Glossary — Agentic AI
What is Embedding?
1 min read Updated
An embedding is a dense vector representation of data in a continuous mathematical space, where semantic similarity is captured by vector proximity — fundamental to search, RAG, and AI reasoning.
WHY IT MATTERS
Embeddings are how AI systems understand meaning numerically. When you embed the word 'wallet,' you get a vector of hundreds of floating-point numbers. Similar concepts have vectors that are close together in this space.
This enables semantic search: instead of keyword matching, you find content by meaning. Embeddings underpin RAG systems, recommendation engines, classification, and clustering.
In the agent stack, they're used to retrieve relevant context, match user intent to tools, and classify transaction types.
FREQUENTLY ASKED QUESTIONS
Which embedding model should I use?
OpenAI's text-embedding-3, Cohere's embed-v3, and open-source models like E5 and BGE are all strong choices. The best depends on your domain and latency requirements.
What's the difference between token and sentence embeddings?
Token embeddings represent individual words and are internal to LLMs. Sentence embeddings represent entire passages as single vectors, used for search and retrieval.
Do embeddings capture financial meaning?
General embeddings capture broad semantic similarity. For financial-specific tasks, domain-adapted or fine-tuned embedding models may better capture nuances.