AI Retail Arbitrage App (Walmart-first)

Goal

Build a mobile (iOS + Android) AI-powered retail arbitrage app that recommends what to buy right now using real-time inventory + pricing and ranks opportunities by profitability + urgency.

Non-Negotiables

• Mobile-first delivery (iOS + Android)
• Real-time data only (freshness validated)
• Walmart-first (expandable to Target/BestBuy/Home Depot/Lowe’s)
• Minimize user effort and decision fatigue
• Exploit every possible edge (speed, inventory collapse, pricing latency)
• No “background work” claims — only count explicit, testable artifacts

Research Status (LOCKED)

Retail arbitrage research is considered COMPLETE and binding:

• Paid Slack arbitrage groups = speed + aggregation, not magic data
• Walmart app/back-end tends to expose richer store-level inventory than the public site
• Clearance/rollback/shelf/app mismatches propagate with latency
• Low inventory + high resale velocity is a top ROI indicator
• Do not re-research unless invalidated by evidence

Current Phase

BUILD PHASE ACTIVE

System Architecture (Modules)

A. Data Acquisition Layer

• Store + ZIP selection
• UPC/EAN, Walmart item ID/SKU
• Live price + availability polling
• Change detector (diff-based)

B. Validation Layer

• Cross-check allowed sources
• Freshness rules (timestamps/TTL/retries)
• Confidence scoring

C. Profit Engine

• Amazon / eBay / pawn-local resale models
• Fees/shipping/taxes/risk buffers (conservative)

D. Ranking & Urgency Engine

• Score = profit × probability × urgency(low stock) × confidence
• Low-inventory indicators + “act now” thresholds

E. Alerts

• Push notifications
• Top 5 “right now”
• Watchlists + thresholds

F. Mobile App UI

• Chat-style input + structured results
• Dashboard of ranked opportunities

G. Learning Loop (later, not blocking v1)

• Outcomes + personalization + false-positive suppression

Tech Stack (Recommended)

• Mobile: React Native + TypeScript
• Backend API: FastAPI (Python)
• Workers/Jobs: Python workers (RQ/Celery or asyncio)
• Queue/Cache: Redis
• DB: Postgres
• Dev: Docker Compose

Repo Structure (Target)

arbitrage-app/ apps/mobile/ services/api/ services/workers/ packages/shared/ infra/ docs/

v1 Milestone (First Executable Demo)

• Store selection (ZIP/store ID)
• One live price/stock signal (even if limited)
• Profit engine v1 (conservative)
• Ranking v1
• Mobile “Top 5 right now” screen + push notification stub

How to Resume This Project in a New Chat

Paste this repo link and say: “Read the repo state and continue the build from the current phase.”

Repo: https://github.com/ajsupplycollc/ChatGPT

Next Task (Do This Next)

1. Create repo skeleton folders
2. Add docs/architecture.md, docs/build_plan.md, docs/data_sources.md
3. Start backend API skeleton (FastAPI)
4. Implement first live inventory/price signal + diff-based polling worker