wiggleverse-ecomm/specs/SD-0001-mvp-sign-up-and-single-storefront.md

# Solution Design: ecomm MVP — sign up and create a single storefront

| | |
| --- | --- |
| **Anchor** | [wiggleverse/wiggleverse-ecomm#1](https://git.wiggleverse.org/wiggleverse/wiggleverse-ecomm/issues/1) (`type/feature`, `priority/P1`) |
| **Author(s)** | Claude (session ecomm-0020), for Ben Stull |
| **Reviewers / approvers** | Ben Stull |
| **Status** | `draft` |
| **Version** | v0.1.0 |
| **Source artifacts** | BDD corpus: [research/shopify](https://git.wiggleverse.org/wiggleverse/wiggleverse-ecomm-prototype-content/src/branch/main/research/shopify) (esp. [flows/first-run-experience.md](https://git.wiggleverse.org/wiggleverse/wiggleverse-ecomm-prototype-content/src/branch/main/research/shopify/flows/first-run-experience.md)) · Prototype: [wiggleverse-ecomm-prototype](https://git.wiggleverse.org/wiggleverse/wiggleverse-ecomm-prototype) (R01–R07) · Reference: Shopify first-run experience · Supersedes: — |

**Change log**

| Date | Version | Change | By |
| --- | --- | --- | --- |
| 2026-06-10 | v0.1.0 | Initial draft (session ecomm-0020) | Claude |

---

## 1. Business Context

*The business lens — solution-agnostic throughout. No mechanism is proposed until §2.*

### 1.1 Executive Summary

Independent merchants who want to sell online need a commerce platform that
treats them honestly — and today Wiggleverse has nothing to offer them: no way
to exist on the platform, no storefront to own, no surface to manage. This
design delivers the first real value of ecomm: a person can become a merchant —
establish their identity, claim their one storefront, and stand on a stable
management surface — in every environment the platform runs in, from day one.
The beneficiaries are the first merchants (who can finally exist on the
platform) and the ecomm line itself (every subsequent capability — catalog,
orders, checkout — gains a real account-and-storefront spine to attach to,
instead of mocks). Just as importantly, it proves the platform can be brought
to life anywhere from nothing: a fresh, empty environment reaches "first
merchant, first storefront" through the product alone, with no hand-seeded
data — so launching the real service is an ordinary, rehearsed act rather than
a one-off ritual.

### 1.2 Background

ecomm is Wiggleverse's multi-tenant, OHM-grounded alternative to Shopify — a
commerce platform whose mechanics are accountable to the
[OHM concept corpus](https://ohm.wiggleverse.org) (no dark patterns, data
minimization, durable addressability, honest pricing). The product was explored
in a first-pass prototype (`wiggleverse-ecomm-prototype`, releases R01–R07,
189 scenario-bound tests) built against a harvested corpus of ~1,238 BDD
scenarios describing Shopify's behavior. That prototype is read-only prior art;
`wiggleverse-ecomm` is the clean greenfield rebuild, stood up 2026-06-10 with
its architecture intentionally undecided until this design.

The strategic moment: the rebuild has repos, an issue tracker, and a charter —
but no app code and no users. Everything in the product line queues behind one
question: *how does a person come to exist on the platform and own a
storefront?* Until that is answered for real — against real persistence, in
real environments — no other capability can ship honestly.

### 1.3 Business Actors / Roles

| Role | Responsible for (in the business) |
| --- | --- |
| **Merchant** | A person running a (small) commerce business who wants to sell under their own storefront: they decide to join a platform, establish their business presence on it, and manage that presence over time. |
| **Platform operator** | The person/org running the ecomm service itself: stands up and maintains the environments the platform runs in, and is accountable for the service being available to merchants. |
| **ecomm team** | Builds the platform; needs each capability to rest on real, shippable foundations rather than mocks. |

(Shoppers — the merchants' customers — are deliberately *not* an actor in this
design; nothing here is shopper-facing yet. See §1.7.)

### 1.4 Problem Statement

A person cannot become a merchant on ecomm. There is no way to establish an
identity on the platform, no way to claim a storefront, and no surface from
which to manage one — the platform has no users because it cannot have users.
And even if it could, there is today no defined path from "fresh, empty
environment" to "working platform serving its first merchant": local
development, pre-production, and production bring-up are all undefined, so
nothing built on top can ever actually launch.

### 1.5 Pain Points

| # | Pain | Who feels it | Cost / frequency today |
| --- | --- | --- | --- |
| PP-1 | No way to establish an identity on the platform — a would-be merchant simply cannot exist on ecomm. | Merchant | Total exclusion; blocks every merchant, always. |
| PP-2 | No way to own a storefront or reach a management surface — even an identity would have nothing to hold. | Merchant | Total; there is no unit of "my business" on the platform. |
| PP-3 | Every subsequent capability (catalog, orders, checkout, policies) has nowhere real to attach — work would have to build on mocks and be redone. | ecomm team | Blocks the entire product line; rework risk on everything shipped against fakes. |
| PP-4 | No defined, repeatable path from an empty environment to a working platform — bring-up of dev, pre-prod, and prod is folklore. | Platform operator | The service cannot launch anywhere; every environment question is open-ended toil. |

### 1.6 Targeted Business Outcomes

| Outcome | Success metric | Baseline → Target | Guardrail (must not regress) | How / when measured |
| --- | --- | --- | --- | --- |
| Merchants can exist on the platform | A real person can go from stranger to merchant-with-storefront, unaided | impossible → possible in every environment | No dark patterns at the door: no trial clock, no payment demand, no plan wall (OHM: Agency & Anti-Manipulation) | Walk the flow end-to-end in each environment at release |
| The product line is unblocked | Subsequent Features attach to a real account + storefront spine | everything mocked → nothing mocked | The spine doesn't need rework to host the next Feature (no throwaway auth/tenancy) | First post-MVP Feature builds on it without spine changes |
| Launching anywhere is routine | A fresh, empty environment reaches "first merchant signs up and creates the first storefront" through the product flows alone | folklore → documented, rehearsed gesture | Zero hand-seeded data in any environment | Bootstrap rehearsal executed on pre-prod, then prod, at release |

### 1.7 Scope (business)

- **In scope:** a person establishing an identity on the platform (joining and
  returning); a merchant establishing their one storefront; the merchant
  standing on a stable management surface for it; the platform being
  bring-up-able from empty in every environment it runs in (local development,
  pre-production, production).
- **Out of scope (later designs):** anything sold or bought — catalog, orders,
  checkout, payments; the shopper-facing public storefront; teams/staff
  (anyone besides the one merchant touching the storefront); plans, billing,
  or any commercial relationship between merchant and platform; storefront
  settings/management beyond simply *having* the surface.
- **Non-goals:** gating who may join (the prototype admitted users by
  invitation; ecomm is an open product — anyone may become a merchant, and
  re-introducing an admission gate is explicitly not pursued); supporting more
  than one storefront per merchant *in this design* (the business wants the
  door left open, not the capability now).

### 1.8 Assumptions · Constraints · Dependencies

- **Assumptions:**
  - One storefront per merchant is the right *experience* bar for the MVP;
    multiple-storefront merchants are a future need, not a current one. Risk
    if wrong: an early merchant genuinely needs a second storefront — accepted;
    the door is designed to stay open (§1.7, #1 acceptance).
  - Merchants are reachable by email and will complete an email-based step to
    join. Risk if wrong: an entry channel rethink — accepted at MVP scale.
- **Constraints:**
  - OHM-groundedness is a charter constraint: entry must be honest — no
    manufactured urgency, no data collected beyond need, no lock-in mechanics.
  - The three environments are a constraint, not a stretch goal: Feature #1's
    acceptance says no environment is "later".
  - Wiggleverse engineering standards apply (handbook): provisioning and
    deploy via flotilla (§8), secrets never in repos or transcripts (§6.3),
    private repos by default (§5.5).
- **Dependencies:**
  - **flotilla / launch-app** for environment provisioning and deploy (the
    operator-run front door, handbook §8.5) — needed when the pre-prod and
    prod environments are stood up.
  - An **email delivery channel** for the entry step in real environments —
    needed by the production-readiness slice (§7).

### 1.9 Business Use Cases

**BUC-1 — As a merchant, I can establish my identity on the platform, so that I can begin doing business there.**

```gherkin
Scenario: BUC-1 — A stranger becomes known to the platform
  Given a person who has never dealt with the platform
  When they present themselves and prove who they are
  Then the platform knows them as a returning individual from then on
  And they were asked for nothing beyond what proving who they are required
```

```gherkin
Scenario: BUC-1a — A person who cannot prove who they are is not admitted as someone else
  Given a person presenting an identity they cannot prove
  When their proof fails
  Then the platform does not mistake them for the claimed identity
  And the genuine owner of that identity is unharmed
```

- **BUC-1 acceptance criteria:** the same person, returning later, is
  recognized as the same merchant; two different people are never conflated;
  no personal data beyond the minimum was demanded (OHM: Privacy & Data
  Minimization).

**BUC-2 — As a returning merchant, I can resume my standing on the platform, so that my business presence persists beyond a single visit.**

```gherkin
Scenario: BUC-2 — A known merchant returns
  Given a merchant the platform already knows
  When they return and prove who they are
  Then they resume exactly the standing they had — same identity, same storefront
```

- **BUC-2 acceptance criteria:** returning costs only the proof-of-identity
  step; nothing about their standing is lost or reset.

**BUC-3 — As a merchant, I can establish my storefront, so that my business has a presence of its own on the platform.**

```gherkin
Scenario: BUC-3 — A merchant claims their storefront
  Given a merchant known to the platform who has no storefront yet
  When they establish their storefront
  Then the platform holds exactly one storefront that is theirs
  And establishing it cost them nothing and committed them to nothing
```

```gherkin
Scenario: BUC-3a — One storefront is the bar
  Given a merchant who already has their storefront
  When they deal with the platform
  Then they are never led toward acquiring a second one
```

- **BUC-3 acceptance criteria:** every merchant who completes the step has
  exactly one storefront; no payment, plan, or commitment was demanded
  (corpus: 14.01.0015, 14.01.0016); the platform's books would still be
  correct if a merchant were someday allowed several (the 1-1 bar is an
  experience rule, not a law of the data — Feature #1 constraint).

**BUC-4 — As a merchant, I can stand on a management surface for my storefront, so that everything I will later do for my business has a home.**

```gherkin
Scenario: BUC-4 — The merchant arrives at their storefront's home
  Given a merchant with a storefront
  When they return to the platform
  Then they arrive directly at their storefront's management surface
  And it is honestly empty — it claims no capability that does not exist yet
```

- **BUC-4 acceptance criteria:** arrival is direct (no dead ends, no limbo —
  a merchant *without* a storefront is guided to establish one instead, never
  stranded); the surface carries the storefront's identity and nothing
  fabricated.

**BUC-5 — As a platform operator, I can bring the platform to life in a fresh environment, so that merchants can be served there without any hand-built state.**

```gherkin
Scenario: BUC-5 — An empty environment reaches its first merchant
  Given a brand-new environment with empty persistence
  When the operator performs the documented bring-up gesture
  And the first merchant walks the ordinary product flows
  Then the environment serves that merchant a working storefront
  And no data was placed by hand at any step
```

```gherkin
Scenario: BUC-5a — Bring-up is rehearsable
  Given the bring-up gesture documented for one environment
  When it is performed in the next environment (dev → pre-prod → prod)
  Then it is the same gesture, and day-one production is simply the bootstrap state
```

- **BUC-5 acceptance criteria:** bring-up is a repeatable, documented gesture
  (not folklore); it holds identically in local development, pre-production,
  and production; the first merchant arrives through the product flows alone.

---

## 2. Solution Proposal

**Build the first vertical slice of the ecomm web application** — a
multi-tenant modular monolith that carries forward the prototype's *proven*
shape and inverts its *disqualifying* assumption:

- **Carry forward** (validated by R01–R07 of the prototype): a four-layer
  Python backend (`entrypoint → api → domains → platform`) with the layer
  contract mechanically enforced; SQLite (WAL) with forward-only numbered
  migrations and no ORM; a React/Vite SPA admin talking to a screen-shaped
  REST BFF; scenario-bound end-to-end tests (one test per BDD scenario ID);
  OHM concepts cited inline next to the rules they ground.
- **Invert** (the prototype's bootstrap was the wrong way round for a real
  product): **no seeded tenant and no admission gate.** The prototype was born
  with one hard-coded store and one invited owner; ecomm is born *empty*.
  Identity is open sign-up via **email + one-time code** (passwordless;
  corpus 14.01.0003–0004), the storefront is created *through the product* by
  its merchant, and an empty database is a first-class, fully working state —
  which is precisely what makes BUC-5's bootstrap story true by construction
  rather than by tooling.
- **Defer** (YAGNI for four screens): the prototype's second API surface
  (GraphQL) is not built in this MVP; the layering keeps the seam so it can be
  added later as a projection over the same domain core. Google/Apple OAuth
  and sign-in-by-emailed-link (corpus 14.01.0005–0007) are deferred the same
  way — the email-canonical identity rule (14.01.0010) keeps them addable
  without account migration.

**Why this approach** over the alternatives:

- *Do nothing / concierge* (hand-create merchants on request): fails BUC-5
  outright (every merchant is hand-seeded state) and validates nothing — the
  prototype already validated the need; the open question is productionizing.
- *Re-architect on a heavier stack* (Postgres, separate services, SSR
  framework): buys scale and rendering properties the MVP measurably does not
  need, at the cost of slower iteration and a bigger bootstrap surface —
  directly against the "launching anywhere is routine" outcome (§1.6). The
  single-VM flotilla deployment standard (handbook §8) favors the lighter
  stack; revisit at real scale (§6.7, §7.4).
- *Resume the prototype codebase*: it bakes in the seeded-tenant and
  invite-gate assumptions at the foundation, and the rebuild exists precisely
  to lay clean foundations. Its *patterns* are kept; its code is reference
  only.

**Solution-specific scope:** the landing page, sign-up/log-in (email +
one-time code), create-storefront (both entry points: post-sign-up and
returning-without-storefront), the empty admin shell, sign-out, and the
bring-up story for localhost/PPE/Prod including real email delivery in
deployed environments.

**Solution-specific non-goals (this release):** GraphQL surface; OAuth
providers and magic-link completion; the onboarding questionnaire and POS
opt-in from the corpus first-run flow (14.01.0017–0024) — the merchant lands
directly on their admin; storefront rename/settings (14.01.0025 makes naming
optional at creation; *changing* it later is the Settings Feature's job);
account-lifecycle polish beyond what passwordless entry gives for free
(password reset does not exist because passwords do not exist; the one-time
code *is* email verification; account deletion is deferred and logged in §9).

---

## 3. Product Personas

| Product persona | In ecomm | Maps to business role(s) |
| --- | --- | --- |
| **Visitor** | An unauthenticated person on the entry surfaces: landing, sign-up/log-in. | Merchant (before they are known) |
| **Merchant** | An authenticated account holder; creates and owns the storefront; lands on its admin. The MVP's only authenticated user type. | Merchant |
| **Operator** | Brings up and deploys environments; interacts through flotilla and the bring-up runbook, not the web UI. | Platform operator; ecomm team |

## 4. Product Use Cases

Corpus grounding: scenario IDs cite the harvested Shopify first-run experience
([flows/first-run-experience.md](https://git.wiggleverse.org/wiggleverse/wiggleverse-ecomm-prototype-content/src/branch/main/research/shopify/flows/first-run-experience.md),
14.01.\*) and the zero-state dashboard (01.01.0008). Divergences are marked.

**PUC-1 — Landing offers the two doors** *(realizes BUC-1, BUC-2; corpus 14.01.0001)*

```gherkin
Scenario: PUC-1 — The landing page offers signing up or logging in
  Given a Visitor on the ecomm landing page
  Then they see what ecomm is, a primary action to get started, and a link to log in
  And no trial, pricing, or fee is presented (corpus 14.01.0011)
```

**PUC-2 — Sign up with email + one-time code** *(realizes BUC-1; corpus 14.01.0002–0004)*

```gherkin
Scenario: PUC-2 — A Visitor signs up
  Given a Visitor who chose to get started
  When they enter their email address and request a code
  Then a one-time code is sent to that address
  And the screen moves to code entry, telling them where the code went
  When they enter the code within its validity window
  Then they are signed in as a new Merchant
  And they are taken directly into creating their storefront (PUC-4)
```

```gherkin
Scenario: PUC-2a — Wrong code
  Given a Visitor on the code-entry screen
  When they enter an incorrect code
  Then the screen says the code didn't match and lets them retry or request a new one
  And after repeated failures the code is invalidated and a fresh request is required
```

```gherkin
Scenario: PUC-2b — Expired code
  When they enter a code past its validity window
  Then the screen says the code expired and offers to send a fresh one
```

```gherkin
Scenario: PUC-2c — Resend is rate-limited
  When they ask for another code immediately after one was sent
  Then the screen asks them to wait briefly before resending
```

**PUC-3 — Log in as a returning Merchant** *(realizes BUC-2; corpus 14.01.0009, 14.01.0010)*

```gherkin
Scenario: PUC-3 — A returning Merchant logs in
  Given a Visitor whose email already belongs to a Merchant account
  When they complete the same email + code flow (from either door)
  Then they are signed in to that same account — the email is the canonical key
  And no duplicate account is created (corpus 14.01.0010)
```

*(Sign-up and log-in are deliberately the same mechanism; the two doors differ
only in framing. A "sign up" with a known email is a log-in; a "log in" with an
unknown email creates the account. Honest copy on the code-entry screen states
which happened.)*

**PUC-4 — Create the storefront after sign-up** *(realizes BUC-3; corpus 14.01.0013, 14.01.0015, 14.01.0016, 14.01.0025, 14.01.0026)*

```gherkin
Scenario: PUC-4 — A new Merchant creates their storefront
  Given a Merchant signed in with no storefront
  Then they are on the create-storefront screen — there is nothing else to do yet
  When they optionally enter a storefront name and confirm
  Then the storefront is created — named as entered, or with a generated default name
  And no payment method, plan, or commitment was requested (14.01.0015/0016)
  And they land on the storefront's admin page (PUC-6)
```

**PUC-5 — Returning without a storefront → prompted to create** *(realizes BUC-3, BUC-4)*

```gherkin
Scenario: PUC-5 — A Merchant who never finished creating one
  Given a returning Merchant who has no storefront
  When they log in
  Then they arrive at the create-storefront screen (PUC-4), not an empty admin or an error
```

**PUC-6 — Returning with a storefront → straight to its admin** *(realizes BUC-4; corpus 14.01.0027)*

```gherkin
Scenario: PUC-6 — A Merchant returns to their storefront
  Given a returning Merchant who has a storefront
  When they log in
  Then they land directly on that storefront's admin page
  And nothing further is mandatory (corpus 14.01.0027)
```

**PUC-7 — One storefront: no second-storefront affordance** *(realizes BUC-3a; **diverges** from corpus 14.01.0012/0014 — multi-store per account is deliberately not offered; see §9 D-2)*

```gherkin
Scenario: PUC-7 — No "create another storefront" exists
  Given a Merchant with a storefront
  Then no surface offers creating another one
  And directly invoking storefront creation anyway is refused with an honest
    explanation that ecomm is one storefront per account today
```

**PUC-8 — The admin shell is honestly empty** *(realizes BUC-4; zero-state per corpus 01.01.0008)*

```gherkin
Scenario: PUC-8 — Empty admin
  Given a Merchant on their storefront's admin page
  Then the page carries the storefront's name and the signed-in account's email
  And states plainly that there is nothing to manage yet
  And shows no fabricated metrics, fake modules, or teaser upsells
```

**PUC-9 — Sign out** *(product-only)*

```gherkin
Scenario: PUC-9 — A Merchant signs out
  Given a signed-in Merchant anywhere in the admin
  When they sign out
  Then their session ends and they are back on the landing page as a Visitor
```

**PUC-10 — A developer bootstraps localhost** *(realizes BUC-5; Operator-facing — the surface is a terminal, not a screen)*

```gherkin
Scenario: PUC-10 — Clean checkout to first storefront, locally
  Given a clean checkout of wiggleverse-ecomm and the documented prerequisites
  When the developer runs the documented dev bring-up command
  Then the app starts against an empty local store, applying schema migrations itself
  And the developer completes sign-up → create storefront → admin in a browser
  And the one-time code reaches them through the local dev channel (no real mail needed)
```

**PUC-11 — The Operator bootstraps a deployed environment (PPE, then Prod)** *(realizes BUC-5, BUC-5a)*

```gherkin
Scenario: PUC-11 — Fresh deployed environment to first merchant
  Given an environment provisioned and deployed through flotilla, with empty persistence
  When the app starts, applying schema migrations itself
  Then the first real merchant can complete sign-up → storefront → admin
    through the public flows, with the code arriving by real email
  And the rehearsal on PPE and the day-one Prod state are the same gesture (BUC-5a)
```

## 5. UX Layout

Text is the source of truth; wireframes are generated from these descriptions
on demand and not committed. One SPA, four surfaces. Visual language: calm,
text-first, honest — no countdowns, no badges, no fake density.

### 5.1 Screen: Landing (serves PUC-1)

- **Purpose:** a Visitor learns what ecomm is and picks a door.
- **Layout (top → bottom):**
  - **Header:** wordmark "ecomm" (left); "Log in" link (right).
  - **Primary content:** one short value statement (what ecomm is, in OHM
    voice — honest commerce, your storefront is yours); primary button
    **"Create your storefront"**.
  - **Footer:** minimal (project identity).
- **States:** happy only — static page. A signed-in Merchant hitting `/` is
  redirected per entry routing (§6.5, PUC-6/PUC-5).
- **Notifications:** none.

### 5.2 Screen: Sign in (serves PUC-2, PUC-3; two steps)

- **Purpose:** the single email + one-time-code flow behind both doors.
- **Step 1 — email entry:**
  - **Header:** wordmark; back-to-landing link.
  - **Primary content:** heading framed by the door used ("Create your
    storefront" / "Log in" — same form); email field (label "Email";
    autofocus); primary button **"Send code"**; one line of honest copy:
    "We'll email you a one-time code. That's all we need."
  - **States:** happy → step 2 · invalid email format: inline field error ·
    sending: button busy state · resend cooldown active: button disabled with
    "you can resend in ~Ns" (PUC-2c) · delivery failure: banner "We couldn't
    send the code — try again" (no fake success; §6.9).
- **Step 2 — code entry:**
  - **Primary content:** "We sent a code to ⟨email⟩" (with a "wrong address?"
    link back to step 1); code field (one-time-code input semantics); primary
    button **"Continue"**; secondary "Resend code" (cooldown-aware).
  - **States:** happy: session starts → entry routing (§6.5) · wrong code:
    inline "That code didn't match" with attempts remaining (PUC-2a) ·
    expired: inline "That code expired" + resend offer (PUC-2b) · attempts
    exhausted: code invalidated, prompt to request a fresh one · loading:
    button busy.
  - **Honest-copy rule (PUC-3):** after verification the screen states which
    happened — "Welcome to ecomm" (new account) vs "Welcome back" (existing).
- **Notifications:** the one-time-code **email** (subject "Your ecomm code:
  ⟨code⟩"; body: the code, its validity window, and "if you didn't request
  this, ignore it"). In localhost/dev the code reaches the developer via the
  dev channel instead (§6.5 PUC-10).

### 5.3 Screen: Create storefront (serves PUC-4, PUC-5, PUC-7)

- **Purpose:** a storefront-less Merchant establishes their one storefront.
- **Layout:**
  - **Header:** wordmark; signed-in email (right) with a "Sign out" action
    (PUC-9).
  - **Primary content:** heading "Create your storefront"; name field (label
    "Storefront name", explicitly optional — helper text "You can leave this
    blank; we'll pick a placeholder name you can change later"); primary
    button **"Create storefront"**.
- **States:** happy: → admin (PUC-6) · blank name: allowed — a default name is
  generated (corpus 14.01.0026) · creating: button busy · already-owns
  (defense in depth, PUC-7): banner "Your account already has its storefront"
  + link to the admin · error: banner with retry.
- **Notifications:** none (no confirmation email — minimal channel use).

### 5.4 Screen: Admin shell (serves PUC-8, PUC-9)

- **Purpose:** the storefront's stable home; honestly empty in this release.
- **Layout:**
  - **Header:** storefront name (left — the admin's anchor identity);
    signed-in email + "Sign out" (right).
  - **Primary content:** empty state — heading with the storefront's name,
    one line: "There's nothing to manage yet. Catalog, orders, and settings
    will appear here as ecomm grows." Nothing else: no zeroed metric tiles,
    no locked-feature teasers (OHM: Agency & Anti-Manipulation — no
    manufactured aspiration).
- **States:** happy/empty are the same state by design · loading: minimal
  skeleton while `/me` resolves · session expired: redirect to sign-in.
- **Notifications:** none.

## 6. Technical Design

### 6.1 Invariants

- **INV-1 — Empty is a working state.** The application boots against empty
  persistence, applies its own schema migrations, and serves the full
  sign-up → storefront → admin flow. No code path assumes a pre-existing row;
  there is no seed step. *Enforced mechanically:* the bootstrap test (§6.8)
  runs the whole flow from a fresh database; it is the first test of the
  suite.
- **INV-2 — Email is the canonical account key.** One account per email
  (case-insensitively normalized); any future auth method resolves to the
  same account by email (corpus 14.01.0010). *Enforced:* unique index on the
  normalized email column.
- **INV-3 — One-time codes are handled like secrets.** Stored only as hashes;
  expire after 10 minutes; single-use; at most 5 verification attempts per
  code; resend no more than once per 60 seconds per email. Codes never appear
  in production logs or API responses. *Enforced:* schema (consumed/attempt
  columns) + service-layer checks + tests.
- **INV-4 — One storefront per account is the MVP rule, not the schema.**
  Ownership lives in a membership relation that admits many storefronts per
  account and many members per storefront; the *service layer* refuses a
  second storefront for the same account today (and the UX never offers it,
  PUC-7). Lifting the rule later is deleting one check, not a migration.
  *Enforced:* service-layer guard + test; no `UNIQUE(account_id)` anywhere in
  the storefront path.
- **INV-5 — Every storefront-scoped row carries its storefront.** All future
  tenant data references `storefront_id`; nothing tenant-scoped is global.
  (Trivially satisfied by this MVP's two tenant tables; stated now because it
  binds every later Feature.) *Enforced:* schema review + FK constraints.
- **INV-6 — One source of truth for rules.** Authentication, authorization,
  and the one-storefront guard live in the domain layer exactly once; every
  API surface (today REST; later GraphQL) delegates to them. *Enforced:*
  layering contract (import-linter) + the API layer owning no business logic.
- **INV-7 — Migrations are forward-only and fail-stop.** Numbered `.sql`
  files applied in order at startup, recorded in a migrations table; a failed
  migration aborts startup (the deploy fails; the prior version keeps
  serving). No down-migrations.
- **INV-8 — No deployment shape in framework code.** URLs, mail relay
  coordinates, secrets: all arrive via environment/deployment configuration
  (resolved from Secret Manager in deployed environments); none are baked
  into the app (handbook §5.3). *Enforced:* config surface table (§6.6) +
  review.
- **INV-9 — Honest failure.** When the platform cannot do what was asked
  (code not sent, storefront not created), the user is told so — no fake
  success, no silent drop (OHM: Transparency & Traceability).

### 6.2 High-level architecture

```mermaid
flowchart LR
  subgraph Browser
    SPA[Admin SPA<br/>React + Vite]
  end
  subgraph "ecomm backend (FastAPI, one process)"
    BFF[REST BFF<br/>main.py]
    subgraph domains
      ACC[accounts<br/>identity · OTC · sessions]
      SF[storefronts<br/>storefront · membership]
    end
    subgraph platform
      DB[db<br/>SQLite WAL + migrations]
      MAIL[mailer<br/>port: log / SMTP]
      DEPS[deps<br/>request wiring]
    end
  end
  STORE[(SQLite file)]
  RELAY[SMTP relay<br/>deployed envs only]

  SPA -->|JSON /api/*| BFF
  BFF --> ACC
  BFF --> SF
  ACC --> DB
  SF --> DB
  ACC --> MAIL
  DB --> STORE
  MAIL -.-> RELAY
```

Layering is the prototype's four-layer contract, mechanically enforced with
import-linter: `main` (entrypoint + REST BFF) → `domains` → `platform`;
imports flow only downward; domains are imported via their package surface
(`__all__`), never by reaching into modules.

- **Admin SPA** — the four surfaces of §5; owns no rules. Renders what the
  BFF returns; must never enforce a business rule the backend doesn't (the
  UX-level one-storefront affordance is presentation; the rule is INV-4's
  service guard).
- **REST BFF (`main.py`)** — screen-shaped JSON endpoints, one round-trip per
  screen; translates HTTP ↔ domain calls; owns no business logic (INV-6).
- **`accounts` domain** — owns identity: account records, one-time-code
  issue/verify (INV-2, INV-3), session establishment, the signed-in-user
  gate. Must never know what a storefront is.
- **`storefronts` domain** — owns the storefront entity and the
  account↔storefront membership (INV-4, INV-5), including the entry-routing
  answer ("which storefront, if any, does this account have?"). Must never
  mint identity.
- **`platform/db`** — SQLite WAL connection + forward-only migration runner
  (INV-7); owns the schema lifecycle, no business rules.
- **`platform/mailer`** — a one-method port (`send(to, subject, body)`) with
  two adapters: `LogMailer` (localhost/dev and tests — the message lands in
  the app log / a test inbox) and `SmtpMailer` (PPE/Prod — relay coordinates
  from deployment config, INV-8). Chosen by configuration at startup.
- **`platform/deps`** — FastAPI dependency wiring: per-request DB connection,
  current-session resolution, configured mailer.

Sessions are signed HTTP-only cookies (server secret from configuration; no
session table). MVP-honest: signing out clears the cookie; a leaked signing
secret is rotated by replacing the configured secret, which invalidates all
sessions at once. Revisit (server-side sessions) when staff/multi-device
management arrives.

### 6.3 Data model & ownership

| Entity | Owned by | Key fields | System of record |
| --- | --- | --- | --- |
| `account` | accounts | `id`, `email` (unique, normalized — INV-2), `created_at` | SQLite |
| `auth_code` | accounts | `id`, `email`, `code_hash`, `expires_at`, `attempts`, `consumed_at`, `created_at` (INV-3) | SQLite |
| `storefront` | storefronts | `id`, `name`, `created_at` | SQLite |
| `storefront_membership` | storefronts | `account_id` (FK), `storefront_id` (FK), `role` (MVP: always `'owner'`), `created_at`; PK `(account_id, storefront_id)` | SQLite |

Notes:

- **Data minimization (OHM):** the account stores the email and nothing else —
  no name, no phone, no profile (corpus sign-up asks for nothing more;
  prototype precedent). A display name arrives only when a Feature needs it.
- `auth_code` rows are keyed by *email*, not account, because the code is
  issued before the account may exist (sign-up and log-in converge, PUC-3).
  Consumed/expired rows are purged opportunistically on issue.
- `storefront_membership.role` is a single-value enum today; it is the seam
  where the corpus's staff & permissions model (13.15.\*) attaches later
  without reshaping ownership (INV-4 rationale).
- `storefront.name`: when the merchant leaves it blank, a generated default
  (e.g. "⟨email local-part⟩'s storefront") is stored — the column is always
  populated; "unnamed" is a naming choice, not a NULL (corpus 14.01.0026).

### 6.4 Interfaces & contracts

All JSON over `/api/*`; session via signed cookie. Errors share one shape:
`{"error": {"code": "<machine-code>", "message": "<honest human text>"}}`.

| Endpoint | In | Out | Errors |
| --- | --- | --- | --- |
| `POST /api/auth/request-code` | `{email}` | `204` — code issued & dispatched. Response is identical whether the email is new or known (no account enumeration). | `400 invalid_email` · `429 resend_cooldown` (with `retry_after_s`) · `502 delivery_failed` (INV-9 — the relay refused; dev `LogMailer` cannot fail) |
| `POST /api/auth/verify` | `{email, code}` | `200 {account: {email}, storefront: {id, name} \| null, created: bool}` — session cookie set; `created` drives the honest welcome copy (PUC-3); `storefront` drives entry routing (§6.5). | `400 code_mismatch` (with `attempts_remaining`) · `400 code_expired` · `400 code_exhausted` |
| `GET /api/auth/me` | — | `200 {account: {email}, storefront: {id, name} \| null}` | `401 unauthenticated` |
| `POST /api/auth/logout` | — | `204` — cookie cleared | — (idempotent) |
| `POST /api/storefronts` | `{name?: string}` | `201 {id, name}` — membership row created with `role='owner'` | `401 unauthenticated` · `409 already_owns_storefront` (INV-4, PUC-7) |
| `GET /healthz` | — | `200 {status: "ok"}` — process up, DB reachable, migrations current | `503` otherwise (deploy gate, §6.9) |

Contract stance: these are the stable seams the SPA and the tests bind to;
adding fields is free, renaming/removing is a spec change (this document,
§3.1-handbook).

### 6.5 Per–Product-Use-Case design

#### Entry routing (the shared spine: PUC-1, PUC-5, PUC-6)

One client-side rule, fed by one server answer, decides where any arrival
lands. The server's answer is `GET /api/auth/me` (or the `verify` response,
which carries the same shape):

| Session | Storefront | Lands on |
| --- | --- | --- |
| none | — | Landing (PUC-1) |
| valid | `null` | Create storefront (PUC-5) |
| valid | present | Admin shell (PUC-6) |

The rule is exhaustive — no state lands nowhere (BUC-4's "never stranded";
OHM: Duty of Care). The SPA router applies it on load and after `verify`.

#### PUC-2 / PUC-3 — Sign up & log in (one flow; realizes BUC-1, BUC-2)

```mermaid
sequenceDiagram
  actor V as Visitor
  participant SPA as Admin SPA
  participant BFF as REST BFF
  participant ACC as accounts
  participant M as mailer (port)
  participant DB as SQLite

  V->>SPA: enter email, "Send code"
  SPA->>BFF: POST /api/auth/request-code {email}
  BFF->>ACC: request_code(email)
  ACC->>DB: upsert auth_code (hash, expiry; cooldown check)
  ACC->>M: send(email, "Your ecomm code: …")
  ACC-->>BFF: ok (uniform, new-or-known)
  BFF-->>SPA: 204 → code-entry step
  V->>SPA: enter code, "Continue"
  SPA->>BFF: POST /api/auth/verify {email, code}
  BFF->>ACC: verify(email, code)
  ACC->>DB: check hash, expiry, attempts; consume (INV-3)
  ACC->>DB: get-or-create account by normalized email (INV-2)
  ACC-->>BFF: account, created?
  BFF->>SF: storefront_for(account) — via domains, for the routing answer
  BFF-->>SPA: 200 {account, storefront|null, created} + session cookie
  SPA-->>V: honest welcome → entry routing
```

- **Implementation:** `request_code` is uniform for new vs known emails (no
  enumeration, §6.6) and enforces the 60s cooldown (INV-3 → PUC-2c).
  `verify` consumes the code transactionally — increment attempts on
  mismatch (PUC-2a), refuse expired (PUC-2b), invalidate at 5 (INV-3) — and
  only then gets-or-creates the account (INV-2). Account creation *is* email
  verification: the address provably received the code. Honors INV-1 (works
  on an empty DB: the first verify creates account #1), INV-6, INV-9.

#### PUC-4 / PUC-7 — Create the storefront (realizes BUC-3, BUC-3a)

```mermaid
sequenceDiagram
  actor Mer as Merchant
  participant SPA as Admin SPA
  participant BFF as REST BFF
  participant SF as storefronts
  participant DB as SQLite

  Mer->>SPA: name (optional), "Create storefront"
  SPA->>BFF: POST /api/storefronts {name?}
  BFF->>SF: create_storefront(account, name?)
  SF->>DB: any membership for account? (INV-4 guard)
  alt already owns one
    SF-->>BFF: refuse
    BFF-->>SPA: 409 already_owns_storefront (PUC-7, honest copy)
  else none yet
    SF->>DB: insert storefront (name or generated default) + membership(owner)
    SF-->>BFF: storefront
    BFF-->>SPA: 201 {id, name}
    SPA-->>Mer: → admin shell (PUC-6)
  end
```

- **Implementation:** guard + insert run in one transaction (two concurrent
  creates cannot both pass the guard). Default naming per §6.3. No payment,
  plan, or extra field exists to ask for (corpus 14.01.0015/0016). Honors
  INV-4, INV-5, INV-6.

#### PUC-8 — Admin shell (realizes BUC-4)

No new backend surface: the shell renders from `GET /api/auth/me`
(storefront name + account email) and is otherwise static honest-empty copy
(§5.4). Honors INV-9 — it claims nothing that doesn't exist.

#### PUC-9 — Sign out

`POST /api/auth/logout` clears the session cookie; SPA returns to Landing.
Idempotent; no server state to destroy (sessions are cookie-borne, §6.2).

#### PUC-10 / PUC-11 — Bootstrap (realizes BUC-5, BUC-5a)

```mermaid
sequenceDiagram
  participant Op as Operator / Developer
  participant App as ecomm process
  participant DB as SQLite (empty file or absent)
  actor Mer as First merchant

  Op->>App: start (dev script · flotilla deploy)
  App->>DB: open/create; apply pending migrations in order (INV-7)
  App-->>Op: /healthz 200 (migrations current)
  Mer->>App: sign up → create storefront → admin (the ordinary PUC-2/4/6 flows)
  Note over App,Mer: first rows created through the product alone (INV-1)
```

- **Implementation:** identical in all three environments; only configuration
  differs (INV-8): **localhost** — `scripts/dev.sh` from a clean checkout
  (venv + deps, backend on :8000, Vite on :5173 proxying `/api`); mailer is
  `LogMailer`, so the one-time code appears in the backend log (the "local
  dev channel" of PUC-10 — auth itself is not stubbed; the same issue/verify
  path runs). **PPE/Prod** — provisioned and deployed through
  flotilla/launch-app (operator gesture, handbook §8.5); mailer is
  `SmtpMailer` with relay coordinates and secrets resolved from Secret
  Manager via deployment config; migrations apply at startup under
  flotilla's fail-stop deploy, gated by `/healthz`. The bring-up runbook
  (per-environment: prerequisites, the one gesture, how to watch `/healthz`)
  ships in-repo as `docs/BOOTSTRAP.md` — the documented gesture BUC-5
  demands.

### 6.6 Non-functional requirements & cross-cutting concerns

<!-- §6.6 pending -->

### 6.7 Key decisions & alternatives considered

<!-- §6.7 pending -->

### 6.8 Testing strategy

<!-- §6.8 pending -->

### 6.9 Failure modes, rollback & flags

<!-- §6.9 pending -->

## 7. Delivery Plan

### 7.1 Approach / strategy

<!-- §7.1 pending -->

### 7.2 Slicing plan

<!-- §7.2 pending -->

### 7.3 Rollout / launch plan

<!-- §7.3 pending -->

### 7.4 Risks & mitigations

<!-- §7.4 pending -->

## 8. Traceability matrix

<!-- §8 pending -->

## 9. Open Questions & Decisions log

<!-- §9 pending -->

## 10. Glossary & References

<!-- §10 pending -->