
sensor_active_percent,
tir_breakdown_5band,
)
SECTION_SUMMARY = "Samenvatting"
SECTION_AGP = "AGP profiel"
SECTION_DAGOVERZICHT = "Dagoverzicht"
SECTION_EVENTS = "Hypo-/hyperevents"
AVAILABLE_SECTIONS = [SECTION_SUMMARY, SECTION_AGP, SECTION_DAGOVERZICHT, SECTION_EVENTS]
_LINE_KWARGS = {"new_x": XPos.LMARGIN, "new_y": YPos.NEXT}
# Zelfde kleuren als het PALETTE["light"] in ui_common.py — bewust hier
# gedupliceerd (5 hex-waarden) in plaats van ui_common.py te importeren, want
# dat zou een streamlit-dependency in deze module trekken die verder bewust
# vrij van UI-framework-imports is (los te unit-testen, zie moduledocstring).
_MUTED = "#898781"
_TIR_BAND_ORDER = ["very_low", "low", "in_range", "high", "very_high"]
_TIR_BAND_LABELS = {
"very_low": "Zeer laag",
"low": "Laag",
"in_range": "In bereik",
"high": "Hoog",
"very_high": "Zeer hoog",
}
_TIR_COLORS = {
"very_low": "#8b1a1a",
"low": "#d03b3b",
"in_range": "#0ca30c",
"high": "#fab219",
"very_high": "#c76f0a",
}
_THRESHOLD_LINES = [VERY_LOW, LOW, HIGH, VERY_HIGH]
DAGOVERZICHT_MAX_DAYS = 14
def _add_threshold_lines(ax) -> None:
"""Teken de 4 klinische drempellijnen (3.0/3.9/10.0/13.9 mmol/L) op een matplotlib-as.
PDF-tegenhanger van ui_common.add_threshold_lines() voor de Plotly-versie
op de dashboardpagina's (gitea #15) — zelfde 4 waarden, zelfde grijstint.
"""
for value in _THRESHOLD_LINES:
ax.axhline(value, color=_MUTED, linewidth=0.7, linestyle=":")
def _agp_figure_png(glucose: pd.DataFrame) -> bytes:
bands = agp_percentile_bands(glucose)
fig, ax = plt.subplots(figsize=(7, 3))
hours = bands["minute_of_day"] / 60
ax.fill_between(hours, bands["p5"], bands["p95"], alpha=0.15, color="#2a78d6", label="5e-95e pct")
ax.fill_between(hours, bands["p25"], bands["p75"], alpha=0.3, color="#2a78d6", label="25e-75e pct")
ax.plot(hours, bands["p50"], color="#2a78d6", linewidth=2, label="Mediaan")
_add_threshold_lines(ax)
ax.set_xlabel("Uur van de dag")
ax.set_ylabel("mmol/L")
ax.set_xlim(0, 24)
ax.legend(loc="upper right", fontsize=8)
fig.tight_layout()
buf = io.BytesIO()
fig.savefig(buf, format="png", dpi=150)
plt.close(fig)
buf.seek(0)
return buf.getvalue()
def _tir_bar_figure_png(tir: dict) -> bytes:
"""Gestapelde staafgrafiek van de 5 TIR-banden — PDF-tegenhanger van
ui_common.build_tir_stacked_bar_figure() (gitea #14)."""
fig, ax = plt.subplots(figsize=(7, 1.1))
left = 0.0
for band in _TIR_BAND_ORDER:
value = tir[band]
ax.barh(0, value, left=left, color=_TIR_COLORS[band], label=_TIR_BAND_LABELS[band])
left += value
ax.set_xlim(0, 100)
ax.set_yticks([])
ax.set_xlabel("%")
ax.legend(loc="upper center", bbox_to_anchor=(0.5, -0.75), ncol=5, fontsize=7, frameon=False)
buf = io.BytesIO()
# bbox_inches="tight" i.p.v. fig.tight_layout(): de legend zit buiten de
# axes (bbox_to_anchor met negatieve y), waar tight_layout() niet voor
# kan compenseren — savefig's eigen bbox-berekening kan dat wel.
fig.savefig(buf, format="png", dpi=150, bbox_inches="tight")
plt.close(fig)
buf.seek(0)
return buf.getvalue()
def _daily_figure_png(day_df: pd.DataFrame, day_label: str) -> bytes:
"""Eén dag als kleine lijngrafiek — PDF-tegenhanger van de Dagoverzicht-pagina (gitea #16)."""
hours = day_df["timestamp"].dt.hour + day_df["timestamp"].dt.minute / 60
fig, ax = plt.subplots(figsize=(7, 1.4))
ax.plot(hours, day_df["mmol"], color="#2a78d6", linewidth=1.2)
_add_threshold_lines(ax)
ax.set_xlim(0, 24)
ax.set_xticks(range(0, 25, 4))
ax.set_ylabel("mmol/L", fontsize=8)
ax.set_title(day_label, fontsize=9, loc="left")
ax.tick_params(labelsize=7)
fig.tight_layout()
buf = io.BytesIO()
fig.savefig(buf, format="png", dpi=150)
plt.close(fig)
buf.seek(0)
return buf.getvalue()
def _event_hour_histogram_png(events: pd.DataFrame, color: str) -> bytes:
"""Aantal events per uur-van-de-dag — PDF-tegenhanger van render_event_level()'s
histogram op de Hypo/Hyper Events-pagina's (gitea #17, #18)."""
hour_counts = events["start"].dt.hour.value_counts().reindex(range(24), fill_value=0).sort_index()
fig, ax = plt.subplots(figsize=(7, 2))
ax.bar(hour_counts.index, hour_counts.to_numpy(), color=color)
ax.set_xlabel("Uur van de dag")
ax.set_ylabel("Aantal events")
ax.set_xlim(-0.5, 23.5)
ax.set_xticks(range(0, 24, 2))
fig.tight_layout()
buf = io.BytesIO()
fig.savefig(buf, format="png", dpi=150)
plt.close(fig)
buf.seek(0)
return buf.getvalue()
def _add_summary_section(pdf: FPDF, glucose: pd.DataFrame, start, end) -> None:
mean_mmol = glucose["mmol"].mean()
tir = tir_breakdown_5band(glucose)
active_pct = sensor_active_percent(glucose, start, end)
pdf.set_font("Helvetica", "B", 13)
pdf.cell(0, 8, "Samenvatting", **_LINE_KWARGS)
warning = completeness_warning(glucose, start, end)
if warning:
pdf.set_font("Helvetica", "BI", 10)
pdf.multi_cell(0, 5, warning)
pdf.ln(2)
pdf.set_font("Helvetica", "", 11)
pdf.cell(0, 6, f"Sensordekking: {active_pct:.0f}%", **_LINE_KWARGS)
pdf.cell(0, 6, f"Gemiddelde glucose: {mean_mmol:.1f} mmol/L", **_LINE_KWARGS)
gmi = gmi_percent(mean_mmol)
pdf.cell(0, 6, f"Geschat HbA1c (GMI): {gmi:.1f}% ({gmi_mmol_per_mol(gmi):.0f} mmol/mol)", **_LINE_KWARGS)
pdf.cell(0, 6, f"Variabiliteit (%CV): {percent_cv(glucose):.0f}%", **_LINE_KWARGS)
pdf.ln(2)
pdf.cell(0, 6, f"Zeer laag (<3.0 mmol/L): {tir['very_low']:.0f}%", **_LINE_KWARGS)
pdf.cell(0, 6, f"Laag (3.0-3.9 mmol/L): {tir['low']:.0f}%", **_LINE_KWARGS)
pdf.cell(0, 6, f"In bereik (3.9-10.0 mmol/L): {tir['in_range']:.0f}%", **_LINE_KWARGS)
pdf.cell(0, 6, f"Hoog (10.0-13.9 mmol/L): {tir['high']:.0f}%", **_LINE_KWARGS)
pdf.cell(0, 6, f"Zeer hoog (>13.9 mmol/L): {tir['very_high']:.0f}%", **_LINE_KWARGS)
pdf.ln(2)
png_bytes = _tir_bar_figure_png(tir)
pdf.image(io.BytesIO(png_bytes), w=160)
pdf.ln(4)
def _add_agp_section(pdf: FPDF, glucose: pd.DataFrame) -> None:
pdf.set_font("Helvetica", "B", 13)
pdf.cell(0, 8, "AGP profiel", **_LINE_KWARGS)
png_bytes = _agp_figure_png(glucose)
pdf.image(io.BytesIO(png_bytes), w=180)
pdf.ln(4)
def _add_dagoverzicht_section(pdf: FPDF, glucose: pd.DataFrame) -> None:
"""Eén dagelijkse mini-grafiek per kalenderdag (gitea #16), max. de meest
recente DAGOVERZICHT_MAX_DAYS dagen — voor langere periodes zou de PDF
anders onwerkbaar lang worden."""
pdf.set_font("Helvetica", "B", 13)
pdf.cell(0, 8, "Dagoverzicht", **_LINE_KWARGS)
df = glucose.copy()
df["date"] = df["timestamp"].dt.date
dates = sorted(df["date"].unique())
if len(dates) > DAGOVERZICHT_MAX_DAYS:
dates = dates[-DAGOVERZICHT_MAX_DAYS:]
pdf.set_font("Helvetica", "I", 9)
pdf.cell(0, 5, f"Beperkt tot de meest recente {DAGOVERZICHT_MAX_DAYS} dagen.", **_LINE_KWARGS)
for day in dates:
day_df = df[df["date"] == day]
png_bytes = _daily_figure_png(day_df, day.strftime("%A %d-%m-%Y"))
pdf.image(io.BytesIO(png_bytes), w=180)
pdf.ln(4)
def _add_events_section(pdf: FPDF, glucose: pd.DataFrame, events_summary: dict[str, dict]) -> None:
pdf.set_font("Helvetica", "B", 13)
pdf.cell(0, 8, "Hypo-/hyperevents", **_LINE_KWARGS)
pdf.set_font("Helvetica", "", 11)
for label, summary in events_summary.items():
pdf.cell(
0,
6,
f"{label}: {summary['count']} events, gemiddeld {summary['avg_duration_minutes']:.0f} min, "
f"{summary['events_per_week']:.1f}/week",
**_LINE_KWARGS,
)
pdf.ln(2)
# Alleen Level 1 krijgt een grafiek in het PDF-rapport (gitea #17, #18) —
# geen tabel met losse events, de samenvattende cijfers staan al hierboven.
if not glucose.empty:
hypo_events = detect_events(glucose, LOW, "below")
if not hypo_events.empty:
pdf.image(io.BytesIO(_event_hour_histogram_png(hypo_events, _TIR_COLORS["low"])), w=180)
hyper_events = detect_events(glucose, HIGH, "above")
if not hyper_events.empty:
pdf.image(io.BytesIO(_event_hour_histogram_png(hyper_events, _TIR_COLORS["high"])), w=180)
pdf.ln(4)
def build_report_pdf(
sections: list[str],
glucose: pd.DataFrame,
events_summary: dict[str, dict],
start,
end,
) -> bytes:
"""Bouw een PDF met de geselecteerde onderdelen.
`events_summary` bevat vooraf berekende hypo/hyper-samenvattingen (label ->
resultaat van reporting.summarize_events), zodat deze module geen
event-detectie hoeft te herhalen.
"""
pdf = FPDF()
pdf.add_page()
pdf.set_font("Helvetica", "B", 16)
pdf.cell(0, 10, "Glucoserapport", **_LINE_KWARGS)
pdf.set_font("Helvetica", "", 11)
pdf.cell(0, 8, f"Periode: {start.strftime('%d-%m-%Y')} t/m {end.strftime('%d-%m-%Y')}", **_LINE_KWARGS)
pdf.ln(4)
if not glucose.empty:
if SECTION_SUMMARY in sections:
_add_summary_section(pdf, glucose, start, end)
if SECTION_AGP in sections:
_add_agp_section(pdf, glucose)
if SECTION_DAGOVERZICHT in sections:
_add_dagoverzicht_section(pdf, glucose)
if SECTION_EVENTS in sections and events_summary:
_add_events_section(pdf, glucose, events_summary)
return bytes(pdf.output())
+172 -6
View File
@@ -11,6 +11,7 @@ from __future__ import annotations
from itertools import pairwise
import numpy as np
import pandas as pd
from nightscout_client import INSULIN_EVENT_TYPES
@@ -35,6 +36,79 @@ def gmi_percent(mean_mmol: float) -> float:
return round(3.31 + 0.4311 * mean_mmol, 1)
def gmi_mmol_per_mol(gmi_pct: float) -> int:
"""Zet een GMI-percentage (DCCT/NGSP) om naar IFCC mmol/mol.
mmol/mol = 10.929 x (% - 2.15) — de standaard IFCC-omrekening
(Hoelzel et al. 2004, Clin Chem 50(1):166-174), gangbare eenheid in
Nederlandse labuitslagen naast het %-getal (gitea #25). Bijv. 6.5% =
48 mmol/mol, de bekende diagnostische drempel.
"""
return round(10.929 * (gmi_pct - 2.15))
GMI_TREND_COLUMNS = ["date", "gmi", "rolling_avg", "trend", "trend_lower", "trend_upper"]
def daily_gmi(glucose: pd.DataFrame) -> pd.DataFrame:
"""Geschat HbA1c (GMI) per kalenderdag — scatterpunten voor de trendgrafiek (gitea #19)."""
if glucose.empty:
return pd.DataFrame(columns=["date", "gmi"])
df = glucose.copy()
df["date"] = df["timestamp"].dt.date
daily_mean = df.groupby("date")["mmol"].mean()
result = daily_mean.apply(gmi_percent).reset_index()
result.columns = ["date", "gmi"]
return result.sort_values("date").reset_index(drop=True)
def gmi_trend(glucose: pd.DataFrame, rolling_window_days: int = 7) -> pd.DataFrame:
"""Dagelijkse GMI-punten plus voortschrijdend gemiddelde en een lineaire
trendlijn met 95%-betrouwbaarheidsband (gitea #19).
Gebruikt een normaalbenadering (z=1.96) voor de betrouwbaarheidsband i.p.v.
een exacte t-verdeling — dat zou een scipy-dependency toevoegen aan een
verder bewust lichte container (dezelfde afweging als het niet gebruiken
van kaleido in pdf_export.py). Bij de hier gebruikelijke steekproefgroottes
(14-90 dagpunten) is het verschil met de exacte t-kritieke waarde
verwaarloosbaar voor een visuele trendindicatie.
Elke keer opnieuw berekend uit de al-opgehaalde periode-data (geen aparte
database) — de bestaande 1-uurs Streamlit-cache op load_range_data()
voorkomt al herhaalde Nightscout-calls voor dezelfde periode. Bij een
veel langere trend (jaren i.p.v. de huidige max. 90 dagen) zou een aparte
opslag wél de moeite waard worden.
"""
daily = daily_gmi(glucose)
if len(daily) < 2:
return daily.reindex(columns=GMI_TREND_COLUMNS)
daily = daily.copy()
daily["rolling_avg"] = daily["gmi"].rolling(window=rolling_window_days, min_periods=1).mean()
x = np.arange(len(daily), dtype=float)
y = daily["gmi"].to_numpy()
slope, intercept = np.polyfit(x, y, 1)
fitted = intercept + slope * x
residuals = y - fitted
degrees_of_freedom = max(len(x) - 2, 1)
residual_std = np.sqrt(np.sum(residuals**2) / degrees_of_freedom)
x_mean = x.mean()
sum_sq_x = np.sum((x - x_mean) ** 2)
standard_error = (
np.zeros_like(x)
if sum_sq_x == 0
else residual_std * np.sqrt(1 / len(x) + (x - x_mean) ** 2 / sum_sq_x)
)
z_95 = 1.96
daily["trend"] = fitted
daily["trend_lower"] = fitted - z_95 * standard_error
daily["trend_upper"] = fitted + z_95 * standard_error
return daily[GMI_TREND_COLUMNS]
def percent_cv(glucose: pd.DataFrame) -> float:
"""Glucosevariabiliteit als %CV (standaarddeviatie / gemiddelde * 100)."""
if glucose.empty:
@@ -95,6 +169,56 @@ def sensor_active_percent(glucose: pd.DataFrame, start, end, expected_interval_m
return round(min(len(glucose) / expected_readings, 1.0) * 100, 1)
def completeness_warning(glucose: pd.DataFrame, start, end) -> str | None:
"""Bouw een waarschuwingstekst als er te weinig data is voor de gekozen periode.
Pure functie zodat zowel de interactieve pagina's (ui_common.py, als
st.warning) als het PDF-rapport (pdf_export.py, als tekstregel) exact
dezelfde beoordeling en bewoording gebruiken — een rapport dat naar het
diabetesteam gaat moet dezelfde disclaimer bevatten als wat de gebruiker
zelf op het scherm ziet.
Onderscheidt twee oorzaken:
- een structureel gat aan het BEGIN van de periode (de vroegste
binnengekomen meting ligt later dan de gekozen startdatum) — data is
simpelweg niet opgehaald/beschikbaar voor dat stuk, geen sensorhiaat.
- een lagere dekking dan de klinische AGP-ondergrens (Danne et al. 2017)
zonder zo'n gat aan het begin — verspreide hiaten.
Returns:
None als er geen waarschuwing nodig is; anders de te tonen tekst.
"""
if glucose.empty:
return "Geen glucosedata gevonden in deze periode."
start_date = pd.Timestamp(start).date()
end_date = pd.Timestamp(end).date()
active_pct = sensor_active_percent(glucose, start, end)
earliest = glucose["timestamp"].min().date()
missing_days_at_start = max((earliest - start_date).days, 0)
if missing_days_at_start > 0:
return (
f"Nightscout heeft pas data vanaf {earliest:%d-%m-%Y} teruggegeven — "
f"de eerste {missing_days_at_start} dag(en) van de gekozen periode "
f"({start_date:%d-%m-%Y} t/m {end_date:%d-%m-%Y}) ontbreken volledig, niet alleen "
f"losse sensorhiaten. Sensordekking over de wél opgehaalde periode: "
f"{active_pct:.0f}%. Onderstaande cijfers zijn dus berekend over een "
"kortere periode dan geselecteerd."
)
if active_pct < 70:
# ">=" i.p.v. "≥": deze tekst komt ook in het PDF-rapport terecht
# (pdf_export.py), en fpdf2's ingebouwde Helvetica-font ondersteunt
# alleen Latin-1 — zie gitea issue #11 voor het opruimen hiervan
# (Unicode-lettertype embedden) zodra dat de moeite waard is.
return (
f"Slechts {active_pct:.0f}% sensordekking in deze periode (streefwaarde "
">=70% voor een klinisch betrouwbaar rapport, Danne et al. 2017). De "
"onderstaande cijfers zijn mogelijk minder betrouwbaar door de vele hiaten."
)
return None
def _state_blocks(df: pd.DataFrame, out: pd.Series) -> list[dict]:
"""Groepeer aaneengesloten metingen met dezelfde in/out-status.
@@ -280,22 +404,35 @@ def day_of_week_pattern(glucose: pd.DataFrame) -> pd.DataFrame:
return pd.DataFrame(rows, columns=columns)
SMB_TYPE = "SMB"
BASAL_EVENT_TYPE = "Temp Basal"
def bolus_stats(treatments: pd.DataFrame) -> dict:
"""Samenvatting van bolus-insuline, opgesplitst naar type.
Let op: dit betreft alleen bolusinsuline (maaltijd/correctie/combo/SMB).
Basaalinsuline wordt hier niet in meegenomen — dat vereist profielgegevens
(zie docs/design-reporting.md, scoping-kanttekening) en levert dus geen
volledige totale-dagdosis op.
AndroidAPS in SMB-looping registreert een automatische micro-bolus niet
als eigen `eventType`, maar als `eventType == "Correction Bolus"` met
`type == "SMB"` — zonder dit onderscheid werden alle SMB's stilzwijgend
als "Correction Bolus" (handmatige correctie) meegeteld, wat gitea #20
meldde als "bolus stats empty/onduidelijk". Geverifieerd rechtstreeks
tegen de productiedata: bij deze gebruiker is vrijwel elke "Correction
Bolus" in werkelijkheid een automatische SMB.
Let op: dit betreft alleen bolusinsuline (voeding/correctie/combo/SMB).
Basaalinsuline zit hier niet in — zie basal_stats().
"""
if treatments.empty:
return {"total_units": 0.0, "count": 0, "by_type": {}}
bolus = treatments[treatments["eventType"].isin(INSULIN_EVENT_TYPES) & treatments["insulin"].notna()]
bolus = treatments[treatments["eventType"].isin(INSULIN_EVENT_TYPES) & treatments["insulin"].notna()].copy()
if bolus.empty:
return {"total_units": 0.0, "count": 0, "by_type": {}}
by_type = bolus.groupby("eventType")["insulin"].sum().round(1).to_dict()
is_smb = (bolus["eventType"] == "Correction Bolus") & (bolus.get("type", "") == SMB_TYPE)
bolus["category"] = bolus["eventType"].where(~is_smb, SMB_TYPE)
by_type = bolus.groupby("category")["insulin"].sum().round(1).to_dict()
return {
"total_units": round(bolus["insulin"].sum(), 1),
"count": len(bolus),
@@ -303,6 +440,35 @@ def bolus_stats(treatments: pd.DataFrame) -> dict:
}
def basal_stats(treatments: pd.DataFrame, period_days: float) -> dict:
"""Totale basaalinsuline uit Temp Basal-treatments (rate x duur), zonder pompprofiel.
AndroidAPS zet in gesloten-lus-modus vrijwel continu Temp Basal-treatments
met een `absolute` snelheid (E/uur) en `duration` (minuten); de som van
afgeleverde eenheden daaruit benadert de daadwerkelijke basaaldosis
nauwkeurig genoeg zonder het pompprofiel te hoeven ophalen (gitea #20).
Periodes zonder actieve temp basal (dan geldt het profiel-basaaltempo)
worden hierin gemist — dit is dus een ondergrens, geen exacte totale-
dagdosis.
"""
if treatments.empty or period_days <= 0:
return {"total_units": 0.0, "avg_units_per_day": 0.0}
temp_basal = treatments[
(treatments["eventType"] == BASAL_EVENT_TYPE)
& treatments["absolute"].notna()
& treatments["duration"].notna()
]
if temp_basal.empty:
return {"total_units": 0.0, "avg_units_per_day": 0.0}
total_units = (temp_basal["absolute"] * temp_basal["duration"] / 60).sum()
return {
"total_units": round(total_units, 1),
"avg_units_per_day": round(total_units / period_days, 1),
}
SITE_INTERVAL_COLUMNS = [
"start",
"end",
+2
View File
@@ -3,5 +3,7 @@ pandas>=2.2,<3.0
requests>=2.32,<3.0
plotly>=5.24,<6.0
python-dotenv>=1.0,<2.0
matplotlib>=3.9,<4.0
fpdf2>=2.7,<3.0
pytest>=8.0,<9.0
ruff>=0.6,<1.0
+245
View File
@@ -90,6 +90,38 @@ def test_fetch_treatments_fills_missing_columns(monkeypatch):
assert result.iloc[0]["eventType"] == "Site Change"
assert pd_isna(result.iloc[0]["insulin"])
assert pd_isna(result.iloc[0]["carbs"])
assert result.iloc[0]["type"] == ""
assert pd_isna(result.iloc[0]["absolute"])
assert pd_isna(result.iloc[0]["duration"])
def test_fetch_treatments_passes_through_type_and_temp_basal_fields(monkeypatch):
"""SMB-classificatie (reporting.bolus_stats) en basaalberekening
(reporting.basal_stats) hebben deze velden nodig — zie gitea #20."""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
now = datetime.now(timezone.utc)
payload = [
{
"created_at": now.isoformat(),
"eventType": "Correction Bolus",
"insulin": 0.5,
"type": "SMB",
},
{
"created_at": now.isoformat(),
"eventType": "Temp Basal",
"absolute": 1.2,
"duration": 30,
},
]
monkeypatch.setattr(ns.requests, "get", lambda *a, **k: FakeResponse(payload))
result = ns.fetch_treatments(hours=24)
smb_row = result[result["eventType"] == "Correction Bolus"].iloc[0]
basal_row = result[result["eventType"] == "Temp Basal"].iloc[0]
assert smb_row["type"] == "SMB"
assert basal_row["absolute"] == pytest.approx(1.2)
assert basal_row["duration"] == pytest.approx(30)
def pd_isna(value):
@@ -134,3 +166,216 @@ def test_fetch_treatments_range_filters_to_date_boundaries(monkeypatch):
result = ns.fetch_treatments_range(date(2026, 1, 1), date(2026, 1, 31))
assert len(result) == 1
assert result.iloc[0]["insulin"] == pytest.approx(4.0)
def _entry(dt, sgv=100):
return {"date": int(dt.timestamp() * 1000), "sgv": sgv, "dateString": dt.isoformat()}
def test_paginated_get_stops_when_a_short_page_is_returned(monkeypatch):
"""Nightscout kan minder dan _PAGE_SIZE teruggeven omdat er simpelweg niet meer data is.
Dit was precies de bug: een enkel groot count-verzoek werd stilzwijgend
afgeknipt. Met paginering moet een pagina die kleiner is dan de gevraagde
paginagrootte de paginering laten stoppen — dat betekent "geen oudere
data beschikbaar", niet "er ging iets mis".
"""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
now = datetime.now(timezone.utc)
payload = [_entry(now - timedelta(minutes=5 * i)) for i in range(5)] # << _PAGE_SIZE
calls = []
monkeypatch.setattr(
ns.requests,
"get",
lambda *a, **k: (calls.append(k), FakeResponse(payload))[1],
)
result = ns._paginated_get(
"/api/v1/entries/sgv.json", now - timedelta(days=1), None, ns.ENTRIES_CURSOR_FIELD
)
assert len(result) == 5
assert len(calls) == 1 # één pagina was genoeg, geen verdere paginering nodig
def test_paginated_get_walks_back_multiple_pages_until_since_is_reached(monkeypatch):
"""De kern-regressietest voor de bug: data die verder terugligt dan één
pagina (~1000 documenten) moet toch volledig worden opgehaald door
meerdere pagina's met find[date][$lte] te doorlopen."""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
now = datetime.now(timezone.utc)
# 3 volle pagina's (elk _PAGE_SIZE groot, 5 minuten uit elkaar, aflopend in tijd)
# + een laatste kleinere pagina die de paginering laat stoppen.
page_size = ns._PAGE_SIZE
pages = []
cursor = now
for _ in range(3):
page = [_entry(cursor - timedelta(minutes=5 * i)) for i in range(page_size)]
pages.append(page)
cursor = cursor - timedelta(minutes=5 * (page_size - 1)) - timedelta(minutes=5)
pages.append([_entry(cursor - timedelta(minutes=5 * i)) for i in range(10)])
call_count = {"n": 0}
def fake_get(*_a, **_k):
response = FakeResponse(pages[call_count["n"]])
call_count["n"] += 1
return response
monkeypatch.setattr(ns.requests, "get", fake_get)
since = now - timedelta(days=365) # ruim voorbij alle test-pagina's
result = ns._paginated_get(
"/api/v1/entries/sgv.json", since, None, ns.ENTRIES_CURSOR_FIELD
)
assert call_count["n"] == 4 # 3 volle pagina's + 1 korte slotpagina
assert len(result) == 3 * page_size + 10
def test_paginated_get_stops_at_since_boundary_without_fetching_further(monkeypatch):
"""Zodra een pagina data ouder dan `since` bevat, hoeft er geen volgende
pagina meer opgehaald te worden, ook al was de pagina zelf vol."""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
now = datetime.now(timezone.utc)
page_size = ns._PAGE_SIZE
# deze ene pagina spant ruim meer dan een dag, dus bevat al data voorbij `since`
page = [_entry(now - timedelta(minutes=5 * i)) for i in range(page_size)]
calls = {"n": 0}
def fake_get(*_a, **_k):
calls["n"] += 1
return FakeResponse(page)
monkeypatch.setattr(ns.requests, "get", fake_get)
since = now - timedelta(hours=6) # ruim binnen de eerste pagina
result = ns._paginated_get(
"/api/v1/entries/sgv.json", since, None, ns.ENTRIES_CURSOR_FIELD
)
assert calls["n"] == 1
assert len(result) == page_size # filtering op since gebeurt pas in _entries_since, niet hier
def test_fetch_entries_range_retrieves_data_spanning_multiple_pages(monkeypatch):
"""End-to-end regressietest: een 90-dagenrapport moet echt 90 dagen data
krijgen, niet stilzwijgend worden afgeknipt tot de meest recente pagina."""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
monkeypatch.delenv("NIGHTSCOUT_TOKEN", raising=False)
monkeypatch.delenv("NIGHTSCOUT_API_SECRET", raising=False)
end_date = date(2026, 3, 31)
start_date = date(2026, 1, 1) # 90 dagen
# simuleer Nightscout: elke 5 minuten een meting over de volle periode,
# opgeknipt in door de server afgedwongen pagina's van _PAGE_SIZE
until = datetime.combine(end_date, datetime.max.time(), tzinfo=timezone.utc)
since = datetime.combine(start_date, datetime.min.time(), tzinfo=timezone.utc)
all_entries = []
t = until
while t >= since:
all_entries.append(_entry(t))
t -= timedelta(minutes=5)
page_size = ns._PAGE_SIZE
def fake_get(*_args, **kwargs):
cursor_ms = kwargs["params"].get("find[date][$lte]")
pool = all_entries if cursor_ms is None else [e for e in all_entries if e["date"] <= cursor_ms]
return FakeResponse(pool[:page_size])
monkeypatch.setattr(ns.requests, "get", fake_get)
result = ns.fetch_entries_range(start_date, end_date)
assert len(result) == len(all_entries)
assert result["timestamp"].min().date() == start_date
assert result["timestamp"].max().date() == end_date
def _treatment(dt, event_type="Temp Basal"):
return {
"created_at": dt.strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3] + "Z",
"date": int(dt.timestamp() * 1000),
"eventType": event_type,
}
def test_fetch_treatments_range_retrieves_full_history_even_with_a_hidden_server_side_window(
monkeypatch,
):
"""Regressietest voor de echte productiebug (2026-07-31): treatments.json
filtert/sorteert intern op `created_at` (string), niet op `date`
(numeriek) zoals entries. Filteren op `date` — of helemaal geen filter
meegeven — leverde stilzwijgend hooguit de laatste ~4 dagen op, ook al
bevatten de documenten een geldig `date`-veld en stond de rest van de
geschiedenis (maanden) gewoon in de database (geverifieerd rechtstreeks
in MongoDB: 123k+ documenten, geen TTL-index, geen capped collection).
Deze test simuleert precies dat servergedrag: een find[date]-filter (of
het ontbreken van elk filter) geeft alleen de laatste 4 dagen terug;
alleen find[created_at] geeft de volledige, langere geschiedenis.
"""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
monkeypatch.delenv("NIGHTSCOUT_TOKEN", raising=False)
monkeypatch.delenv("NIGHTSCOUT_API_SECRET", raising=False)
end_date = date(2026, 7, 31)
start_date = date(2026, 5, 1) # 92 dagen, ruim voorbij het 4-dagen-venster
until = datetime.combine(end_date, datetime.max.time(), tzinfo=timezone.utc)
since = datetime.combine(start_date, datetime.min.time(), tzinfo=timezone.utc)
all_treatments = []
t = until
while t >= since:
all_treatments.append(_treatment(t))
t -= timedelta(minutes=30)
hidden_window_start = until - timedelta(days=4)
page_size = ns._PAGE_SIZE
def fake_get(*_args, **kwargs):
params = kwargs["params"]
cursor = params.get("find[created_at][$lte]")
if cursor is None:
# geen (of het verkeerde) filter -> verborgen 4-dagenvenster
pool = [e for e in all_treatments if e["created_at"] >= hidden_window_start.strftime(
"%Y-%m-%dT%H:%M:%S.%f"
)[:-3] + "Z"]
else:
pool = [e for e in all_treatments if e["created_at"] <= cursor]
pool = sorted(pool, key=lambda e: e["created_at"], reverse=True)
return FakeResponse(pool[:page_size])
monkeypatch.setattr(ns.requests, "get", fake_get)
result = ns.fetch_treatments_range(start_date, end_date)
assert len(result) == len(all_treatments)
assert result["timestamp"].min().date() == start_date
assert result["timestamp"].max().date() == end_date
def test_paginated_get_uses_created_at_field_for_treatments_cursor(monkeypatch):
"""Losstaande, gerichte assertie op het daadwerkelijke queryparameter-veld
— moet `find[created_at][$lte]` zijn, nooit `find[date][$lte]`, voor het
treatments-endpoint."""
monkeypatch.setenv("NIGHTSCOUT_URL", "https://example.com")
now = datetime.now(timezone.utc)
payload = [_treatment(now - timedelta(minutes=30 * i)) for i in range(5)]
seen_params = []
def fake_get(*_a, **kwargs):
seen_params.append(kwargs["params"])
return FakeResponse(payload)
monkeypatch.setattr(ns.requests, "get", fake_get)
ns._paginated_get(
"/api/v1/treatments.json", now - timedelta(days=1), None, ns.TREATMENTS_CURSOR_FIELD
)
assert "find[created_at][$lte]" in seen_params[0]
assert "find[date][$lte]" not in seen_params[0]
+92
View File
@@ -0,0 +1,92 @@
from datetime import date
import pandas as pd
from pdf_export import (
DAGOVERZICHT_MAX_DAYS,
SECTION_AGP,
SECTION_DAGOVERZICHT,
SECTION_EVENTS,
SECTION_SUMMARY,
build_report_pdf,
)
def _sample_glucose():
timestamps = pd.date_range("2026-01-01", periods=48, freq="30min")
return pd.DataFrame({"timestamp": timestamps, "mmol": [6.0 + (i % 5) for i in range(48)]})
def _glucose_with_hypo_and_hyper_events():
"""3 dagen data met een echte hypo- (<3.9, >=15 min) en hyperepisode (>10.0)."""
timestamps = pd.date_range("2026-01-01", periods=3 * 24 * 12, freq="5min")
values = [7.0] * len(timestamps)
values[10:15] = [3.0] * 5 # 25 min hypo
values[100:106] = [12.0] * 6 # 30 min hyper
return pd.DataFrame({"timestamp": timestamps, "mmol": values})
def test_build_report_pdf_produces_valid_pdf_bytes():
glucose = _sample_glucose()
events_summary = {
"Hypo Level 1": {"count": 2, "avg_duration_minutes": 20.0, "events_per_week": 1.0},
}
pdf_bytes = build_report_pdf(
[SECTION_SUMMARY, SECTION_AGP, SECTION_EVENTS],
glucose,
events_summary,
date(2026, 1, 1),
date(2026, 1, 2),
)
assert isinstance(pdf_bytes, bytes)
assert pdf_bytes.startswith(b"%PDF")
assert len(pdf_bytes) > 500
def test_build_report_pdf_handles_empty_glucose():
empty = pd.DataFrame(columns=["timestamp", "mmol"])
pdf_bytes = build_report_pdf(
[SECTION_SUMMARY, SECTION_AGP], empty, {}, date(2026, 1, 1), date(2026, 1, 2)
)
assert pdf_bytes.startswith(b"%PDF")
def test_build_report_pdf_handles_no_sections_selected():
glucose = _sample_glucose()
pdf_bytes = build_report_pdf([], glucose, {}, date(2026, 1, 1), date(2026, 1, 2))
assert pdf_bytes.startswith(b"%PDF")
def test_build_report_pdf_includes_dagoverzicht_section():
glucose = _sample_glucose()
pdf_bytes = build_report_pdf(
[SECTION_DAGOVERZICHT], glucose, {}, date(2026, 1, 1), date(2026, 1, 2)
)
assert pdf_bytes.startswith(b"%PDF")
assert len(pdf_bytes) > 500
def test_build_report_pdf_caps_dagoverzicht_at_max_days():
timestamps = pd.date_range("2026-01-01", periods=(DAGOVERZICHT_MAX_DAYS + 10) * 24, freq="1h")
glucose = pd.DataFrame({"timestamp": timestamps, "mmol": [7.0] * len(timestamps)})
pdf_bytes = build_report_pdf(
[SECTION_DAGOVERZICHT],
glucose,
{},
timestamps[0].date(),
timestamps[-1].date(),
)
assert pdf_bytes.startswith(b"%PDF")
def test_build_report_pdf_draws_hypo_and_hyper_charts_when_events_present():
glucose = _glucose_with_hypo_and_hyper_events()
events_summary = {
"Hypo Level 1": {"count": 1, "avg_duration_minutes": 25.0, "events_per_week": 2.3},
"Hyper Level 1": {"count": 1, "avg_duration_minutes": 30.0, "events_per_week": 2.3},
}
pdf_bytes = build_report_pdf(
[SECTION_EVENTS], glucose, events_summary, date(2026, 1, 1), date(2026, 1, 4)
)
assert pdf_bytes.startswith(b"%PDF")
assert len(pdf_bytes) > 500
+167 -1
View File
@@ -1,12 +1,20 @@
from datetime import date
import pandas as pd
import pytest
from reporting import (
GMI_TREND_COLUMNS,
agp_percentile_bands,
basal_stats,
bolus_stats,
completeness_warning,
daily_gmi,
day_of_week_pattern,
detect_events,
gmi_mmol_per_mol,
gmi_percent,
gmi_trend,
minutes_per_day,
overnight_daytime_split,
percent_cv,
@@ -28,6 +36,63 @@ def test_gmi_percent_matches_bergenstal_formula():
assert gmi_percent(7.0) == pytest.approx(6.3, abs=0.05)
def test_gmi_mmol_per_mol_matches_known_ifcc_reference_values():
# 6.5% is de bekende diagnostische drempel = 48 mmol/mol (IFCC).
assert gmi_mmol_per_mol(6.5) == 48
assert gmi_mmol_per_mol(7.0) == 53
def test_daily_gmi_aggregates_per_calendar_day():
glucose = _series(
[
("2026-01-01 08:00", 6.0),
("2026-01-01 20:00", 8.0),
("2026-01-02 08:00", 7.0),
]
)
result = daily_gmi(glucose)
assert list(result["date"].astype(str)) == ["2026-01-01", "2026-01-02"]
assert result["gmi"].iloc[0] == pytest.approx(gmi_percent(7.0)) # gemiddelde van 6.0 en 8.0
assert result["gmi"].iloc[1] == pytest.approx(gmi_percent(7.0))
def test_daily_gmi_handles_empty_input():
empty = pd.DataFrame(columns=["timestamp", "mmol"])
result = daily_gmi(empty)
assert result.empty
assert list(result.columns) == ["date", "gmi"]
def test_gmi_trend_handles_empty_input():
empty = pd.DataFrame(columns=["timestamp", "mmol"])
result = gmi_trend(empty)
assert result.empty
assert list(result.columns) == GMI_TREND_COLUMNS
def test_gmi_trend_returns_no_line_for_a_single_day():
glucose = _series([("2026-01-01 08:00", 6.0)])
result = gmi_trend(glucose)
assert list(result.columns) == GMI_TREND_COLUMNS
assert len(result) == 1
assert pd.isna(result["trend"].iloc[0])
def test_gmi_trend_fits_increasing_linear_trend_with_band_around_it():
pairs = [(f"2026-01-{i + 1:02d} 08:00", 6.0 + i * 0.2) for i in range(10)]
glucose = _series(pairs)
result = gmi_trend(glucose, rolling_window_days=3)
assert len(result) == 10
assert result["trend"].iloc[-1] > result["trend"].iloc[0]
assert (result["trend_lower"] <= result["trend"]).all()
assert (result["trend"] <= result["trend_upper"]).all()
# Vrijwel lineaire input (gmi_percent rondt af op 1 decimaal, dus een
# kleine restfout blijft over) -> smalle betrouwbaarheidsband.
assert (result["trend_upper"] - result["trend_lower"]).max() == pytest.approx(0, abs=0.15)
assert result["rolling_avg"].iloc[-1] > result["rolling_avg"].iloc[0]
def test_percent_cv_handles_empty_input():
empty = pd.DataFrame(columns=["timestamp", "mmol"])
assert percent_cv(empty) == 0.0
@@ -93,6 +158,42 @@ def test_sensor_active_percent_handles_gaps():
assert result == pytest.approx(50.0, abs=1.0)
def test_completeness_warning_none_for_empty_glucose():
empty = pd.DataFrame(columns=["timestamp", "mmol"])
assert completeness_warning(empty, "2026-01-01", "2026-01-14") == "Geen glucosedata gevonden in deze periode."
def test_completeness_warning_none_when_period_is_fully_covered():
# 14 volle dagen aan 5-minuten-metingen, precies de gekozen periode
timestamps = pd.date_range("2026-01-01 00:00", "2026-01-14 23:55", freq="5min")
glucose = pd.DataFrame({"timestamp": timestamps, "mmol": [6.0] * len(timestamps)})
assert completeness_warning(glucose, date(2026, 1, 1), date(2026, 1, 14)) is None
def test_completeness_warning_flags_missing_days_at_start_of_period():
# gevraagd: 1 t/m 14 januari, maar data begint pas op 11 januari (de
# exacte bug: alleen de laatste ~4 dagen van een langere periode zijn
# opgehaald)
timestamps = pd.date_range("2026-01-11 00:00", "2026-01-14 23:55", freq="5min")
glucose = pd.DataFrame({"timestamp": timestamps, "mmol": [6.0] * len(timestamps)})
message = completeness_warning(glucose, date(2026, 1, 1), date(2026, 1, 14))
assert message is not None
assert "11-01-2026" in message
assert "10 dag" in message # 1 t/m 10 januari ontbreken volledig = 10 dagen
assert "sensorhiaten" in message
def test_completeness_warning_flags_low_coverage_without_start_gap():
# data start op tijd, maar met veel verspreide hiaten -> lage dekking
# zonder dat er een aaneengesloten gat aan het begin is
timestamps = pd.date_range("2026-01-01 00:00", periods=20, freq="5min")
glucose = pd.DataFrame({"timestamp": timestamps, "mmol": [6.0] * 20})
message = completeness_warning(glucose, date(2026, 1, 1), date(2026, 1, 14))
assert message is not None
assert "sensordekking" in message
assert "sensorhiaten" not in message # dit is de generieke, niet de "ontbreekt volledig"-boodschap
def test_detect_events_ignores_dip_shorter_than_minimum_duration():
# dip van 10 minuten (< 15 min drempel) mag niet meetellen
glucose = _series(
@@ -300,6 +401,7 @@ def test_bolus_stats_splits_by_event_type():
"eventType": ["Meal Bolus", "Correction Bolus", "Site Change"],
"insulin": [4.0, 1.5, None],
"carbs": [45.0, None, None],
"type": ["NORMAL", "NORMAL", ""],
}
)
result = bolus_stats(treatments)
@@ -308,12 +410,76 @@ def test_bolus_stats_splits_by_event_type():
assert result["by_type"] == {"Meal Bolus": 4.0, "Correction Bolus": 1.5}
def test_bolus_stats_reclassifies_smb_correction_boluses_separately():
"""AndroidAPS registreert een automatische SMB-microbolus als
eventType == "Correction Bolus" met type == "SMB" — dit moet apart van
een handmatige correctiebolus (type != "SMB") geteld worden (gitea #20)."""
treatments = pd.DataFrame(
{
"timestamp": pd.to_datetime(["2026-01-01 08:00", "2026-01-01 12:00"]),
"eventType": ["Correction Bolus", "Correction Bolus"],
"insulin": [0.3, 0.9],
"carbs": [None, None],
"type": ["SMB", "NORMAL"],
}
)
result = bolus_stats(treatments)
assert result["by_type"] == {"SMB": 0.3, "Correction Bolus": 0.9}
def test_bolus_stats_treats_missing_type_column_as_non_smb():
treatments = pd.DataFrame(
{
"timestamp": pd.to_datetime(["2026-01-01 08:00"]),
"eventType": ["Correction Bolus"],
"insulin": [0.9],
"carbs": [None],
}
)
result = bolus_stats(treatments)
assert result["by_type"] == {"Correction Bolus": 0.9}
def test_bolus_stats_handles_empty_input():
empty = pd.DataFrame(columns=["timestamp", "eventType", "insulin", "carbs"])
empty = pd.DataFrame(columns=["timestamp", "eventType", "insulin", "carbs", "type"])
result = bolus_stats(empty)
assert result == {"total_units": 0.0, "count": 0, "by_type": {}}
def test_basal_stats_sums_temp_basal_rate_times_duration():
treatments = pd.DataFrame(
{
"timestamp": pd.to_datetime(["2026-01-01 08:00", "2026-01-01 08:30"]),
"eventType": ["Temp Basal", "Temp Basal"],
"absolute": [1.2, 0.6],
"duration": [30, 60],
}
)
# 1.2 E/u * 0.5u + 0.6 E/u * 1u = 0.6 + 0.6 = 1.2 E over 1 dag
result = basal_stats(treatments, period_days=1)
assert result["total_units"] == pytest.approx(1.2)
assert result["avg_units_per_day"] == pytest.approx(1.2)
def test_basal_stats_ignores_non_temp_basal_rows_and_missing_fields():
treatments = pd.DataFrame(
{
"timestamp": pd.to_datetime(["2026-01-01 08:00", "2026-01-01 08:30"]),
"eventType": ["Meal Bolus", "Temp Basal"],
"absolute": [None, None],
"duration": [None, 30],
}
)
result = basal_stats(treatments, period_days=1)
assert result == {"total_units": 0.0, "avg_units_per_day": 0.0}
def test_basal_stats_handles_empty_input():
empty = pd.DataFrame(columns=["timestamp", "eventType", "absolute", "duration"])
result = basal_stats(empty, period_days=1)
assert result == {"total_units": 0.0, "avg_units_per_day": 0.0}
def test_site_interval_stats_splits_by_site_change():
glucose = _series(
[
+30 -1
View File
@@ -1,4 +1,12 @@
from ui_common import hex_to_rgba
import plotly.graph_objects as go
from ui_common import (
PALETTE,
THRESHOLD_LINES,
add_threshold_lines,
build_tir_stacked_bar_figure,
hex_to_rgba,
)
def test_hex_to_rgba_converts_correctly():
@@ -7,3 +15,24 @@ def test_hex_to_rgba_converts_correctly():
def test_hex_to_rgba_handles_hash_optional():
assert hex_to_rgba("2a78d6", 0.3) == "rgba(42,120,214,0.3)"
def test_add_threshold_lines_adds_one_line_per_clinical_threshold():
fig = go.Figure()
add_threshold_lines(fig, PALETTE["light"])
assert len(fig.layout.shapes) == len(THRESHOLD_LINES)
def test_add_threshold_lines_targets_subplot_row_when_given():
fig = go.Figure()
fig.add_trace(go.Scatter(x=[0], y=[0]))
add_threshold_lines(fig, PALETTE["light"], row=1)
assert len(fig.layout.shapes) == len(THRESHOLD_LINES)
def test_build_tir_stacked_bar_figure_has_one_bar_trace_per_band():
tir = {"very_low": 1.0, "low": 3.0, "in_range": 80.0, "high": 12.0, "very_high": 4.0}
fig = build_tir_stacked_bar_figure(tir, PALETTE["light"])
assert len(fig.data) == 5
assert [trace.y[0] for trace in fig.data] == [1.0, 3.0, 80.0, 12.0, 4.0]
assert fig.layout.barmode == "stack"
+104 -1
View File
@@ -15,7 +15,15 @@ import plotly.graph_objects as go
import streamlit as st
from nightscout_client import fetch_entries_range, fetch_treatments_range
from reporting import detect_events, summarize_events
from reporting import (
HIGH,
LOW,
VERY_HIGH,
VERY_LOW,
completeness_warning,
detect_events,
summarize_events,
)
MMOL_HOVERTEMPLATE = "%{x|%d-%m %H:%M} · %{y:.1f} mmol/L<extra></extra>"
@@ -41,6 +49,14 @@ PALETTE = {
"muted": "#898781",
"text": "#52514e",
"good": "#0ca30c",
# Losstaand van "low"/"high" hierboven (die blijven gedeeld tussen
# andere grafieken): de 5-bands TIR-grafiek (gitea #14) heeft juist
# zichtbaar onderscheid tussen laag/zeer-laag en hoog/zeer-hoog nodig.
"tir_very_low": "#8b1a1a",
"tir_low": "#d03b3b",
"tir_in_range": "#0ca30c",
"tir_high": "#fab219",
"tir_very_high": "#c76f0a",
},
"dark": {
"glucose": "#3987e5",
@@ -55,9 +71,24 @@ PALETTE = {
"muted": "#898781",
"text": "#c3c2b7",
"good": "#0ca30c",
"tir_very_low": "#a32626",
"tir_low": "#d95d5d",
"tir_in_range": "#0ca30c",
"tir_high": "#fab219",
"tir_very_high": "#d9852e",
},
}
# Klinische drempellijnen (Battelino et al. 2019), gedeeld tussen de AGP- en
# Dagoverzicht-pagina's (FR3, FR4 — gitea #15) zodat beide grafieken exact
# dezelfde 4 lijnen op dezelfde manier tonen.
THRESHOLD_LINES = [
(VERY_LOW, "3.0"),
(LOW, "3.9"),
(HIGH, "10.0"),
(VERY_HIGH, "13.9"),
]
def hex_to_rgba(hex_color: str, alpha: float) -> str:
"""Zet een '#rrggbb'-kleur om naar een 'rgba(r,g,b,a)'-string voor Plotly-fills."""
@@ -98,6 +129,78 @@ def load_range_data(start: date, end: date):
return glucose, treatments
def add_threshold_lines(fig: go.Figure, colors: dict, row: int | None = None) -> None:
"""Teken de 4 klinische drempellijnen (3.0/3.9/10.0/13.9 mmol/L) op een grafiek.
Gedeeld door de AGP- en Dagoverzicht-pagina's (gitea #15). `row` wordt
doorgegeven aan `add_hline` voor subplot-figuren (Dagoverzicht); voor een
los-staande figuur (AGP) blijft dit `None`.
"""
kwargs = {} if row is None else {"row": row, "col": 1}
for value, label in THRESHOLD_LINES:
fig.add_hline(
y=value,
line={"color": colors["muted"], "width": 1, "dash": "dot"},
annotation_text=label,
annotation_position="right",
annotation_font={"size": 9, "color": colors["muted"]},
**kwargs,
)
TIR_BAND_ORDER = ["very_low", "low", "in_range", "high", "very_high"]
TIR_BAND_LABELS = {
"very_low": "Zeer laag",
"low": "Laag",
"in_range": "In bereik",
"high": "Hoog",
"very_high": "Zeer hoog",
}
def build_tir_stacked_bar_figure(tir: dict, colors: dict) -> go.Figure:
"""Bouw de 5-bands tijd-in-bereik als één gestapelde staafgrafiek (gitea #14).
Eén staaf, geen x-as-label, percentage op de y-as; van onder naar boven
"Zeer laag" -> "Zeer hoog" — dezelfde volgorde als de klinische banden.
"""
fig = go.Figure()
for band in TIR_BAND_ORDER:
fig.add_trace(
go.Bar(
x=[""],
y=[tir[band]],
name=TIR_BAND_LABELS[band],
marker_color=colors[f"tir_{band}"],
hovertemplate=f"{TIR_BAND_LABELS[band]}: %{{y:.0f}}%<extra></extra>",
)
)
fig.update_layout(
barmode="stack",
paper_bgcolor="rgba(0,0,0,0)",
plot_bgcolor="rgba(0,0,0,0)",
font_color=colors["text"],
height=140,
margin={"l": 10, "r": 10, "t": 10, "b": 10},
yaxis={"title": "%", "range": [0, 100], "gridcolor": colors["grid"]},
xaxis={"visible": False},
legend={"orientation": "h", "yanchor": "bottom", "y": 1.02, "xanchor": "left", "x": 0},
)
return fig
def render_completeness_warning(glucose: pd.DataFrame, start: date, end: date) -> None:
"""Toon reporting.completeness_warning()'s oordeel als st.warning, indien nodig.
Gebruikt door elke rapportagepagina, direct na de load_range_data-call.
De beoordelingslogica zelf staat in reporting.py (pure functie) zodat het
PDF-rapport (pdf_export.py) exact dezelfde disclaimer kan hergebruiken.
"""
message = completeness_warning(glucose, start, end)
if message:
st.warning(message)
def render_event_level(
glucose: pd.DataFrame,
threshold: float,