- event-builder: called at every batch, reads the whole analysis and
decides on its own, against the event base, what is new; an event
keeps the bounds of the event, not of the batch (plus 6 h).
- the "Nouveautés" section leaves the analysis and the upgrader's
output; the summary lists contradictions with the base.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- analyste: the most precise analysis of the batch, detectors allowed as
a tool; reading, labels, known events, novelties in analyse.md; skill
critiquer-lot renamed analyser-lot.
- detectors-critic: compares the detectors to the analysis event by
event, without looking at the data; verdict and at most two
suggestions in critique.md, read by the detectors-upgrader.
- /boucle step 2, builder, review page, /new-event and score.py import
follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- detectors-upgrader: makes the detectors better on the batch from the
analyst's critique; adds, removes or modifies, removing first; targets
the significant events missed or badly bounded, not a perfect match.
- detectors-control: the guard against overfitting; never looks at the
batch, judges the diff and the benchmark on the event base, decides,
and sends its remarks back to the upgrader.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The event base is the analyst's knowledge base, read before or during
its reading; a known event is a lead to check on the data, not a
conclusion. No measurement of the analyst's progress.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Catch-up: E-004, E-006, E-007, E-008, E-010, E-011 in the map; new
section 5 "Faux amis"; rejected E-009 no longer cited.
- /review-event and /new-event: what the event teaches goes into the
map at validation, shown to the expert with the fiche.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- read-event: map and index first, then the card of close events only,
then their data with the reading tools; never all cards at once.
Diagnostic agent, critiquer-lot and builder point to it; the index
header too. Signature length: about 250 characters, a guideline.
- "Cas" becomes "event": events/E-xxx/event.md, skills new-event,
review-event, anonymize-event.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The diagnostic agent is kept in case it is needed again, under a name that
says what it is; entry point still .claude/agents/diagnostic.md; paths updated
in skills, agents, spec, contexts, the living plan and CLAUDE.md; reports and
history keep old paths. Also: two stale tool paths in the living plan.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The reading tools leave agents/diagnostic/outils/ for a root folder of their
own; paths updated in skills, agents, detectors, scripts, case fiches,
support/, contexts and CLAUDE.md; graphique, vignettes and gif find the repo
root one level up; history keeps old paths.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The case base leaves agents/diagnostic/cas/ for a root folder of its own,
with the event map and its graph; paths updated in skills, agents, scripts,
case fiches, contexts, the living plan and CLAUDE.md; history keeps old paths.
Also: plan point 51, an analyst that improves from session to session.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The same ameliorateur agent runs after the releveur; its scope now depends
on the process named in its request.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Détecteur degradee : une cible rejetée la plupart du temps, jour après jour,
quand les autres sont vues (catégorie cible_degradee, grisée, comptée par la
boucle). Page detecteurs.py : error_value moyenne de tous les capteurs à chaque
instant, du vert au rouge, sous les barres.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- End-of-session page for /boucle: the session's curves, the analyst's
labels against the detectors' events per category, a plain score table
and the benchmark (comparaison.py, D-066)
- Detectors moved to vues-deplacements/detectors/ (D-067)
- Detectors must detect everything; what the client sees is a separate
layer (D-068)
- Benchmark: each case keeps its labels from creation (etiquettes.json),
benchmark.py scores the detectors on every case, history in
vues-deplacements/benchmark/historique.csv; run by /boucle at each lot,
after each change and at the end, read by the upgrader and validator;
labels for C-001..C-005 from one analyst pass corrected by their files,
C-006..C-008 from lot 01 (D-069)
- Fixes from the releveur's notes: /boucle arguments and other sessions,
branch name in the spec, detector name in categories.py output, learned
case header, anonymiser.py (camera/, no empty CSVs, --etiquettes),
ensemble.py (clock-aligned steps, title), plages.py (same-sign merge),
labeling request written into /nouveau-cas
- Also carries the other session's uncommitted D-065 edits in the same
files; data/, .gitignore, pomper.py and stations.py are left out
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The station manual says the DIFF image is the contrast-enhanced view of the
reflectors that the displacement algorithm measures, not a difference between
two instants. The reading tools only look at IRON images, so run.md describes
--diff as an option instead of putting it in the pump command.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Camera pan, tilt, scale, rotation and sigma0, one row per camera measurement,
read like the platform viewer (cam_movements aggregate), one CSV per day in
camera/. Tried on NORDEND 30/07: 70 poses out of 95 instants, sigma0 on the
same 7 as the advanced export. run.md holds the command to copy into the
app's dev container.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two days (29-30/07) pumped by the user in the app's dev container: system
metrics, displacements and camera images match the received exports, and
the journal gives the same events. The container mounts ~ on /workspaces,
so the script now writes straight into Support-Agent's data/ (no copies).
Add --delai for days that exceed the 60 s query timeout, and report the
root database error instead of pandas' wrapper.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each batch now leaves etiquettes.json next to the critique: the analyst's
own reading, written before the detectors' output, one entry per moment,
checked by critiquer-lot/etiquettes.py against the event map and the
station's sensors. Groundwork for a future learned model.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
pomper.py now also pumps every system metric of a station, in the format of
the System page download (hourly mean, text lists, empty hour), one CSV per
metric and per day; displacements are written one file per day, today is
never pumped, and a rerun only fetches missing days. Fix image resume: an
image already on disk now counts for its time slot. load_stations merges
several files of the same metric instead of failing on a duplicate.
Not run against the database: first run by the user, with point 27.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each /boucle subagent now ends its output with a Difficultés section;
at the end of a session the orchestrator calls the releveur once, which
locates the instruction or file at fault and records it in
docs/ameliorations-processus.md, without fixing anything.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The agent now serves one precise case, reads the project's displacement
knowledge and the whole case base, starts from the detectors' events
without judging them, and no longer has conclusions imposed on it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plan points 18 to 26, 30, 39, 40 and the point 37 housekeeping:
- greyed zones per target, vertical scale without the zones to ignore, Fit all,
mm by default with a px switch, Support view, Expand (one row per category);
- cleaned curves: cut in the zones to ignore, raw data kept there in thin grey;
- target02 to target05 count as targets; smoothing questions answered;
- categories.py sorts in slices with 3 days of history by default;
- the pumping script points to ~/monitoring-dashboard and skips diff images;
- CLAUDE.md back under 200 lines; a case page is named after its case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Event map regrouped into six families, 33 events including hypothetical ones, with the journal categories (section 4)
- categories.py turns detector episodes into categorised events: absorption of brief jumps, merging, catégorique phrases
- Journal tab: one "What happened" row coloured by family, cards, and an Images button playing the event's timelapse
- Plan: user-validated points marked, points 39 and 40 added; anomaly definition clarified in D-056
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Displacements tab now puts the list on the left and the curves on
the right, kept on screen while the list scrolls. Hovering a row keeps
only its target in color and draws it on the curves, with fixed scales:
a jump gets its 2 hours highlighted, the 24 h levels before and after
and an arrow with the jump in mm; a trend gets its period, the
Theil-Sen line, the daily levels, its start and end and the drift as
an arrow. #focus=j0-0, t0 or e0 opens with a row hovered, for captures.
Documented in the detection note: on NORDEND, PR17 and PR06 clear the
thresholds by a wide margin, the next curves fall short (PR03 nearest).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The jumps and trends now sit in a Displacements tab after Diagnosis
(#tab=displacements), no longer in a separate sidebar page; a click on
a row opens the Diagnosis tab on that target.
Documented: what the client looks at (jumps and trends, not errors,
handled upstream) in docs/displacement/domaine/ce-que-regarde-le-client.md;
the expert's reading of NORDEND in the case journey; the experimental
way of working in vues-deplacements/CLAUDE.md; D-055, the spec, the
note, the README and the project map.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mechanical anomaly detection on the hourly displacements in mm, data
unchanged: for each target and axis, the largest gap between the median
of the next 24 h and the previous 24 h, kept above 3 times the target's
daily swing (provisional). Jumps within 6 h form one event, so a common
jump is not several anomalies. Grey curves for all targets, colour for
those that jumped; a click opens Explore, Diagnosis.
Note on the idea, first trial and leads in
vues-deplacements/contexte/detection-d-anomalies.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- docs/displacement/domaine/incident-de-barcelone.md: the AIMsight
regression case is data/CLARIDEN/ (case C-004); what the data show,
the documentation's reading (reference/compensation instability)
against the expert's (real movement of the structure, confirmed by
other instruments, amplitude likely overestimated), open questions.
- nouveau-cas: platform zips date images in UTC, exports cover local
days; keep every image of the range, even at 10 min.
- colonnes-des-deplacements: native cadence is 15 or 10 min.
- CLAUDE.md: four cases in the base, displacement data of NORDEND,
WILDSTRUBEL and CLARIDEN.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>