CA Revise · Ss

Active-recall revision app for CA Final Group 1 (FR · AFM · AAA), Nov 2026 attempt.

Each chapter is a casebook: rule on real cases, hunt the clause, sprint through MCQs, then read the law. Handwritten notes, SM-2 flip cards and typed practice sit alongside.

Stack

• Next.js 14 (App Router, TypeScript) — deploys to Vercel
• Plain CSS design system (app/globals.css) from the Ethics Casebook artifact: Bricolage Grotesque, Public Sans, IBM Plex Mono, Special Elite. Light and dark themes, flat hairline surfaces, 44px tap targets, phone bottom bar.
• Handwritten fonts (Kalam / Caveat / Shadows Into Light) are used only in the Notes view (scoped under .hand).
• localStorage for progress (v0.1 — no backend needed)
• Phase 2: Railway (Postgres + FastAPI) for cross-device sync and Claude API for grading typed answers

Local dev

npm install
npm run dev
# → http://localhost:3000

Deploy to Vercel

1. Push this folder to a fresh GitHub repo:

   git init -b main
   git add .
   git commit -m "ca-revise: v0.1 with AAA Ch 19"
   git remote add origin [email protected]:<you>/ca-revise.git
   git push -u origin main
   

2. On vercel.com → Add New Project → import the repo.
3. Framework: Next.js (auto-detected). No env vars needed for v0.1.
4. Click Deploy. First deploy takes ~90 seconds.
5. On your phone: open the deployed URL → Add to Home Screen (Safari share menu / Chrome menu). Behaves like a native app.

Adding a chapter (any subject)

Chapters are data only. No component changes are needed.

npm run new-chapter -- fr 3 "Ind AS 115 Revenue" --casebook

This creates data/fr/ch03-ind-as-115-revenue.json (notes, flip cards, written practice) and, with --casebook, data/fr/ch03-casebook.json (case drill, provision hunt, MCQs, law library, cheat sheet), then registers the chapter in lib/registry.ts. Fill in the JSON and it appears on the home page.

Everything chapter-specific lives in the casebook's meta block:

key	what it controls	example
sources	drill filter pills, [kind, label]	[["rtp","RTP Nov 2026"]]
verdicts	ruling buttons and stamps; omit for reveal-and-self-mark	[{"label":"Guilty","tone":"bad","hint":"misconduct"}]
caseNoun / provisionNoun	wording across the app	"case" / "clause", "question" / "standard"
huntQuestion	Provision Hunt prompt
tagline / intro	chapter home headline (*word* is highlighted; {cases} is the case count)
scopeTitle / scopeNote / methodTitle	home page section headings

Provisions with "hunt": true are the answer options in Provision Hunt (all provisions if none are flagged).

A casebook may also carry formulas (a formula bank: families led by a master formula, each formula with its logic, derivation, exam cue and trap). It adds a Formulas tab to the chapter and feeds the subject-wide bank at /<subject>/formulas. Every AFM chapter has one. See data/_CASEBOOK_TEMPLATE.json for the full shape.

Building a chapter from the study material (Claude Code skill)

Drop the ICAI study-material PDF in inbox/ and ask Claude Code, for example: "add FR chapter 3, SM is inbox/fr-ch3.pdf". The ca-chapter skill (.claude/skills/ca-chapter/) runs the same steps every time: extract every illustration and TYK question, verify the attempt's amendments, collect RTP / MTP / past-paper questions from the ICAI BoS portal, rank topics by exam frequency, write both data files, validate and build. Its research report lands in research/<chapter-id>.md.

The programs it uses can also be run by hand:

npm run unlock-pdf                                  # inbox/*.pdf -> inbox/unlocked/ (removes copy restrictions)
npm run extract-sm -- inbox/unlocked/sm.pdf work/fr-ch03/sm   # text + inventory of illustrations / TYK
npm run assemble -- fr-ch03                          # work/fr-ch03/parts/*.json -> data files
npm run validate -- fr-ch03                          # schema, links, source labels, coverage

Chapter JSON schema

{
  "id": "aaa-ch19",             // unique across all chapters
  "subject": "aaa",              // "aaa" | "fr" | "afm"
  "number": 19,                  // ICAI SM chapter number
  "title": "Professional Ethics",
  "summary": "...",              // one line for the list
  "currency": "...",             // optional — Nov 2026 flag
  "sections": [
    {
      "id": "sch1-part1",
      "title": "First Schedule Part I",
      "emoji": "📙",
      "cards": [
        {
          "id": "s1p1c11",
          "type": "clause",       // clause | concept | illustration | mnemonic | section | case
          "tags": ["Cl 11", "Reg 190A"],
          "front": "Cl 11 — Other business or occupation",
          "back": {
            "bareText": "engages in any business or occupation…",
            "simplified": "Cannot engage in any business…",
            "exception": "General list…",
            "explanation": "…",
            "trap": "25-hour cap…",
            "cases": ["Nov 2023: Editor 'Company Audit' → NOT GUILTY"],
            "cross": ["Reg 190A", "Sec 2(2)(iv)"]
          }
        }
      ]
    }
  ],
  "practice": [
    {
      "id": "q-cl11-editor",
      "prompt": "CA X becomes the editor of a professional journal…",
      "clauseRef": "Cl 11 Part I 1st Sch",
      "hints": ["Reg 190A general list…"],
      "modelAnswer": "Step 1 — CITE: …\nStep 2 — RULE: …\nStep 3 — APPLY: …\nStep 4 — CONCLUDE: …",
      "markingPoints": ["Cites Clause 11", "Refs Reg 190A", "Concludes: not guilty"]
    }
  ]
}

Modes (per chapter)

Casebook modes — available when the chapter registers a casebook JSON (see data/aaa/ch19-casebook.json, shape in lib/casebook.ts):

• Home — coverage ring, accuracy, best streak, weak clauses (<60%), RTP scope rail, ICAI 4-step answer method.
• Case Drill — read the facts, rule Guilty / Not guilty / Partly, get stamped, read the model answer and examiner's trap. Filter by source (RTP, SM Illustrations, SM TYK, exam pattern, Code 2026) and New / Missed.
• Clause Hunt — 10 rounds: which provision do the facts trigger?
• MCQ Sprint — shuffled MCQs with instant feedback.
• Law Library — every provision as a tile; tap for the law, notes, updates, memory hook and linked questions.
• Cheat Sheet — numbers table and facts → clause finder.

Core modes — every chapter (from the chapter JSON):

• Notes — handwritten, lined-paper notes, section by section.
• Flip Cards — SM-2 spaced repetition (Again / Hard / Good / Easy).
• Practice — type the full answer, reveal the model answer and marking points, self-mark.

Roadmap

• Phase 2 — Railway backend (FastAPI + Postgres): sync progress across devices, historical charts, weak-clause radar.
• Claude API grader for the Practice mode — rubric-based scoring of your typed answer against the marking points.
• Visual advertisements — chapter-summary infographics generated with Claude to keep on the phone.
• All chapters — full FR / AFM / AAA syllabus, one JSON file at a time.

Attribution

Built for personal revision — Ss ([email protected]). Not for redistribution.

Mastery layer (path, compare, games, step solver)

Optional per chapter: data/<subject>/chNN-learn.json, registered with a learn: loader in lib/registry.ts. Shapes are in lib/learnData.ts; views are components/views/{Path,Compare,Games,Solve}View.tsx. Adds four menu items (Mastery path, a side-by-side compare, Games, Step solver) and a floating scratch pad (Jotter). Progress is stored per chapter (ca-revise:v1:learn:<id>) and synced like everything else.

• Levels: sequential lessons (teach blocks, a picture table, how it connects, recall check, "twist").
• Compare: rows plus a "which is which" sorter.
• Games: match, sort and order types, all data-driven.
• Step solver: each numerical is cut into small steps with round numbers, plus twisted versions.

Ch 14 is generated from scripts/learn-src/ch14-*.mjs: node scripts/build-learn.mjs aaa-ch14 (recomputes every solver figure before writing) and node scripts/merge-ch14-additions.mjs (CFA-style and one-line questions, with a figure check).