CaseSheet — front end for SIH26047 (Patient Case-Taking Software)
A UI-only Django project for the problem statement Patient Case-Taking Software
(All India Institute of Ayurveda, Ministry of Ayush). A "case sheet" is what the
patient record is called in an Indian hospital, and case-taking is what this software does.
Django and nothing else. Django templates, Django static files, hand-written CSS
and a little plain browser JavaScript. No React, no Tailwind, no Bootstrap, no CDN,
no build step, no Node. No models, no database, no external services — every screen
renders from fixed demo content in frontend/demo_data.py. English only.
What it looks like
The patient says where it hurts — a labelled body map beside a picture grid, for a
patient who may not read well.

The interview listens, and escalates — every question can be spoken or tapped, and a
red-flag answer pages triage before the patient has left the kiosk.

The physician gets a draft, not a diagnosis — structured history, scanned documents and
out-of-range values, to accept, amend or reject.

Accessibility is not a settings page — high contrast and three text sizes are in the
header of every kiosk screen.

All 19 screenshots are in docs/screenshots/, numbered in the order a
patient walks through them.
Run it
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
python manage.py runserver
Then open http://127.0.0.1:8000/screens/ for an index of every screen.
There is no database, so migrate is neither needed nor possible: DATABASES is empty
and INSTALLED_APPS holds only django.contrib.staticfiles and this project's frontend app.
Screens
| URL |
Screen |
/ |
Welcome |
/identify/ |
ABHA scan, ABHA number, Aadhaar fingerprint, or new registration |
/consent/ |
Four consent choices, each revocable |
/complaint/ |
Body map with labelled regions, plus a picture grid |
/interview/?step=1..6 |
The interview, in speak or tap mode |
/ayush/ |
Dashavidha Pariksha |
/documents/ |
Scanning, extraction and the medical timeline |
/summary/ |
The structured history the patient confirms |
/done/ |
Token, room and priority |
/clinician/ |
OPD queue with red-flag escalation |
/clinician/patient/ |
The history summary a physician accepts or amends |
/portal/login/ |
Patient portal login |
/portal/ |
Patient dashboard |
/portal/records/ |
Patient documents |
/screens/ |
Index of every screen |
How the problem statement maps onto the UI
- Module A, conversational multimodal history —
/interview/. Every question can be
spoken or tapped (Speak / Tap). Each carries the SOCRATES probe it belongs to.
Step 4 fires the red-flag path.
- Module B, document digitisation —
/documents/. Scanner view, per-document extraction
with a confidence figure, out-of-range lab values highlighted, auto-ordered timeline.
- Module C, structured summary —
/summary/ for the patient, /clinician/patient/ for
the physician, in the standard clinical order. The physician view is an editable draft
with accept / amend / reject, never a diagnosis.
- Module D, consent, privacy and ABDM —
/consent/, plus the ABHA identity and consent
panels that recur across the flow.
- AYUSH history mode —
/ayush/: all ten Dashavidha Pariksha parameters.
- Accessibility — the header of every kiosk screen: three text sizes, high contrast,
audio-guide toggle, call a helper. Preferences persist in
localStorage.
Writing rules used here
One idea per screen, one line per idea. Headings are questions the patient would ask
themselves. No paragraph of explanation where a label will do, and nothing repeated on
two screens. If a screen has to teach the patient something, it is doing too much.
Layout
casesheet/
manage.py
casesheet/ project settings, urls, wsgi, asgi
frontend/ app: urls, views, demo_data, context_processors
templates/
base.html document shell
screens.html reviewer index
partials/ icon sprite, accessibility bar, progress rail
kiosk/ patient-facing screens
clinician/ clinician workspace
patient/ patient portal
static/
css/base.css design tokens, reset, shared components
css/kiosk.css kiosk surface
css/clinician.css clinician workspace and patient portal
js/a11y.js text size, contrast, audio
js/kiosk.js selection states, consent switches, microphone
Notes for whoever wires up the back end
- Views take no POST and hold no state. Each builds a context dict and renders.
- All display content lives in
frontend/demo_data.py; swapping it for real querysets
should not require template changes beyond field names.
- The interview advances through
?step=, so server-driven adaptive branching drops in
without touching the front end.
- The kiosk is English-only by design for now. Nothing is hard-coded against English:
the strings live in templates, so
{% trans %} and Django's i18n machinery can be
layered on when the other languages are added.