Embedding Visualization with LangChain + OpenAI

Generate text embeddings using OpenAI models through LangChain, reduce them to 3D with PCA, and plot them with Matplotlib so you can eyeball the semantic relationships between words.

Features

• Generate embeddings for arbitrary text using OpenAI's embedding models.
• Choose between:
  • text-embedding-3-large
  • text-embedding-3-small
  • text-embedding-ada-002 (default)
• Dimensionality reduction with PCA.
• 3D scatter plot saved as a high-resolution PNG.

Requirements

Install dependencies with:

pip install -r requirements.txt

You will also need an OpenAI API key.

Setup

1. Clone this repository:

   git clone https://github.com/MandalAutomations/Vector-Visualizer-OpenAI.git
   cd Vector-Visualizer-OpenAI

2. Create a .env file in the project root with your OpenAI API key:

   echo "OPENAI_API_KEY=your_api_key_here" > .env

3. (Optional) choose a different embedding model by editing EMBEDDING_MODEL near the top of main.py:

   EMBEDDING_MODEL = "text-embedding-ada-002"

4. Add the words/phrases you want to embed to words.txt, one per line:

   nfl
   football
   soccer
   basketball
   baseball

Usage

Run the script to generate embeddings and plot them:

python main.py

This will:

1. Read each line from words.txt and request an embedding for it.
2. Reduce the embeddings to 3D with PCA.
3. Save the visualization to 3d_plot_small.png in the project root.

Project Structure

.
├── main.py              # Main script — embeds words.txt and plots them
├── words.txt            # Words/phrases to embed (one per line)
├── requirements.txt     # Dependencies
├── 3d_plot_small.png    # Example output plot
└── .env                 # API key (not committed)

Customization

• Edit words.txt to change which words are compared.
• Change EMBEDDING_MODEL in main.py to swap models.
• Adjust the dpi argument inside plot_embeddings_3d to control output resolution.

Example Plot