Home / Embeddings
Embeddings guides
14 published guides. Start with What are embeddings.
What embeddings are, how they encode meaning, and every decision that affects search quality.
Read the guideLearning path
All guides
Understand it2 guides
Compare and choose3 guides
ComparisonChoosing an embedding modelCompare embedding models on retrieval quality, dimensions, cost, context length and languages, with a decision path.ComparisonTokenization vs embeddingTokens split text, embeddings represent meaning. How the two steps connect and why it matters for search.ComparisonWord vs sentence embeddingsWhy word2vec-style vectors fell short for search and how sentence embeddings fixed it.
How it works6 guides
ExplainerDomain-specific embeddingsWhen a general embedding model fails your domain and the options: pick a specialist, adapt, or fine-tune.ExplainerEmbedding dimensionsWhat the number of dimensions means, how it affects quality, memory and speed, and when fewer is better.ExplainerAsymmetric vs symmetric semantic searchShort queries against long passages need different handling than matching like with like. How to tell and what to change.ExplainerEmbedding and vector quantizationShrink vectors 4x to 32x with quantization, and measure how much recall each method costs.ExplainerMatryoshka embeddingsHow Matryoshka training lets one embedding be cut to fewer dimensions, and when to use it in search.ExplainerMultimodal embeddingsHow models like CLIP put text, images, audio and video in one space so a text query can find any of them.
Build and run it3 guides
TutorialFine-tuning embedding modelsHow to fine-tune an embedding model on your own queries, how much data it needs, and how to prove it helped.TutorialRe-embedding when the model changesA zero-downtime migration plan for switching embedding models: dual indexes, backfill, cutover and rollback.TutorialBatch embeddingHow to embed large corpora efficiently with batch APIs, and how batch size affects throughput, cost and failures.