Prototype Demo

Run the working prototype locally and test personalization with sample data

GitHub Repository

Clone the repo to get started with the prototype implementation

Quick Start

Clone the repository and run the prototype server locally. The demo uses in-memory mock data to demonstrate personalization logic without requiring database setup.

Installation

# Navigate to prototype directory
cd prototype

# Install dependencies
npm install

# Start the server
npm start

Server will start on http://localhost:3001

Prerequisites

• Node.js 16 or higher
• npm or yarn package manager
• Terminal access

Mock Data Reference

Demo Users

All user IDs are SHA-256 hashed for privacy compliance.

User ID (Hashed)	Profile	Watch History
06d6cbdc...	Gaming Enthusiast	8 gaming videos
3e0e7f64...	Cooking Enthusiast	5 cooking videos
b1ee9a41...	Fitness Enthusiast	5 fitness videos
4911e04c...	Tech Enthusiast	5 tech videos
6f015e46...	New User (Cold Start)	No history

Tenants

Tenant ID	Personalization	Weights
tenant1	Enabled	50% watch, 30% engage, 20% editorial
tenant2	Enabled	70% watch, 20% engage, 10% editorial
tenant3	Disabled (Flag Off)	Editorial only

Video Categories

40+ mock videos across 8 categories:

GamingCookingFitnessTechMusicTravelComedyEducation

Try It Now

Copy and paste these curl commands to see different personalization scenarios in action.

Scenario 1: Gaming Enthusiast

User with 8 gaming video watches. Expects gaming content ranked highest.

curl "http://localhost:3001/v1/feed?user_id=06d6cbdcfc221d2f4460c17193442b9db221f30950f1c17af4e73e6e1788002b&tenant_id=tenant1&limit=10"

Scenario 2: Cooking Enthusiast

User with 5 cooking video watches. Expects cooking content ranked highest.

curl "http://localhost:3001/v1/feed?user_id=3e0e7f64a2495659941e0b704069bcb310d8dfcab850ba1aa992669ef6f55bcb&tenant_id=tenant1&limit=10"

Scenario 3: Cold Start (New User)

User with no watch history. Expects popular content with high editorial boost.

curl "http://localhost:3001/v1/feed?user_id=6f015e465db03f8a847292bf8624f567167f87f0ea0aa223d1e31779cad855c7&tenant_id=tenant1&limit=10"

Scenario 4: Feature Flag Disabled

Same gaming user, but tenant has personalization disabled. Expects editorial-only ranking.

curl "http://localhost:3001/v1/feed?user_id=06d6cbdcfc221d2f4460c17193442b9db221f30950f1c17af4e73e6e1788002b&tenant_id=tenant3&limit=10"

Scenario 5: Different Tenant Weights

Same gaming user with tenant2 (70% watch history weight vs 50% default). Notice different rankings.

curl "http://localhost:3001/v1/feed?user_id=06d6cbdcfc221d2f4460c17193442b9db221f30950f1c17af4e73e6e1788002b&tenant_id=tenant2&limit=10"

Run the Demo Script

For an automated walkthrough of all scenarios with formatted output, run the included demo script.

# Make sure server is running in another terminal, then:
cd prototype
./tests/demo.sh

The demo script will show all five scenarios with timing information and highlight the differences between personalized and non-personalized feeds.

Test Gradual Rollout

Test the percentage-based rollout feature to see how users are consistently bucketed into personalized vs. non-personalized experiences during gradual deployment.

# Make sure server is running in another terminal, then:
cd prototype
./tests/test-rollout.sh

What This Tests

• 50% rollout on tenant4: Half of users get personalized feeds
• Consistent bucketing: Same user always gets same experience
• Hash-based assignment: Uses user_id hash for deterministic splitting
• Gradual rollout strategy: Simulates 0% → 10% → 50% → 100% deployment

This demonstrates the safe rollout approach described in the Rollout & Observability documentation.