Yaumiyya — Application Documentation

Last updated 30 September 2026

On this page

Who this is for
Contents
How to read the verification labels
How this package was produced (tools)
Regenerating the portal

Yaumiyya (ޔައުމިއްޔާ, "by Atoll Mind") turns Dhivehi and English meetings into transcripts, summaries, formal minutes, decisions and tracked actions. This package documents the actual running application as it behaved during a documentation run on 30 September 2026 against the local development server (npm run dev, http://localhost:3000, branch user-decisions-0927).

Every screenshot in this package was captured from the real application during that run, using synthetic demonstration data. No screenshot is a mock-up.

Open the portal: double-click documentation/index.html (works offline, no server needed). It has navigation, search, every page below and a filterable screenshot gallery.

Who this is for#

Reader	Start here
New user / staff in training	Getting started → Complete meeting workflow
Support staff	Troubleshooting, FAQ, Status reference
Administrators	Roles & permissions, Plans & credits, Admin console
Management	Feature inventory, Coverage audit
Developers / QA	Feature inventory (with source locations), Coverage audit, Screen atlas

Contents#

Getting started — Getting started · Navigation

Features

Workflows (end-to-end tutorials)

Scenarios — Demonstration scenario · Sample data

Reference — Roles & permissions · Status reference · Troubleshooting · FAQ · Screen atlas

Audit — Feature inventory · Coverage audit

Sample output — samples/digital-services-improvement-meeting-minutes.pdf (printed from the app's own Print / PDF view)

How to read the verification labels#

Each feature page marks what was actually checked:

Label	Meaning
VERIFIED THROUGH UI	Performed in a real browser during this run; the screenshot shows the result.
VERIFIED THROUGH CODE	Behaviour established by reading the implementation; not exercised in the browser.
NOT VERIFIED	Neither exercised nor conclusively established; the reason is given.

How this package was produced (tools)#

Requested tool	Available?	What was used instead
AppAtlas plugin (`sleepyw33kday/appatlas`)	Not installed in this Claude Code environment (not in local plugins, `~/.claude/skills`, or the claude.ai plugin list).	Its published methodology was followed by hand: recon → section-by-section crawl → review & feature audit (walked / shallow / skipped / missed) → browsable atlas. See Screen atlas and Coverage. AppAtlas itself was not invoked.
Playwright Claude Code plugin	Not installed as a plugin.	The installed gstack `browse` skill, a Playwright-driven headless Chromium, was invoked and used for all browser interaction (navigation, clicks, forms, uploads, dialogs, viewports, screenshots, PDF).
app-screenshots (`alexanderop/app-screenshots`)	Not installed.	Its technique was reproduced: an SVG overlay drawn from DOM bounding boxes (numbered badges, dashed boxes, arrows, labels, rotating colours) injected before capture, then removed. Originals are kept next to every `-annotated` copy. app-screenshots itself was not invoked.

Account used: a throwaway QA account created by the project's own _prepare-qa.mjs script (run by the owner), placed on a Team plan for this run. Its email is redacted in screenshots, as are the one-time invitation link and the business bank account number.

Regenerating the portal#

The online documentation (/documentation in the app) and the offline index.html are both built from these Markdown files and screenshots/manifest.tsv. One command regenerates the inventory, the image variants and the offline copy, then checks links, anchors, images and animations and scans for anything private:

bash

npm run docs

The recorded animations are made by a separate pipeline in documentation/animations/ (see its README); nothing about them is rendered when the app is built or deployed.