MVP: Amazon as Initial Vendor Source #21

New issue

Closed

opened 2026-05-02 19:13:23 +00:00 by agent-cos · 2 comments

agent-cos commented

2026-05-02 19:13:23 +00:00

Update: Amazon as Initial Vendor for MVP

To prove the concept quickly, we'll use Amazon as the initial vendor source. This gives us:

Advantages

Massive selection to test curation algorithms
Fast shipping (Prime) for quick iteration
Easy returns (Amazon's return policy)
No vendor negotiation needed for MVP
Product data available via Amazon Product Advertising API or scraping

MVP Approach

Manual curation initially — Scott/Mary pick items from Amazon that match the style profile
Amazon affiliate links — generate revenue on kept items (2-4% commission)
System tracks — which items were suggested, which were kept, feedback ratings
Learn — use this data to train the matching algorithm

Transition Plan

Phase 1 (Now): Amazon for MVP, manual curation
Phase 2 (Working algorithm): Amazon API integration for automated suggestions
Phase 3 (Proven concept): Add boutique vendors, reduce/eliminate Amazon

Technical Needs

Amazon Product API access (or scraping fallback)
Way to store Amazon product IDs with style metadata
Affiliate link generation
Price tracking (Amazon prices fluctuate)

This is a smart pivot — prove the curation engine works before investing in vendor relationships.

## Update: Amazon as Initial Vendor for MVP To prove the concept quickly, we'll use Amazon as the initial vendor source. This gives us: ### Advantages - Massive selection to test curation algorithms - Fast shipping (Prime) for quick iteration - Easy returns (Amazon's return policy) - No vendor negotiation needed for MVP - Product data available via Amazon Product Advertising API or scraping ### MVP Approach 1. **Manual curation initially** — Scott/Mary pick items from Amazon that match the style profile 2. **Amazon affiliate links** — generate revenue on kept items (2-4% commission) 3. **System tracks** — which items were suggested, which were kept, feedback ratings 4. **Learn** — use this data to train the matching algorithm ### Transition Plan - Phase 1 (Now): Amazon for MVP, manual curation - Phase 2 (Working algorithm): Amazon API integration for automated suggestions - Phase 3 (Proven concept): Add boutique vendors, reduce/eliminate Amazon ### Technical Needs - Amazon Product API access (or scraping fallback) - Way to store Amazon product IDs with style metadata - Affiliate link generation - Price tracking (Amazon prices fluctuate) This is a smart pivot — prove the curation engine works before investing in vendor relationships.

agent-cos commented

2026-05-02 19:20:09 +00:00

Author

Engineering Spec Update

Amazon Associates ID: greybeardhold-20
Domain: wardrobeai.scoble.me

Affiliate Link Format

Standard Amazon affiliate URL structure:

https://www.amazon.com/dp/{ASIN}?tag=greybeardhold-20

Or for search results pages:

https://www.amazon.com/s?k={search-term}&tag=greybeardhold-20

Required Disclosures

Footer on every page: "As an Amazon Associate I earn from qualifying purchases"
On box display page: "WardrobeAI is a participant in the Amazon Services LLC Associates Program"

Domain Setup

Deploy to Storj static hosting (consistent with speckl.scoble.me, scott.scoble.me)
CNAME: wardrobeai.scoble.me → storj custom domain
SSL/TLS via Storj

@engineering — use this affiliate ID for all Amazon product links in the MVP.

## Engineering Spec Update **Amazon Associates ID**: `greybeardhold-20` **Domain**: `wardrobeai.scoble.me` ### Affiliate Link Format Standard Amazon affiliate URL structure: ``` https://www.amazon.com/dp/{ASIN}?tag=greybeardhold-20 ``` Or for search results pages: ``` https://www.amazon.com/s?k={search-term}&tag=greybeardhold-20 ``` ### Required Disclosures - Footer on every page: "As an Amazon Associate I earn from qualifying purchases" - On box display page: "WardrobeAI is a participant in the Amazon Services LLC Associates Program" ### Domain Setup - Deploy to Storj static hosting (consistent with speckl.scoble.me, scott.scoble.me) - CNAME: wardrobeai.scoble.me → storj custom domain - SSL/TLS via Storj @engineering — use this affiliate ID for all Amazon product links in the MVP.

agent-cos commented

2026-05-02 19:22:18 +00:00

Author

Architecture Update: Split Deployment

Static Site (Storj)

Domain: wardrobeai.scoble.me
Purpose: Landing page, style form, box display, product browsing
Tech: Astro SSG or plain HTML/CSS/JS
Hosting: Storj static hosting (CNAME + TXT records)

Real-Time AI Chat Service

Purpose: Conversational style interview, live curation assistance
Tech: Node.js/Python + WebSocket or Server-Sent Events
Hosting: Railway, Fly.io, or Render (real-time capable)
Domain: api.wardrobeai.scoble.me or chat.wardrobeai.scoble.me

Integration Flow

User visits wardrobeai.scoble.me → static landing page
Clicks "Start Style Interview" → loads chat interface (iframe or JS embed) from real-time service
Chat service conducts AI-powered style interview, stores profile
Chat service redirects back to static site with profile ID
Static site displays curated box using data from API
"Buy on Amazon" links use greybeardhold-20 affiliate ID

Backend API (could be same as chat service or separate)

Store style profiles
Store product catalog (Amazon items with metadata)
Matching algorithm endpoint
Feedback capture endpoint

Database

PostgreSQL on Railway or Supabase (free tier sufficient for MVP)
Or SQLite if keeping it ultra-simple

This split keeps costs low (Storj static is cheap, chat service scales as needed) and separates concerns.

## Architecture Update: Split Deployment ### Static Site (Storj) **Domain**: `wardrobeai.scoble.me` **Purpose**: Landing page, style form, box display, product browsing **Tech**: Astro SSG or plain HTML/CSS/JS **Hosting**: Storj static hosting (CNAME + TXT records) ### Real-Time AI Chat Service **Purpose**: Conversational style interview, live curation assistance **Tech**: Node.js/Python + WebSocket or Server-Sent Events **Hosting**: Railway, Fly.io, or Render (real-time capable) **Domain**: `api.wardrobeai.scoble.me` or `chat.wardrobeai.scoble.me` ### Integration Flow 1. User visits `wardrobeai.scoble.me` → static landing page 2. Clicks "Start Style Interview" → loads chat interface (iframe or JS embed) from real-time service 3. Chat service conducts AI-powered style interview, stores profile 4. Chat service redirects back to static site with profile ID 5. Static site displays curated box using data from API 6. "Buy on Amazon" links use `greybeardhold-20` affiliate ID ### Backend API (could be same as chat service or separate) - Store style profiles - Store product catalog (Amazon items with metadata) - Matching algorithm endpoint - Feedback capture endpoint ### Database - PostgreSQL on Railway or Supabase (free tier sufficient for MVP) - Or SQLite if keeping it ultra-simple This split keeps costs low (Storj static is cheap, chat service scales as needed) and separates concerns.