Embeddings are what make semantic search possible in Autonomy Realms. Instead of searching by exact keywords, you search by meaning — and the system finds content that is conceptually similar to your query.
A vector embedding is a list of numbers that represents the meaning of a piece of text. When you send text to an embedding model (Autonomy Realms uses OpenAI's text-embedding-3-small), it returns a high-dimensional vector — a point in mathematical space where similar meanings cluster together.
Two signals about "walking through a forest at dawn" and "hiking in the woods at sunrise" would produce vectors that are close together, even though they share almost no exact words. A signal about "quarterly revenue projections" would produce a vector far away from both.
When you search in Autonomy Realms, your query is converted into a vector using the same embedding model. The system then compares your query vector against every stored embedding and returns the closest matches — ranked by similarity.
This means you can search for:
The search finds relevant results even when the exact words do not match.
What gets embedded is not just the raw signal text. An embedding prompt defines how the signal's data is formatted before being sent to the embedding model. This prompt assembles fields like the title, summary, transcript, tags, and themes into a single text block optimized for embedding.
Each entity type — signal, cluster, shape — has its own embedding prompt. Clusters and shapes also include sub-templates that define how their children (signals or clusters) are rendered within the embedding text.
You can customize embedding prompts in Settings > Templates to control exactly what text is embedded and how it is structured.
Embeddings are generated for all three entity types:
This means semantic search works at every level of your archive. You can find individual signals, thematic clusters, or broad patterns — all by meaning.
Embeddings can be regenerated at any time. If you update a signal's content, change the embedding prompt template, or add new signals to a cluster, regenerating the embedding ensures the vector reflects the current state of the content.