Fabric IQ Doc 3 — Product & Build Plan

0. Two-line summary

Fabric IQ is a new multi-tenant garment SaaS (fabriciq.ai / xyz.fabriciq.ai) that rebuilds FactoryTracking’s MES on Postgres and ports FactoriQ’s Buy/Demand/Risk IQ, WhatsApp, and GST patterns — sold at ₹5,000/month (yearly ~₹50,000) with a hard access gate. First paid MVP is achievable in ~8 weeks; full vision ~20–24 weeks.

Can we build it? Yes — shop-floor E2E already runs locally; subscription E2E is standard Razorpay + webhook + tenant provision; WhatsApp/GST need Meta/NIC lead time but are proven patterns in FactoriQ.

1. Product definition

Two-line summary

One product for Indian garment/fabric manufacturers: buy materials intelligently, make on the floor with QR, talk on WhatsApp, invoice buyers with GST, and pay you a subscription for the software.

Positioning table

Pricing (initial)

• Monthly: ₹5,000 / factory / month (all Phase-1 features).
• Yearly: ~₹50,000 / year (≈2 months free — finalize commercially).
• Trial: 14 days, then paywall (hard).
• Later: add-ons (extra plants, WhatsApp high volume, GST live NIC support).

2. Architecture (chosen)

Two-line summary

New monorepo — Next.js UI + Express/TypeScript API + PostgreSQL + Redis — FactoriQ-style tenancy; do not ship the Flask monolith.

flowchart LR
  subgraph public [Public]
    Mkt["fabriciq.ai marketing"]
    Rz[Razorpay]
  end
  subgraph tenant [Tenant]
    Sub["xyz.fabriciq.ai"]
    Phone["Floor phones"]
    WA["WhatsApp Cloud"]
  end
  subgraph plat [Platform]
    FE["Next.js"]
    API["Express API"]
    PG[(Postgres)]
    Redis[(Redis)]
    AI["AI service"]
  end
  Mkt --> Rz
  Rz --> API
  Sub --> FE
  Phone --> FE
  FE --> API
  API --> PG
  API --> Redis
  API --> WA
  API --> AI

Tenancy

• Resolve host xyz.fabriciq.ai → tenants.slug = xyz → tenant_id on every query.
• Fallback path /t/xyz if DNS/TLS not ready.
• Middleware: auth + subscription_active or trial not expired — else redirect billing.

Data

• Postgres as system of record (Neon/Supabase/RDS).
• Schema starts from Doc 1 tables, renamed company_id → tenant_id for consistency with FactoriQ patterns.
• Migrations via Prisma or Drizzle (pick one and stick).

Why not evolve Flask?

Untested PG shim, no tests, secrets hygiene failure, Jinja UI bar, and FactoriQ modules are already TS/Python services — rewrite cost of MES domain is lower than forever maintaining the monolith.

3. End-to-end: SaaS subscription (buildable)

Two-line summary

Owner picks plan on marketing site → Razorpay charges → webhook creates tenant + admin → DNS already covers wildcard → owner logs into xyz.fabriciq.ai; every request checks subscription.

Detailed flow

sequenceDiagram
  participant Owner as FactoryOwner
  participant Web as MarketingSite
  participant Pay as Razorpay
  participant API as FabricAPI
  participant App as TenantApp

  Owner->>Web: View pricing Monthly or Yearly
  Owner->>Web: Start trial or Pay now
  alt Pay now
    Web->>Pay: Create subscription
    Pay->>Owner: Pay with UPI or card
    Pay->>API: Webhook subscription activated
  else Trial
    API->>API: Create tenant with 14 day trial
  end
  API->>API: Insert tenant slug and admin user
  API->>Owner: Email login link
  Owner->>App: Login on subdomain
  App->>API: Check subscription active
  alt Active or trial valid
    API-->>App: Allow MES
  else Lapsed
    API-->>App: Redirect to billing
  end

Implementation checklist

1. Razorpay plans: plan_monthly_5000, plan_yearly_50000.
2. Webhook signature verify; idempotent tenant create.
3. Tables: tenants, subscriptions, subscription_events.
4. Hard middleware on all app routes except marketing, signup, webhook, billing.
5. Customer portal link to manage payment method.
6. Dunning: WhatsApp/email 3 days before trial end (port FT daily-notifications idea).

Trial → paid conversion

Same checkout; on success set subscription_status=active, clear trial lock. No soft access after expiry.

4. End-to-end: factory shop floor (target Fabric IQ)

Two-line summary

Same MES loop as FactoryTracking, plus WhatsApp status pings and (Phase 3) GST invoice after dispatch.

Detailed flow

flowchart TD
  O["Customer order entered"] --> C["Create batch at CUTTING"]
  C --> Q["Print QR label"]
  Q --> S["Operator scans phone"]
  S --> MV["Move with sent qty"]
  MV --> WA1["WhatsApp stage alert"]
  MV --> RV["Receive with received qty"]
  RV --> ALERT{"Qty mismatch?"}
  ALERT -->|yes| FLAG["ALERT on dashboard"]
  ALERT -->|no| OK[OK]
  OK --> MORE{"More stages?"}
  FLAG --> MORE
  MORE -->|continue| MV
  MORE -->|QC| DEF["Log defect types"]
  DEF --> MORE
  MORE -->|PACK| PACK["Packing sheet"]
  PACK --> DISP[DISPATCH]
  DISP --> WA2["WhatsApp buyer Ready"]
  DISP --> GST["GST tax invoice PDF"]
  DISP --> REP["TV and reports"]

Stage pipeline (unchanged domain)

flowchart LR
  CUTTING --> STITCHING --> WASHING --> QC --> PACKING --> DISPATCH

Roles on the floor

5. End-to-end: Buy + inventory (Phase 2 target)

Two-line summary

Demand or low stock creates a requisition; Buy IQ drafts a PO to a scored mill; WhatsApp sends it; GRN receives fabric into inventory; cutting consumes BOM.

flowchart TD
  LS[Low stock or Demand IQ] --> RQ[Requisition]
  RQ --> BI[Buy IQ vendor score]
  BI --> PO[Purchase order]
  PO --> WA[WhatsApp to mill]
  WA --> GRN[Goods receipt]
  GRN --> INV[Inventory ledger]
  INV --> BOM[BOM consume at CUTTING]
  BOM --> MES[Shop floor MES]

6. Phased delivery & timeline

First revenue: end of Phase 1 (~8 weeks). Full vision as described: ~20–24 weeks with 1–2 senior engineers + you.

Phase 0 detail

• Create FabricIQ monorepo (apps/web, apps/api, packages/db).
• Tables: tenants, users, memberships, subscriptions, factory_stages, production_batches, movement_history, scan_logs, inventory_*, customer_orders (minimum).
• Cloudflare fabriciq.ai + wildcard *.fabriciq.ai.
• CI: lint, typecheck, tenant-isolation tests.

Phase 1 detail (sellable)

1. Onboarding + roles
2. MES recreate (create/move/receive/QR/defects/targets/dashboards)
3. Inventory + style BOM
4. Orders + thin buyer portal
5. Razorpay monthly/yearly + hard gate
6. WhatsApp Cloud: low_stock + stage_move + daily_brief templates
7. Marketing site pricing page

Out of Phase 1: full BuyIQ, Demand/Risk, GST IRN, roll tracking, native app.

Phase 2–4

As in Doc 2 priority table (Buy IQ → WhatsApp inbound → Demand/Risk/GST → polish).

7. Target schema notes (Postgres)

Two-line summary

Port every Doc 1 table; rename company_id → tenant_id; add subscriptions and WhatsApp/integration tables; add fabric attributes on items when Phase 2 starts.

New tables beyond FT

Fabric item extensions (Phase 2)

On inventory_items or fabric_rolls: gsm, width_cm, composition, shade, lot_number, roll_barcode, meters.

8. Deployment strategy

Two-line summary

Cloudflare DNS/TLS in front; managed Postgres; app on Railway/Render/Fly or a VPS; staging + prod; never commit secrets.

9. Engineering rules

• Tenant isolation on every query (automated tests).
• No secrets in git; rotate FT-exposed keys before demos.
• MES domain tests first: move/receive qty, stage permissions, subscription middleware.
• UI: one design system (Tailwind; FactoriQ look-and-feel OK).
• India-first: Razorpay, +91 WhatsApp, GST schema hooks early.
• Small PRs; do not paste entire app.py.

10. Risks & mitigations

11. Success metrics (first 90 days post-MVP)

12. Immediate next steps (after docs approved)

1. Confirm domain fabriciq.ai + Cloudflare.
2. Scaffold FabricIQ monorepo.
3. Schema ADR + first migrations.
4. Razorpay test subscription E2E.
5. Vertical slice: subdomain login → create batch → move → QR scan.
6. Start Meta Business verification in parallel.

13. FAQ

Is FactoryTracking deployed? No confirmed live URL; use as domain reference only. Is subscription E2E buildable? Yes — Razorpay + webhook + tenant + hard gate. Is shop-floor E2E buildable? Yes — already demoable locally; rebuild for production. Do we need FactoriQ monorepo as dependency? No — port patterns/modules into FabricIQ repos; keep products separate. HTML versions of these docs? Markdown is complete with Mermaid; say if you want matching single-file HTML next.

End of Doc 3.