AI Seller Copilot for Small Ecommerce Sellers

AI Seller Copilot is a full-stack MVP that helps small ecommerce sellers upload Meesho, Flipkart, or Amazon order CSVs and answer:

What should I do today to increase profitable orders and reduce losses?

Tech Stack

• Frontend: React + Vite + Tailwind CSS
• Backend: FastAPI
• Data processing: pandas
• Database: SQLite
• Charts: Recharts
• AI: modular ai_service.py with fallback rule-based summary and listing generation

Features

• CSV upload and automatic cleaning for messy marketplace columns
• Dashboard summary with total, delivered, cancelled, RTO, revenue, top SKU, worst SKU, ad orders, and natural orders
• SKU score table with Push, Watch, and Pause actions
• RTO Risk Shield for states, SKUs, state + SKU combinations, and source-wise RTO
• Promotion recommender for SKUs to promote, pause, improve, and safe state/product combinations
• Daily AI business summary with rule-based fallback
• Listing Doctor for SEO title, description, bullets, keywords, and platform-specific text
• Daily action checklist stored in SQLite
• Inventory alerts using average daily orders
• Usage tracking for uploads, dashboard views, completed actions, listing generation, and stock updates

Setup

Backend

cd seller-copilot/backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn app.main:app --reload

Backend runs at http://localhost:8000.

Frontend

cd seller-copilot/frontend
npm install
npm run dev

Frontend runs at http://localhost:5173.

Sample CSV

Use:

sample_data/sample_orders.csv

The frontend also exposes the same sample at /sample_data/sample_orders.csv.

SKU Scoring Formula

For each SKU:

score = 50
  + delivered_rate * 30
  + normalized_order_volume * 20
  - rto_rate * 25
  - cancellation_rate * 15

Rates are decimal values from 0 to 1. normalized_order_volume is the SKU order volume divided by the highest SKU order volume in the uploaded file. The score is clamped between 0 and 100.

Actions:

• Push: score >= 75
• Watch: score >= 45 and < 75
• Pause: score < 45

RTO Risk Logic

High risk means:

• RTO rate >= 20%
• Minimum 5 orders in the segment

The app checks risk for states, SKUs, state + SKU combinations, and order sources.

AI Fallback

backend/app/services/ai_service.py checks for OPENAI_API_KEY or LLM_API_KEY. If no key exists, the app uses rule-based summary generation so the MVP works locally without external services.

backend/app/services/listing_generator.py follows the same fallback pattern for the Listing Doctor.

Future Improvements

• Real LLM integration for richer business recommendations
• PostgreSQL for multi-seller production usage
• Meesho, Flipkart, and Amazon platform API integrations
• WhatsApp daily report for sellers
• Advanced ML RTO prediction using customer location, price, SKU history, and source quality update 26531 update 29480 update 24593 update 32514 update 27629 update 26431 update 1460 update 1561 update 21548 update 25955 update 9813 update 10076 update 13998 update 17488 update 23045 update 12565 update 13728 update 28740 update 19845 update 23544 update 8532 update 19404 update 19738 update 13896 update 27953 update 26456 update 25112 update 6863 update 5247 update 26646 update 19437 update 18990 update 9996 update 6333 update 30662 update 27614 update 29514 update 2919 update 26939 update 14316 update 5676 update 28794 update 4774 update 28358 update 30505