embeddings_analyzer

Edit in StackBlitz next generation editor ⚡️

Overview

This repository illustrates how segmenting lengthy documents and averaging their embedding vectors can affect similarity scores and search outcomes when using Azure OpenAI models.

Usage

• npm install
• npm run dev
• Provide your Azure OpenAI credentials (including the deployment name for a v3 large embedding model).
• Enter a lengthy text snippet (under 8k tokens).
• Enter a sample search query.
• Click “Go.”

The strategy dropdown is for visualization only—when you select “Go,” all strategies are evaluated and displayed side by side in the results table.