upload-artifact@v4/download-artifact@v4 use the newer Artifact API v2,
which Gitea's act_runner doesn't support yet (GHESNotSupportedError).
v3 uses the older, Gitea-compatible API.
pytest-cov + pyproject.toml [tool.coverage.*] measure reporting.py,
nightscout_client.py, data_processing.py, pdf_export.py, profile_data.py,
ui_common.py, and auth.py; app.py and pages/*.py are omitted (thin
Streamlit UI scripts, not meaningfully unit-testable without a running
Streamlit session — the handful of similarly UI-coupled functions left
in ui_common.py/auth.py are marked "# pragma: no cover" with the same
rationale, since the logic they call into is already fully tested).
The Gitea Actions test job now runs with --cov and fails the build
below 80%; the coverage XML is uploaded as a build artifact and fed
into the SonarQube scan (closing the gap flagged in
homelab-cicd-werkinstructie.md). Backfilled a few missing branches
(get_colors() theme fallback, secondary-y-axis overlay traces) to reach
98% on the measured modules — comfortable headroom above the 80% gate.
Builds on #29/#30/#31.
Every report section now gets its own landscape A4 page (or, for
Dagoverzicht, one page per day) via a GlucoseReportPDF(FPDF) subclass
with a real header (section title, report version, insulin/pump/sensor
type from new REPORT_* env vars) and footer (page number, generation
timestamp, Nightscout link) on every page, plus an original SVG icon
in the header — inspired by the LibreView AGP report.
New "Insulineprofiel" PDF section mirrors the dashboard page (basal
bar chart, ISF/carb-ratio segments, DIA). Each Dagoverzicht day now
draws the same basal/bolus/carb/target overlays as the interactive
charts (matplotlib port of the Plotly helpers from #30) plus a new
per-day stats block (reporting.daily_stats: avg/min/max glucose,
total/basal/bolus/SMB insulin, hypo/hyper counts and durations).
Builds on #29 (profile data) and #30 (overlay logic/colors).
pd.Timedelta(days=1) triggers a DeprecationWarning about NumPy's
"generic" timedelta unit in this pandas version; pd.Timedelta(1, unit="D")
is equivalent and warning-free.
Overlay the full insulin picture onto the interactive glucose charts:
basal rate as a semi-transparent step-area (profile schedule overridden
by active temp basal), bolus doses as size-scaled dots split into
food/correction/SMB categories (food and correction labeled with units,
SMB left unlabeled), carbs as bars, and the pump's target glucose as a
line — each with its own color, alpha, and legend entry. Shared trace-
building helpers live in ui_common.py so both pages (and the PDF report
in a follow-up PR) render identically.
Builds on the pump profile work in #29 (profile_data.py, load_profile()).
Fetch and parse the latest-used Nightscout profile (basal schedule, ISF,
carb ratio, DIA, targets) and show it as a new dashboard page: a list of
basal/ISF/carb-ratio segments plus a bar chart of basal rate by hour of
day. This is the data foundation for treatment overlays on the glucose
charts and the redesigned PDF report in follow-up PRs.
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.
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.
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.
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).
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.
_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.
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]>
reporting.py gains overnight_daytime_split, day_of_week_pattern,
bolus_stats, and site_interval_stats. The last two are explicitly scoped
to bolus-only insulin (no basal reconstruction) since that needs the
Nightscout profile, which is the separate profile-tuning feature's
territory — documented as a caveat in both the UI and the design doc.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Dagoverzicht stacks one glucose trace per calendar day (small multiples
via a shared 0-24h x-axis). The two event pages share a new
render_event_level() helper in ui_common.py so Level 1/Level 2 hypo and
hyper are rendered identically: count/avg/longest/events-per-week
metrics, a time-of-day histogram, and a table of individual events.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Wires reporting.py's stats into two new pages: a summary panel (data
completeness, GMI, %CV, 5-band time-in-range vs. consensus targets) and
the standard AGP percentile chart (5th-95th and 25th-75th bands plus
median, aggregated by time-of-day across the selected period).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
First increment of the clinical reporting feature (see
docs/requirements-reporting.md and docs/design-reporting.md): moves the
existing live view to pages/1_Live.py, adds a password gate since the
dashboard URL will be shared with the diabetes team, adds
fetch_entries_range/fetch_treatments_range for multi-day report queries,
and adds reporting.py with the core clinical calculations (GMI, %CV,
5-band time-in-range, hypo/hyper event detection per the Battelino et al.
2019 persistence rule). No report pages wired up yet — that's next.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
@master resolves to a version that installs Trivy via a nested
setup-trivy composite action, which does an actions/checkout against
github.com with a token. On Gitea's act_runner that fails with "Input
required and not supplied: token" even when the token is explicitly
set to an empty string (act_runner doesn't distinguish empty from
unset). v0.26.0 introduced that composite/setup-trivy split; v0.25.0
is the last release that ships as a self-contained Docker image with
Trivy already baked in, so it needs no token at all.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
gitea.zb.local resolves but is plain HTTP, which Docker refuses as a
registry without daemon-level insecure-registry config. gitea.zb.lan
resolves but nothing answers there yet (proxy not configured for it).
gitea.zb-kw.duckdns.org is confirmed reachable end-to-end over HTTPS,
including the registry v2 API, from both the runner and the deploy
target, so use that consistently. Also fix aquasecurity/trivy-action's
token-setup-trivy input, which defaults to gitea.token and gets
rejected by github.com when it tries to clone the install script.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
CI invokes the pytest console script directly (pytest tests/ -v), which
unlike `python -m pytest` does not add the repo root to sys.path. That
broke imports of the top-level data_processing/nightscout_client
modules from the test files. Add pythonpath = ["."] so it works
regardless of invocation style.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
gitea.thuis.local doesn't resolve on this network; the real internal
Gitea hostname is gitea.zb.local. Also load .env via python-dotenv so
the app works the same under PowerShell as bash, and add .dockerignore
so .env never ends up baked into the image.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
willemjan
