Semantic Search: Find Content by Meaning

Cosmic now understands what your content is about. Semantic search (Content RAG) embeds your objects into a vector index so a natural-language query returns the most relevant content, even when the wording is different from what's stored. It's a new API endpoint, and it also powers your Agents.

Your content is embedded on write and kept in sync automatically as you create, edit, publish, unpublish, and delete objects, and embedding is free: only queries consume AI tokens. Semantic search is in Beta, on by default for new projects and workspaces, and an admin or manager can turn it on for existing ones under Settings → Semantic search.

What's New

• Semantic search endpoint. embeds your query and returns the most semantically relevant objects in the bucket, ranked by score, with optional , , , , and filters.
• Agents that retrieve by meaning. With CMS read access, Agents get a new tool so they can find the right content by meaning instead of scanning titles, making support and Q&A answers more accurate.
• Always in sync. Your content is embedded on write and kept current as you create, edit, publish, unpublish, and delete objects. Embedding your content is free.

Why This Matters

Keyword and structured queries are great when you know the exact field or phrase. But real questions ("how do refunds work on annual plans?") rarely match the exact words in your content. Semantic search bridges that gap: it matches intent, so your apps, support bots, and Agents surface the right objects the first time.

How It Works

1. When semantic search is enabled for your account, Cosmic embeds your objects' searchable text into a vector index and keeps it in sync automatically.
2. Call with a natural-language (bucket write key required) to get ranked results with a text snippet for each match.
3. Give an Agent CMS read access and it can use to ground its answers in your content.

Notes:

• Embedding your content on write is free. Query-time search consumes AI tokens like other AI features.
• Semantic search complements structured queries: use queries for exact filters, semantic search for meaning. See the Semantic search docs.
• In Beta. On by default for new projects and workspaces; existing ones can turn it on under Settings → Semantic search.