Cringe Influencer RAG

A Next.js application that uses RAG (Retrieval Augmented Generation) to search through LinkedIn posts, create embeddings, and generate content in an authentic voice.

Live Walkthrough Video

🚀 Quick Start

Prerequisites

• Node.js (v22+)
• Yarn package manager
• OpenAI API key
• Pinecone account (free tier available)

Step 1: Clone and Install

git clone https://github.com/yourusername/cringe-influencer-rag.git
cd cringe-influencer-rag
yarn install

Step 2: Set Up Environment Variables

Create a .env file in the project root:

# Required for vector database
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_INDEX_NAME=your_index_name

# Required for OpenAI embeddings and search
OPENAI_API_KEY=your_openai_api_key

Step 3: Set Up Pinecone (Free Tier)

1. Sign up at Pinecone (free tier includes 1 index)
2. Create a new index with these settings:
   • Name: Choose a name (use this as PINECONE_INDEX_NAME in .env)
   • Dimensions: 512 (for OpenAI text-embedding-3-small)
   • Metric: cosine
   • Cloud: Any region (e.g., aws/us-east-1)
3. Copy your API key from the Pinecone console to your .env file

Step 4: Get OpenAI API Key

1. Sign up at OpenAI Platform
2. Navigate to API Keys
3. Create a new API key and copy it to your .env file
4. Add $5 credit to your account (OpenAI offers pay-as-you-go pricing)
   • Go to Billing
   • Click "Add to credit balance"
   • Add $5 (this is enough for thousands of embeddings and queries)

Step 5: Upload Vectors to Pinecone

yarn upload

This uploads pre-generated vectors from output/brian_posts_vectors.json to your Pinecone index. Please use the correct index name.

Step 6: Run the Application

yarn dev

Visit http://localhost:3000 to use the application.

🔍 Using the Application

1. Enter a search query in the text area (e.g., "AI startup advice")
2. Click the SEARCH button
3. View results from both basic vector search and re-ranked results
4. Compare how the re-ranking improves search relevance

📁 Project Structure

• app/ - Next.js application files
  • api/ - API routes for search and re-ranking
  • components/ - React components
• data/ - Source data files
• libs/ - Utility libraries for OpenAI and Pinecone
• scripts/ - Scripts for embedding and uploading vectors
• output/ - Generated vector files

🛠️ Available Scripts

• yarn dev - Start development server
• yarn build - Build for production
• yarn start - Start production server
• yarn embed - Generate embeddings from source data
• yarn upload - Upload vectors to Pinecone

📚 Learning Resources

For those interested in the technology behind this application:

• Essence of Linear Algebra - Visual introduction to vectors
• Neural Networks - How neural networks work
• Transformers, explained - Understanding the transformer architecture

Join Parsity.io if you want to learn the skills to create production-grade full stack AI applications.