Business logic understanding — Lovable dashboard
Accumulated understanding of the Lovable dashboard’s business logic, built up day-by-day until everything is clear (tracked as a blocker task in .local/tasks/). Primary sources: Malcolm’s walkthrough transcripts/videos, the Lovable source clone (.local/lovable-source/, analyze origin/main), and its docs/*.md (application-overview, database-overview, test-plan-happy-paths).
Files
| File | What |
|---|---|
2026-07-08-malcolm-walkthrough-demo.md |
FULL notes from Malcolm’s team demo (complete after screenshot capture 2026-07-09): scanner shipping flow + partnership prerequisite + implied chain of custody + mass balance, supply-chain page privacy rules, claim process A→Z (PO → link → confirm → certificate w/ QR), programme owner mini-CMS, fibre programme roles, tiers, hardcoded-apparel decision, certificates, timeline + open questions |
2026-07-09-claims-and-product-flow.md |
What a claim IS (credit-based model, production_claims lifecycle proposed→confirmed/rejected, validate_production_claim_limits gate) + full end-to-end flow (programme → pigment scans → chain-of-custody shipping pairs → PO → link → propose → confirm → evidence pack → public URL/QR tokens) — reconstructs the untranscribed walkthrough segment from source |
2026-07-09-walkthrough-2-linking-retirement-rules.md |
Walkthrough #2 (Hung’s live test + Malcolm Q&A): scan→claim linking is DELIBERATELY unrestricted (any own-facility scan attachable, even addressed to another company), retirement scope (tier-1 linked scans retire on confirm; sub-tier scans NEVER retire), unlink lifecycle (reject/cancel), partial claims (re-submit PO, no over-claim, buggy), tier-1 sets claim amount (not tier-0), tier = self-declared company field (company-vs-facility debate), Malcolm’s export stance: do NOT let another AI re-interpret business logic |
flow-explained.md |
Plain-language Vietnamese explainer (INTENTIONALLY Vietnamese — the one exception to English-only here): 4-sentence summary, glossary (18 English terms explained for non-experts), running story example (Sundown Gin → ACME chain → Target → US customs), 12 steps, 10 rules. Source of truth for flow-diagram.html |
flow-diagram.html |
Rendered human-readable version of flow-explained.md (open in browser): glossary cards, story banner, tier strip, 12 steps grouped by actor with plain-VI descriptions + gates, 10-rule panel. Regenerate BOTH files when flow knowledge changes |
persona-case-matrix.md |
EXHAUSTIVE persona × route × state × trigger matrix (A-Z 100% of UI cases) built via code-first fan-out over the Lovable clone origin/main (7 Explore agents, 2026-07-16). Covers the persona-derivation model + demo switchUser mechanism + nav-gating, then every page’s render branches (loading/empty/status enums/modals/errors) with file:line, plus Part C findings/bugs (Owner route 404, /admin open-by-default, dead branches, SAT error-sheets absent). Companion HTML: persona-case-matrix.html |
gui-walkthrough/ |
LIVE-GUI walkthrough via Claude-in-Chrome (2026-07-16, full “explore everything” pass — ALL 6 personas + key wizards). README.md = reusable case library (per-persona numbered step tables + routes + what-you-see, the happy-path claim flow mapped across personas, the link-verifications + make-a-claim wizards step-by-step, evidence-pack viewer, scanner simulator, admin console, shared settings, confirmed-live bugs, raw UX observations parking-lot, coverage status). 11 GIF recordings (cases 01-11, incl. the authorised real E2E claim) + screenshots/ (39 stills) + walkthrough-report.html (interactive report: case grid → GIF modal + screenshot lightbox, a Start-here learning path, and mirrored UX-Findings + Q&A sections). State-mutating final actions exercised only in the authorised E2E (cases 10-11, test record CLM-D0DD0958 — safe to delete) |
2026-07-16-code-resolved-open-questions.md |
SOURCE-ONLY pass (3 Explore agents, no browser) closing every code-resolvable open question: full supply-chain/shipping flow (direction columns, chainWalk tier derivation, partner-masking, non-partner visibility), full PO→claims→evidence-pack flow (po_status trigger mechanics, partial-claim volume tracking, non-PO path, pack_types 3-DB-vs-8-UI, public-URL/QR real + webhook NOT), Part C bug triage (verified PRESENT/FIXED + monet-fixable vs Lovable-owned), and an explicit “NOT code-resolvable — needs Malcolm/team/design” list. Satisfies T74 done-when part 2 (missing walkthrough segment documented from source) |
knowledge-base/ |
CONSOLIDATED knowledge base (2026-08-19): from-scratch re-derivation of ALL business logic against the QC Test Script v1.0, Lovable origin/main (ffab85ca), migrations, canvas, SAT, backend-specs and memory — 15 markdown files with per-claim confidence markers (VERIFIED/STATED/INFERRED/UNKNOWN), hand-drawn SVG ERD/lifecycles/persona-map, per-persona deep dives incl. roles the QC script omits, implementation-state matrix, 20 ranked open questions. Explorable offline portal: knowledge-base/portal.html (self-contained, file:// safe); each .md also rendered standalone; rebuild portal via bun knowledge-base/build-portal.mjs after re-rendering the .md with md2html. Supersedes nothing — the dated walkthrough files remain the raw sources it cites |
ux-findings-for-malcolm.md |
The T75 deliverable (drafted 2026-07-17 from the GUI walkthrough): 8 UX/logic findings each TRIAGED (bug / fix-in-monet / fix-in-Lovable / verify / out-of-scope / ask-Malcolm) with persona + route + screenshot, the 10 OPEN questions still needing Malcolm/team, and a one-line pointer list of the 13 RESOLVED answers (full detail + citations in 2026-07-16-code-resolved-open-questions.md). Mirrored in gui-walkthrough/walkthrough-report.html (UX Findings + Q&A sections) |
Conventions
- One file per session/source, date-prefixed. English only.
- Every claim that can be verified against the Lovable clone should be cross-checked (
git show origin/main:<path>) and cited. - Open questions live at the bottom of each file as checklists; resolve them in later files or in place with a dated note.