Embeddings

Overview

An embedding turns a piece of text into a fixed-size numeric vector, so that Zilla can compare two pieces of text by their meaning instead of by their exact wording. Embeddings are registered under a namespace-scoped embeddings map, the same way catalogs, vaults, and stores are registered, and are referenced by name from wherever a semantic comparison is needed.

embeddings:
  moderator0:
    type: glove

Glove

Providers

Type	Purpose
glove	Averages pre-trained GloVe word vectors. Runs in-process, no external service. [Incubator]
openai	Resolves text to a vector via OpenAI's embeddings API, or a self-hosted OpenAI-compatible server. [Plus]
aws-bedrock	Resolves text to a vector via a managed AWS Bedrock embedding model (Titan or Cohere). [Plus]
ibm-watsonx-ai	Resolves text to a vector via an IBM watsonx.ai-hosted embedding model (SaaS or CP4D). [Plus]

For a hosted alternative backed by IBM watsonx.ai's text-embeddings models, see ibm-watsonx-ai.

Comparing embeddings

Two embeddings are compared using cosine similarity: a value from -1 to 1 describing how closely their meanings align, independent of vector magnitude or shared vocabulary. A higher similarity means the two pieces of text mean something closer to the same thing.

Semantic message moderation

The vector model references a named embedding to reject a message value whose meaning is close enough to one of a configured set of example phrases, even when the message shares no vocabulary at all with those phrases.

embeddings:
  moderator0:
    type: glove
bindings:
  north_echo_server:
    type: echo
    kind: server
    options:
      value:
        model: vector
        embedding: moderator0
        reject:
          - "You will never believe what happened next."
        threshold: 0.94