What is Vector Embeddings?
Vector Embeddings — Vector embeddings are numerical representations of text (or other data) in a high-dimensional space where semantically similar content is positioned close together. They enable AI systems to perform semantic search — finding content by meaning rather than exact keyword matches — by converting text into fixed-length arrays of numbers that capture semantic relationships.
How Embeddings Work
An embedding model converts text into a vector (array of numbers). For example, a 512-dimensional embedding converts "how to write unit tests" into an array of 512 floating-point numbers. Similar texts produce similar vectors, enabling mathematical comparison.
Common Embedding Models (2026)
| Model | Dimensions | Best For |
|---|---|---|
| Jina v3 | 512-1024d | Code + text (multilingual) |
| Voyage 4 Lite | 1024d | Code-focused |
| OpenAI text-embedding-3-small | 1536d | General purpose |
| Cohere embed-v4 | 1024d | Multilingual |
Embeddings in Quoth
Quoth uses two embedding systems:
- Cloud (SaaS): Jina v3 at 512 dimensions for semantic search across the shared knowledge base
- Local (Plugin): Voyage 4 Lite at 1024 dimensions for pattern matching in the self-learning pipeline
The local HNSW index (M=16, cosine distance) enables O(log n) approximate nearest-neighbor search, making pattern lookups fast even with thousands of entries.