Skip to content

Caedral technology

Available

Caedral Embed

Turn text into vectors so search, retrieval, and RAG can compare meaning instead of keywords. Caedral Embed is Caedral's hosted embedding model: same API key as chat, billed from the Caedral Models pool.

caedral-embed · caedral-embed-e1-small-v1 · 384 dimensions · $0.001 / 1M tokens

Get API keyCaedral Models docs

What

An embedding is a numeric representation of a piece of text. Similar sentences sit close together in that space. Caedral Embed returns 384-dimensional vectors over POST /v1/embeddings. Responses echo the canonical model id caedral-embed-e1-small-v1.

Why

Keyword search misses paraphrases. Embeddings let you store documents once, query them later, and retrieve the passages that actually answer the question — the retrieval step in RAG, semantic search, clustering, and duplicate detection.

How it fits in Caedral

Embed is a Caedral-operated product, not a third-party lab alias. Usage draws the Caedral Models pool. It does not run through Notre — Notre applies to chat completions only. Pair retrieved chunks with Caedral Rerank when you need a second pass over candidates.

Model IDs
Request caedral-embed or caedral-embed-e1-small-v1. Responses use the canonical id.
Context
512 tokens per input. Longer text should be chunked before embed.
Vectors
384 dimensions, L2-normalized.
Promo
Limited free use through 28 September 2026 at 130 RPM, then billed at the published rate.

Use

Send one string or an array. Store the vectors in your own index. At query time, embed the question, retrieve nearest neighbors, then optionally rerank. OpenAI-compatible clients work if you set the Caedral base URL.

  • Semantic search over docs, tickets, or knowledge bases
  • RAG: retrieve, then generate with a chat model
  • Clustering and near-duplicate detection

POST /v1/embeddings

curl https://api.caedral.com/v1/embeddings \
  -H "Authorization: Bearer $CAEDRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "caedral-embed",
            "input": "Quarterly pipeline report for retrieval"
  }'

Next