Embedding Visualizer
Paste multiple texts and see them visualized as points in 2D embedding space. Understand semantic similarity, clustering, and distance between concepts. Great for debugging RAG pipelines.
⚠️ This is a demonstration visualization using hash-based pseudo-embeddings. Real embeddings (from models like text-embedding-3 or claude's embeddings) produce more meaningful semantic clusters. Hover over points to see similarity scores.
Frequently Asked Questions
What are text embeddings?
Text embeddings are numerical representations (vectors) of text that capture semantic meaning. Semantically similar texts produce vectors that are close together in the embedding space. They're used in RAG pipelines, semantic search, clustering, and recommendation systems.
What does cosine similarity mean?
Cosine similarity measures the angle between two vectors. A score of 1.0 means identical direction (very similar), 0.0 means perpendicular (unrelated), and -1.0 means opposite. For text, scores above 0.8 typically indicate very similar meaning.
How do I generate real embeddings?
Use OpenAI's text-embedding-3-small, Anthropic's embeddings via Amazon Bedrock, or open-source models like sentence-transformers (all-MiniLM-L6-v2) which run locally. For RAG applications, Nomic Embed and BGE models offer excellent performance.