AI Workshop: Luzmo IQ Streaming Implementation

This project demonstrates how to create an "Executive Summary" generated by Luzmo IQ, accompanied by an embedded dashboard. The project is mainly created for use during a workshop at Luzmo's Data Talks event in Paris.

• AI-Powered Business Analytics: Automatically generates an AI summary from your sales data using Luzmo IQ, with live status updates
• Interactive Dashboard: Embeds a Luzmo dashboard for deeper data exploration

📁 Project Structure

data-talks_ai-workshop/
├── iq_summary_with_embedded_dashboard.html    # Main demo application
├── neon_db_workshop_sales_schema_and_data.sql # Database setup with sample sales data
└── workshop_sql-dataset-definition.sql        # SQL query for denormalized Luzmo dataset

🚀 Quick Start Guide

1. Set Up Neon Database (Serverless PostgreSQL)

You can skip this if you'd like to connect your own data source to Luzmo!

1. Create Neon Account

   • Go to https://console.neon.tech/signup
   • Sign up for a free account

2. Set Up Database Schema & Data

   • In your Neon console, go to the SQL Editor
   • Copy and paste the entire content of neon_db_workshop_sales_schema_and_data.sql
   • Execute the script to create tables, insert sample data, and set up a read-only user

3. Get Connection Details

   • In Neon console, go to Dashboard → Connection Details
   • Note down your connection string (make sure to swith to the "Role" luzmo_readonly_user)

2. Set Up Luzmo Account

1. Start Free Trial

   Sign up for a free trial account (or log in to your existing Luzmo account)

2. Connect Your Data Source

   • In Luzmo, click "Connect" to add a new data source
   • Select the PostgreSQL connector
   • Use the connection details from Neon:
     • Host: Your Neon host (e.g., ep-cool-name-123456-pooler.us-east-1.aws.neon.tech)
     • Port: 5432
     • Database: neondb (or your database name)
     • Username: luzmo_readonly_user
     • Password: SuperSecretPassword123
   • Create the connection

3. Create Dataset

   • In Luzmo, create a new SQL Dataset
   • Use the query from workshop_sql-dataset-definition.sql:

   SELECT
     po.*,
     -- Case statement to map sales codes into meaningful descriptions
     CASE 
         WHEN po.sales_code = 'S001' THEN 'No discount'
         WHEN po.sales_code = 'S020' THEN 'Vendor discount'
         WHEN po.sales_code = 'S500' THEN 'End-of-day discount'
         ELSE 'Other / Unknown'
     END AS sales_code_description,
     c.customer_first_name,
     c.customer_last_name,
     c.customer_email,
     c.customer_date_of_birth,
     p.product_name,
     p.product_category,
     p.product_release_date,
     s.store_name,
     s.store_country
   FROM product_orders po
   JOIN customers c ON po.customer_id = c.customer_id
   JOIN products p ON po.product_id = p.product_id
   JOIN stores s ON po.store_id = s.store_id

4. Create a Dashboard

   • Create a new dashboard in Luzmo and add some charts using your dataset (e.g., revenue by store, top products)
   • Note down your Dashboard ID (found in the dashboard URL, or dashboard settings at the bottom)

3. Generate Embed Token

1. Generate token via Embed button

   • In your Luzmo dashboard, click the "Embed" button
   • Copy the Embed Key and Embed Token
   • Important: These are temporary tokens for demo purposes and are only valid for 30 minutes!

2. Recommended: Generate via API
   In production setups, Embed Authorization tokens are requested server-side in your application; for the sake of this demo application, we can use tools like Postman or Insomia (or just plain curl requests) to manually generate an Embed key-token pair. This might be necessary if you'd like to e.g. specify extra context for Luzmo IQ (somewhat like a "Master prompt").

   1. Create a Luzmo API key-token pair in your account settings: https://app.luzmo.com/settings/api-tokens
   2. Use your API key-token pair to generate a temporary Embed key-token pair via an HTTP request to the /authorization API endpoint (see this guide). Make sure to give access to both your dashboard and dataset (e.g. via a Collection).

   curl https://api.luzmo.com/0.1.0/authorization  -H "Content-Type: application/json" -d @- << EOF
    {
      "action": "create",
      "version": "0.1.0",
      "key": "<your Luzmo API key>",
      "token": "<your Luzmo API token>",
      "properties": {
        "type": "embed",
        "username": "luzmo-iq-embed@example.com",
        "name": "Embed Luzmo IQ Test User",
        "suborganization": "Company A",
        "email": "luzmo-iq-embed@example.com",
        "access": {
          "collections": [
            {
              "id": "<collection_id>",
              "inheritRights": "use"
            }
          ]
        }
      }
    }
    EOF

4. Configure the HTML File

Open iq_summary_with_embedded_dashboard.html and update the following values in the CONFIG object (around line 84):

const CONFIG = {
  // Replace with your actual embed credentials
  authKey: "YOUR_EMBED_KEY_HERE",
  authToken: "YOUR_EMBED_TOKEN_HERE",
  dashboardId: "YOUR_DASHBOARD_ID_HERE"
};

You can also customize the "Executive Summary" prompt by modifying BUSINESS_SUMMARY_PROMPT in the HTML file (right below the CONFIG object):

const BUSINESS_SUMMARY_PROMPT = `Your custom prompt here...`;

5. Run the Application

Simply open iq_summary_with_embedded_dashboard.html in your web browser. The application will automatically start generating AI insights while allowing you to already slice-and-dice in the embedded dashboard!

🔧 Technical Details

Luzmo IQ Streaming API

The project uses Luzmo's Message streaming API (/ai/v1/message) to get real-time AI responses:

• Streaming States: Analyzing → Querying → Summarizing → Completed
• Response Modes: text_only (used in this demo), chart_only, or mixed

🛠️ Troubleshooting

Common Issues

1. "Failed to load dashboard"

   • Check that your embed key-token pair is correctly set
   • Ensure your dashboard ID is valid
   • If you're not on the EU multitenant VPC, make sure to adapt the API Host URL and App Server URL in the CONFIG object

2. "No data found"

   • Confirm your Neon database has data
   • Check that your Luzmo dataset is properly configured
   • Verify the SQL query in your dataset

3. "Connection failed"

   • Verify your connection string is correct
   • Check that the read-only database user has the proper permissions to query the schema, tables, etc.

Getting Help

• Luzmo Documentation: https://developer.luzmo.com
• Luzmo Academy: https://academy.luzmo.com
• Neon Documentation: https://neon.tech/docs