LLM Video Analysis

An AI-powered YouTube video analysis application built with TanStack Start, Bun, and Google's Gemini models. Analyze video content with custom prompts to extract insights, summaries, and structured information.

🚀 Features

Video Analysis

• YouTube Video Analysis: Analyze any YouTube video with custom prompts
• AI-Powered Insights: Extract summaries, key points, and structured information
• Real-time Processing: See results as they're generated
• Flexible Prompts: Ask specific questions about video content
• Structured Output: Get organized sections with timestamps

Technical Features

• TanStack Start + Form: Modern React SSR with form validation
• Bun Runtime: Fast, efficient JavaScript runtime
• Type-Safe API: Zod schemas for request/response validation
• Stub Mode: Test without API keys using mock data
• Production-Ready Server: Optimized static asset serving with intelligent caching

📦 Installation

This project was created with TanStack Start:

bunx create-start-app@latest

Install dependencies:

bun install

🏃‍♂️ Development

Prerequisites

• Bun 1.3+ (install with curl -fsSL https://bun.sh/install | bash)
• Node.js 20+ (for compatibility)
• Google API Key (optional, for production mode)

Quick Start

1. Install dependencies:

   ~/.bun/bin/bun install

2. Copy environment variables:

   cp .env.local.example .env.local

3. Start development server:

   ~/.bun/bin/bun run dev

4. Open http://localhost:3000 in your browser

Environment Configuration

Create a .env.local file with the following variables:

# Stub Mode (default: true)
# Set to false to use real Google AI API
USE_STUB=true

# Google AI Configuration (required for production)
GOOGLE_API_KEY=your-gemini-api-key-here

# Vertex AI Configuration (optional, for advanced features)
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_CLOUD_LOCATION=us-central1
ENABLE_CHUNKING=false
SEGMENT_DURATION=180

# Model Configuration
DEFAULT_MODEL=gemini-2.0-flash-exp

Getting API Keys

1. Google AI (Gemini) API Key:

   • Visit Google AI Studio
   • Create a new API key
   • Add to .env.local as GOOGLE_API_KEY

2. Vertex AI (Optional):

   • Set up a Google Cloud Project
   • Enable the Vertex AI API
   • Configure authentication
   • Add project details to .env.local

🔨 Production Build

Build the application for production:

bun run build

🚀 Production Server with server.ts

Quick Start - Use in Your Project

You can easily use this production server in your own TanStack Start project:

1. Copy the server.ts file into your project root
2. Build your project with bun run build
3. Start the server directly with:

   bun run server.ts

Or add it to your package.json scripts:

{
  "scripts": {
    "start": "bun run server.ts"
  }
}

Then run with:

bun run start

Server Features

The server.ts implements a high-performance production server with the following features:

1. Intelligent Asset Loading

The server automatically decides which files to preload into memory and which to serve on-demand:

• In-Memory Loading: Small files (default < 5MB) are loaded into memory at startup
• On-Demand Loading: Large files are loaded from disk only when requested
• Optimized Performance: Frequently used assets are served from memory

2. Configuration via Environment Variables

# Server Port (default: 3000)
PORT=3000

# Maximum file size for in-memory loading (in bytes, default: 5MB)
STATIC_PRELOAD_MAX_BYTES=5242880

# Include patterns (comma-separated, only these files will be preloaded)
STATIC_PRELOAD_INCLUDE="*.js,*.css,*.woff2"

# Exclude patterns (comma-separated, these files will be excluded)
STATIC_PRELOAD_EXCLUDE="*.map,*.txt"

# Enable detailed logging
STATIC_PRELOAD_VERBOSE=true

Example Configurations

Minimal Memory Footprint

# Preload only critical assets
STATIC_PRELOAD_MAX_BYTES=1048576 \
STATIC_PRELOAD_INCLUDE="*.js,*.css" \
STATIC_PRELOAD_EXCLUDE="*.map,vendor-*" \
bun run start

Maximum Performance

# Preload all small assets
STATIC_PRELOAD_MAX_BYTES=10485760 \
bun run start

Debug Mode

# With detailed logging
STATIC_PRELOAD_VERBOSE=true \
bun run start

Server Output

The server displays a clear overview of all loaded assets at startup:

📦 Loading static assets from ./dist/client...
   Max preload size: 5.00 MB
   Include patterns: *.js,*.css,*.woff2

📁 Preloaded into memory:
   /assets/index-a1b2c3d4.js           45.23 kB │ gzip:  15.83 kB
   /assets/index-e5f6g7h8.css           12.45 kB │ gzip:   4.36 kB

💾 Served on-demand:
   /assets/vendor-i9j0k1l2.js          245.67 kB │ gzip:  86.98 kB

✅ Preloaded 2 files (57.68 KB) into memory
ℹ️  1 files will be served on-demand (1 too large, 0 filtered)

🚀 Server running at http://localhost:3000

Testing

This project uses Vitest for testing. You can run the tests with:

bun run test

Styling

This project uses Tailwind CSS for styling.

🚀 Deployment

Railway (Recommended)

This project is configured for one-click deployment to Railway:

1. Connect Repository: Push your code to GitHub
2. Deploy to Railway:

   railway login
   railway init -n llm-video-analysis
   railway up

3. Set Environment Variables in Railway dashboard:
   • USE_STUB=false (for production)
   • GOOGLE_API_KEY=your-api-key
4. Generate Domain:

   railway domain

Cloudflare Pages/Workers

For Cloudflare deployment (coming soon - see ticket 004):

• Static assets → Cloudflare Pages
• Server functions → Cloudflare Workers
• Use Cloudflare KV for caching

Manual Deployment

1. Build the application:

   bun run build

2. Run the production server:

   PORT=3000 bun server.ts

Linting & Formatting

This project uses eslint and prettier for linting and formatting. Eslint is configured using tanstack/eslint-config. The following scripts are available:

bun run lint
bun run format
bun run check

📋 Project Tickets

The tickets/ directory contains development tickets tracking the implementation:

• 001: MVP Landing Page ✅ - Video analysis form with validation
• 002: Server Analysis Stub ✅ - Typed server endpoint with mock responses
• 003: Analyzer Pipeline Integration - Port Python analyzer to TypeScript
• 004: Cloudflare Deployment - Production deployment configuration