vectors

vectors

Store child: vector/embedding index config

Translations

• en: vectors
• de: Vektoren
• es: vectores
• fr: vecteurs
• ja: ベクトル
• ko: 벡터
• zh: 向量

Example — Remembering and storing data

Keeps notes in memory and owns a small contact database with rules about who can read it.

machine crm_notes
  accepts
    note as text, is required
  responds with
    matches as list
  behaves
    remember keep
      collection: "notes"
      content: input.note
    recall find
      query: input.note
      limit: 5
    save snapshot
      to: contacts
    query lookup
      from: contacts
    compute done
      {matches: steps.find.results}
  stores
    store contacts
      source: managed
      resource contact
        id as uuid, primary_key
        name as text, is required
        email as text
        timestamps
        create add
          accept: [name, email]
        read by_name
          argument name: string, required
          filter: name == arg(name)
        update rename
          accept: [name]
        destroy remove
        policy action_type(read)
          authorize_if actor_attribute(access_level, "team")
        vectors
          source: qdrant
          embedding_model: "nomic-embed-text"
  verifies
    test "keeps the note and finds related ones"
      assuming keep {ok: true}
      assuming find {results: ["Call Dana back"]}
      assuming snapshot {ok: true}
      assuming lookup {rows: []}
      given {note: "Call Dana back"}
      expect {matches: ["Call Dana back"]}