Embeddings

An embedding is a list of floating-point numbers (a vector) that represents the semantic meaning of a piece of text. By comparing these vectors using distance metrics (like Cosine Similarity), Jabrod can determine which chunks of text are most relevant to a user’s query, even if they don’t share the exact same keywords.

Supported Models

Jabrod integrates with several embedding providers to give you flexibility between cost, speed, and quality.

Model	Dimensions	Provider	Tier	Notes
`text-embedding-3-small`	1536	OpenAI (via OpenRouter)	Free / Pro	Fast, cheap, industry standard.
`nomic-embed-text`	768	Nomic (via OpenRouter)	Free / Pro	Open source alternative.
`cohere/embed-english-v3.0`	1024	Cohere (via OpenRouter)	Pro	Excellent for English RAG.
`voyage/voyage-large-2`	1536	Voyage (via OpenRouter)	Pro	State-of-the-art retrieval performance.

The Embedding Process

When a Data Source is chunked, Jabrod sends the chunks in batches to the selected embedding model.
The model returns a vector for each chunk.
Jabrod stores both the original text chunk and the vector in the Vector Database (Upstash).
When you submit a Query, Jabrod embeds the query text using the same model, and asks the Vector Database to find the closest matching vectors.