JUNO

🚀 Modern commerce & AI‑powered support infrastructure for Shopify‑based dropshipping businesses.

Tech stack

---

JUNO is a full‑stack Next.js application that powers:

• A Vendor Portal for Shopify merchants
• A Supplier Portal for product providers
• An Admin view for operations
• An AI‑powered helpdesk and knowledge engine called JUNOCHAT / Juno Engine

Everything runs inside a single Next.js App Router project (frontend + backend + background jobs). 💻

---

📚 Table of Contents

1. High‑Level Overview
2. Core Features
   • Vendor Portal
   • Supplier Portal
   • Admin / Ops
   • JUNOCHAT (AI Helpdesk)
3. Architecture & Tech Stack
4. Problems & Solutions
5. Future Enhancements
   • JUNOCHAT & AI
6. Local Development
7. Docker & Deployment
8. AWS Bedrock & AI Setup
9. Prisma & Database
10. License

---

🧭 High‑Level Overview

JUNO centralizes communication and operations between vendors (Shopify stores) and suppliers:

• Vendors can:

  • Connect Shopify stores and sync products.
  • Receive products and price offers from suppliers.
  • See new products in a clean, auto‑refreshing New Products view.
  • Get notifications when suppliers change prices or withdraw products.

• Suppliers can:

  • Manage their product catalog with images and pricing.
  • Accept product sync requests from vendors.
  • Provide products and custom pricing per vendor.
  • Edit synced product price/quantity per vendor with inline controls.

• Both sides are supported by:

  • JUNOCHAT – an AI assistant powered by AWS Bedrock (Titan + Claude) that can answer questions based on the knowledge base and FAQs.
  • A Juno Engine area for Gmail and ticketing workflows.

---

✨ Core Features

🛍️ Vendor Portal

• Shopify connections

  • Connect one or more Shopify stores to JUNO.
  • Webhooks keep orders, inventory, and products in sync.

• New Products

  • Consolidated list of:
    • Products vendors requested and had synced by suppliers.
    • Products provided directly by suppliers via offers.
  • Clean row‑based layout (no heavy cards) that:
    • Auto‑refreshes in the background on an interval and when the tab becomes visible.
    • Avoids UI flicker by only updating when data actually changes.
  • Product detail sheet with:
    • Price
    • Supplier
    • Optional quantity/amount
    • Description and image
  • Clicking outside the detail closes it (no need to aim at a tiny “X”).

• Suppliers Directory

  • Discover global suppliers.
  • Send connection invitations.
  • Track connection status (PENDING / CONNECTED / REJECTED).

• Notifications

  • Vendors are notified when:
    • A supplier accepts a product sync request.
    • A supplier provides a new product.
    • A supplier updates the price or quantity of an offered/synced product.

🤝 Supplier Portal

• Vendor Connections

  • Accept invitations using secure tokens.
  • See all connected vendors with status filters.
  • Accept / reject product sync requests from vendors with a dedicated panel.

• Synced Products Per Vendor

  • For each vendor, suppliers see a synced products block:
    • Compact, scrollable list for large catalogs.
    • Shows image, title, SKU, description.
    • Inline edit controls for:
      • Price (for that vendor)
      • Available amount / quantity (for that vendor)
  • Updates flow through ProductSync so the vendor’s New Products view reflects the new price/amount.

• Provided Products (Offers)

  • Suppliers can:
    • Select products from their catalog.
    • Specify per‑vendor price and total quantity.
    • See all offers in a tidy list with image, SKU, and timestamps.
  • Edit offer:
    • Opens a modal (no page jump) to change price and quantity.
    • Sends a notification to the vendor about the update.
  • Withdraw offer:
    • Opens a custom confirmation dialog:
      • “You are about to withdraw {productName} from {vendorName}…”
    • On confirm, removes the offer and the product disappears from the vendor’s New Products list.

• Products Catalog

  • Manage own products (create/update/delete).
  • Upload images via:
    • S3 in hosted environments, or
    • Local public/uploads in pure local dev.
  • Safeguards:
    • Synced products from vendors are editable but not deletable.
    • UI labels clarify when a product is “Synced” and where it came from.

🛠️ Admin / Ops

• Admin page with an overview of:

  • Stores
  • Vendors
  • Suppliers
  • Juno Engine activity

• Email‑centric ops:

  • Gmail polling and webhook endpoints.
  • Email logs and templates.
  • MFA setup, notifications, and account hygiene flows.

🤖 JUNOCHAT (AI Helpdesk)

JUNOCHAT (a.k.a. Juno Engine chat) is the AI layer that:

• Uses AWS Bedrock:
  • Amazon Titan Text Embeddings V2 for vector embeddings.
  • Anthropic Claude 3 Haiku for chat and knowledge‑base answers.
• Supports:
  • FAQ ingestion and retrieval.
  • Knowledge document upload + chunking.
  • RAG‑style responses for store‑specific questions (orders, products, policies, etc.).
• Integrates with the rest of JUNO:
  • Can respond based on vendor/supplier context.
  • Ties into the Juno Engine Gmail / ticketing workflow so AI can assist with replies.

---

Future Enhancements

Future Enhancements for JUNOCHAT & AI

• Deeper Shopify awareness

  • Let JUNOCHAT answer questions like:
    • “Which supplier provides the top‑selling items in the last 30 days?”
    • “What is the margin per vendor on synced products?”

• Multi‑channel support

  • Extend Juno Engine beyond Gmail:
    • WhatsApp / SMS
    • Messenger / Instagram DMs
    • Shopify Inbox

• Active suggestion engine

  • Use embeddings + usage data to:
    • Suggest products suppliers should offer to specific vendors.
    • Recommend pricing bands per market/region.

• Self‑service playbooks

  • Let admins define reusable “AI playbooks”, e.g.:
    • “Handle delayed shipment” flow.
    • “Out‑of‑stock product” flow.
  • JUNOCHAT would follow these playbooks when generating replies.

---

Architecture & Tech Stack

• Framework: Next.js (App Router, TypeScript)
• UI: React + Tailwind‑style utility classes
• Auth: NextAuth
• Database: PostgreSQL via Prisma ORM
• Background / Integrations:
  • Gmail APIs (OAuth)
  • Shopify APIs (orders, products, inventory)
  • AWS Bedrock (Titan embeddings + Claude 3 Haiku chat)
• Deployment: Docker + GitHub Actions → AWS EC2

---

Problems & Solutions

1. Dropshipping is glued together with spreadsheets and chats 😵‍💫

Typical Shopify dropshipping today:

• Vendors discover suppliers in WhatsApp groups, email threads, or random marketplaces.
• Product info, pricing, and availability live in:
  • Google Sheets
  • Excel files
  • Long chat histories and voice notes
• When something changes (price increases, product discontinued), it’s easy to miss and very hard to trace.

How JUNO helps:

• Formal connections between Shopify stores and suppliers.
• Structured models (ProductSync, SupplierProductOffer) instead of ad‑hoc spreadsheets.
• Every new product or change becomes:
  • A record in the database.
  • A visible entry in the New Products list.
  • A notification to the affected party (vendor or supplier).

👉 Result: less “Did you see my message?” and more “The system already updated this.”

2. No single source of truth for cost, price, and margin per vendor 💸

In a typical Shopify setup:

• Supplier sends a Google Sheet with “latest prices”.
• Vendor copies that into Shopify, maybe with manual mark‑ups.
• If the supplier changes cost for one vendor, everyone else may still see old prices.
• Nobody is sure what the current agreed cost is for a given product/vendor pair.

How JUNO helps:

• Suppliers manage per‑vendor offers directly in the Supplier Portal:
  • Price and quantity are editable inline.
  • Synced products from vendors can be overridden per store.
• Vendors see the resulting price and available amount in one place:
  • The New Products page, refreshed in the background.
  • Notifications whenever suppliers change price/quantity.

👉 Result: vendors always see the latest cost per product per supplier, and suppliers stay in control of what each partner gets.

3. Operational noise across email, tickets, and knowledge 💬

Pain today:

• Support lives in Gmail, while operations live in Shopify, and documentation is buried in Notion/Docs.
• Agents manually copy/paste order IDs, tracking links, and policy snippets into replies.
• Repeated questions (“Where is my order?”, “What’s your return policy?”) steal time from real issues.

How JUNO helps:

• JUNOCHAT + Juno Engine:
  • Ingest FAQs and internal docs into a vectorized knowledge base (via Bedrock embeddings).
  • Answer questions using RAG against that knowledge.
  • Integrate with Gmail so tickets and AI context live in the same place.

👉 Result: fewer repetitive replies, faster accurate responses, and a single surface for “what we know” + “what customers ask”.

---

Local Development

1. Install dependencies

npm install

2. Create .env

Copy .env.example (or the variables listed below) into a local .env file and fill in your own values:

• DATABASE_URL
• NEXTAUTH_SECRET
• NEXTAUTH_URL
• GOOGLE_CLIENT_ID
• GOOGLE_CLIENT_SECRET
• SMTP_HOST
• SMTP_PORT
• SMTP_SECURE
• SMTP_USER
• SMTP_PASSWORD
• AWS_ACCESS_KEY_ID (for AWS Bedrock)
• AWS_SECRET_ACCESS_KEY (for AWS Bedrock)
• AWS_REGION (default: us-east-1)
• ADMIN_EMAIL
• ADMIN_PASSWORD

Never commit real secrets; .env* files are ignored by git.

3. Run the dev server

npm run dev

Visit http://localhost:3000 in your browser.

---

Docker & Deployment

The project can be run via Docker using the provided Dockerfile and docker-compose.yml.

Typical pattern:

docker compose up -d --build

Make sure your .env is present next to the app code so the container can read it.

CI/CD on AWS EC2

This repo includes a GitHub Actions workflow at .github/workflows/deploy-ec2.yml that:

1. Runs on pushes to the main branch.
2. Installs dependencies and builds the Next.js app.
3. SSHes into an EC2 instance.
4. Writes an .env file on the EC2 instance from a single GitHub secret EC2_ENV.
5. Uses docker compose / docker-compose to build and run the app.
6. Performs a health check and automatically rolls back to the previous commit if the deployment or health check fails.

Required GitHub Secrets

Set the following secrets in GitHub → Settings → Secrets and variables → Actions:

• EC2_HOST – public IP or DNS of the EC2 instance.
• EC2_USER – SSH user (e.g. ubuntu).
• EC2_SSH_KEY – private SSH key for that user.
• EC2_SSH_PORT – (optional) port, defaults to 22 if not set.
• EC2_ENV – multi-line secret containing the full contents of the .env file used in production.
  This is written to .env on the server during deployment.

Deployment Flow

• Push to main.
• GitHub Actions:
  • Builds the app.
  • Connects to EC2.
  • Writes .env from EC2_ENV.
  • Runs docker compose up -d --build.
  • Hits the health check URL (by default http://localhost:3000/api/health).
• If the health check fails, the workflow:
  • Resets the repo on EC2 to the previous commit.
  • Re-runs docker compose up -d --build.

This gives you a sharp deployment with an automatic rollback path.

---

AWS Bedrock & AI Setup

The knowledge base and JUNOCHAT features use AWS Bedrock for AI‑powered embeddings and chat completion.

Prerequisites

1. AWS Account: Create an account at https://aws.amazon.com/
2. Enable Bedrock Access:
   • Go to https://console.aws.amazon.com/bedrock/
   • Navigate to Model access
   • Request access to:
     • Amazon Titan Text Embeddings V2 (for embeddings)
     • Anthropic Claude 3 Haiku (for chat completion)

IAM Setup

Create an IAM user or role with Bedrock permissions:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": ["bedrock:InvokeModel"],
      "Resource": [
        "arn:aws:bedrock:*::foundation-model/amazon.titan-embed-text-v2:0",
        "arn:aws:bedrock:*::foundation-model/anthropic.claude-3-haiku-20240307-v1:0"
      ]
    }
  ]
}

Create access keys for that user and add them to .env:

AWS_ACCESS_KEY_ID="your-access-key-id"
AWS_SECRET_ACCESS_KEY="your-secret-access-key"
AWS_REGION="us-east-1"  # or your preferred region

Database Migration & Re‑Embedding

After setting up AWS Bedrock:

1. Apply Prisma migrations:

npx prisma migrate deploy

2. Re‑embed existing knowledge base data:

npm run re-embed-knowledge-base

This script:

• Reads all existing KnowledgeChunk records.
• Generates new embeddings using Titan V2.
• Updates the database in batches to avoid rate limiting.

---

Prisma & Database

To inspect the database locally:

export DATABASE_URL='<YOUR_DB_URL>'
npx prisma studio

---

License

Private / All rights reserved unless explicitly stated otherwise.