Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/simonw/LLM/llms.txt

Use this file to discover all available pages before exploring further.

Embedding models convert a piece of text — a word, sentence, paragraph, or entire article — into an array of floating point numbers called an embedding vector. This vector is a numerical representation of the semantic meaning of the content in a high-dimensional space. By calculating the distance between two embedding vectors, you can determine how semantically similar two pieces of content are, even if they share no exact keywords.

​
What embeddings enable

  • Semantic search — users search by meaning rather than exact keywords; results match the intent of a query.
  • Related content — find articles, documents, or records that are conceptually close to a given item.
  • Image search — models like CLIP operate on binary inputs such as images, enabling image-to-image or text-to-image similarity queries.
  • Clustering and classification — group or label content based on the geometric structure of its embedding space.

​
How LLM supports embeddings

LLM supports multiple embedding models through its plugin system. Once an embedding model plugin is installed, you can use it on the command line or through the Python API to:
  1. Calculate embedding vectors for individual strings or binary files.
  2. Store those vectors in named collections inside a SQLite database.
  3. Run similarity searches across a collection to find the nearest neighbors to a query.
Embeddings are stored as compact binary blobs (little-endian 32-bit floats) in SQLite. See the Storage Format page for technical details.
For an extended explanation of embeddings and why they are useful, see LLM now provides tools for working with embeddings by Simon Willison.

​
Explore the embeddings section

CLI Reference

Use llm embed, llm embed-multi, llm similar, and llm embed-models from the terminal to generate, store, and search embeddings.

Python API

Work with the llm.Collection class to store and search embeddings programmatically from Python.

Writing Plugins

Implement a custom embedding model by writing an LLM plugin that exposes an EmbeddingModel class.

Storage Format

Understand the SQLite schema, the binary blob encoding, and the encode/decode helpers used internally.

Build docs developers (and LLMs) love

Get started for freeTalk to us