docs: named vectors by default — vector-less collections and the auto-schema default vector

docs/deploy/migration/index.md

@@ -49,6 +49,15 @@ Between `v1.25` and `v1.27`, there are two minor versions, `v1.26` and `v1.27`.
 
 :::
 
+### Vector configuration defaults (v1.39.1+, v1.40+) {#vector-config-defaults}
+
+Two defaults for **new** collections changed. Existing collections are untouched, and there is no migration step.
+
+- From `v1.39.1`, [auto-schema](/weaviate/config-refs/collections.mdx#auto-schema) creates a single named vector called `default` with the vectorizer set to `none`, instead of a [single vector collection](/weaviate/config-refs/collections.mdx#single-vector-collections). [`DEFAULT_VECTORIZER_MODULE`](/deploy/configuration/env-vars/index.md#DEFAULT_VECTORIZER_MODULE) is not applied to it, so supply the vectors yourself for auto-created collections or [add a vector](/weaviate/manage-collections/vector-config.mdx#add-new-named-vectors) after the collection is created.
+- From `v1.40`, a collection definition that sets no vector parameters at all creates a [collection without a vector](/weaviate/config-refs/collections.mdx#no-vector). Previously the server defaults filled in the top-level parameters and created a single vector collection. A definition that sets `vectorizer`, `vectorIndexType` or `vectorIndexConfig` is unaffected and still gets the defaults.
+
+The `v1.40` change is only visible with a client that omits `vectorIndexType` when you do not set one. A client that still sends `vectorIndexType: hnsw` on every collection create keeps getting a single vector collection. The Python client stopped sending it in `v4.21.2`.
+
 ### Raft Migration (v1.25.0+)
 
 Weaviate `v1.25.0` introduced Raft [as the consensus algorithm for cluster metadata](/weaviate/concepts/replication-architecture/cluster-architecture#metadata-replication-raft). This requires a one-time migration of the cluster metadata.

docs/weaviate/config-refs/collections.mdx

@@ -573,7 +573,7 @@ Weaviate supports two approaches for vector configuration:
 - **Single vector collections**: One vector space per object using top-level parameters (`vectorizer`, `vectorIndexType`, `vectorIndexConfig`)
 - **Multiple named vectors**: Multiple vector spaces per object using the `vectorConfig` parameter (**recommended**)
 
-You cannot combine both approaches in the same collection.
+You cannot combine both approaches in the same collection. A definition that uses neither creates a [collection without a vector](#no-vector).
 
 :::tip We recommend using `vectorConfig`
 
@@ -598,7 +598,7 @@ Using the `vectorConfig` parameter allows you to start with one vector per colle
 
 #### Single vector collections
 
-If you don't explicitly define a [named vector](#named-vectors) in your collection definition, Weaviate automatically creates what's known as a _single vector_ collection. These vectors are stored internally under the named vector `default` (which is a reserved vector name).
+A collection that sets any of the top-level `vectorizer`, `vectorIndexType` or `vectorIndexConfig` parameters has one vector space for the whole collection. Weaviate fills in the remaining top-level parameters from the server defaults.
 
 To learn which properties of your data are vectorized, refer to the [Configure semantic indexing](./indexing/vector-index.mdx#configure-semantic-indexing) section.
 
@@ -714,6 +714,35 @@ For more code example and configuration guides visit the [How-to: Vectorizer and
 
 :::
 
+#### The `default` vector name {#default-vector-name}
+
+`default` is a reserved vector name, and it means different things in the two approaches.
+
+**Single vector collections.** `default` is an alias for the collection's one vector space. Reads and writes that target `default` resolve to it, and a write that sends `vectors: {"default": [...]}` is stored there. The alias is not part of the collection definition, so `GET /v1/schema` returns the top-level parameters and no `vectorConfig` entry. You cannot add a named vector called `default` to such a collection.
+
+**Collections that use `vectorConfig`.** `default` is an ordinary named vector with no special behavior. It is the name Weaviate gives the vector that [auto-schema](#auto-schema) creates.
+
+#### Collections without a vector {#no-vector}
+
+:::info Changed in `v1.40`
+:::
+
+A collection definition that sets none of `vectorizer`, `vectorIndexType`, `vectorIndexConfig` or `vectorConfig` creates a collection with no vector. Before `v1.40`, the server defaults filled in the top-level parameters, so such a definition produced a single vector collection.
+
+For a collection without a vector:
+
+- `GET /v1/schema/{collection}` returns none of the vector parameters.
+- Objects without vectors are stored as usual. A write that carries a vector is rejected with a `422` error.
+- Vector search reports that the collection has no vector index.
+- You can [add a named vector](../manage-collections/vector-config.mdx#add-new-named-vectors) later.
+- You cannot add the top-level parameters later. An update that sets `vectorizer`, `vectorIndexType` or `vectorIndexConfig` is rejected with a `422` error:
+
+```text
+a class-level vectorizer or vector index cannot be added through an update, add a named vector instead
+```
+
+Existing collections are unchanged, and a definition that does set the top-level parameters still creates a single vector collection.
+
 ---
 
 ### Module configuration
@@ -1065,16 +1094,22 @@ After you create a collection, you can [add new properties](../manage-collection
 
 ## Auto-schema
 
+:::info Changed in `v1.39.1`
+:::
+
 The "Auto-schema" feature generates a collection definition automatically by inferring parameters from data being added. It is enabled by default, and can be disabled (e.g. in `docker-compose.yml`) by setting the environment variable [`AUTOSCHEMA_ENABLED`](/docs/deploy/configuration/env-vars/index.md#AUTOSCHEMA_ENABLED) to `'false'`.
 
 It will:
 
 - Create a collection if an object is added to a non-existent collection.
+- Create a single [named vector](#named-vectors) called `default` for that collection, with the vectorizer set to `none` and the index type taken from [`DEFAULT_VECTOR_INDEX`](/deploy/configuration/env-vars/index.md#DEFAULT_VECTOR_INDEX), or `hnsw` if that variable is unset. Before `v1.39.1`, it created a [single vector collection](#single-vector-collections) instead.
 - Add any missing property from an object being added.
 - Infer array data types, such as `int[]`, `text[]`, `number[]`, `boolean[]`, `date[]` and `object[]`.
 - Infer nested properties for `object` and `object[]` data types.
 - Throw an error if an object being added contains a property that conflicts with an existing schema type. (e.g. trying to import text into a field that exists in the schema as `int`).
 
+[`DEFAULT_VECTORIZER_MODULE`](/deploy/configuration/env-vars/index.md#DEFAULT_VECTORIZER_MODULE) is not applied to the `default` vector, so auto-schema does not vectorize your data. Supply the vectors yourself, or define the collection before you import.
+
 :::tip Define the collection manually for production use
 
 Generally speaking, we recommend that you disable auto-schema for production use.

docs/weaviate/manage-collections/collection-operations.mdx

@@ -29,6 +29,8 @@ import VectorConfigSyntax from "/_includes/vector-config-syntax.mdx";
 
 To create a collection, specify at least the collection name. If you don't specify any properties, [`auto-schema`](../config-refs/collections.mdx#auto-schema) creates them.
 
+From `v1.40`, a definition that sets only the name creates a [collection without a vector](../config-refs/collections.mdx#no-vector).
+
 import InitialCaps from "/_includes/schemas/initial-capitalization.md";
 
 <InitialCaps />

docs/weaviate/manage-collections/vector-config.mdx

@@ -129,6 +129,17 @@ To configure how a vectorizer works (i.e. what model to use) with a specific col
   </TabItem>
 </Tabs>
 
+## Create a collection without a vector
+
+:::info Changed in `v1.40`
+:::
+
+To create a collection without a vector, set no vector parameters at all — no `vectorizer`, `vectorIndexType` or `vectorIndexConfig`, and no named vectors. From `v1.40`, Weaviate no longer fills these in from the server defaults, so vector search is not available on the collection. For the rest of the behavior, see [Collections without a vector](/weaviate/config-refs/collections.mdx#no-vector).
+
+You can [add a named vector](#add-new-named-vectors) to it later. The top-level `vectorizer`, `vectorIndexType` and `vectorIndexConfig` parameters cannot be added after creation.
+
+The change is only visible with a client that omits `vectorIndexType` when you do not set one; the Python client does so from `v4.21.2`. See [Vector configuration defaults](/deploy/migration/index.md#vector-config-defaults).
+
 ## Define named vectors
 
 You can define multiple [named vectors](../concepts/data.md#multiple-vector-embeddings-named-vectors) per collection. This allows each object to be represented by multiple vector embeddings, each with its own vector index.
@@ -182,7 +193,7 @@ As such, each named vector configuration can include its own vectorizer and vect
 
 <AddNamedVectors/>
 
-Named vectors can be added to existing collection definitions with named vectors. (This is not possible for collections without named vectors.)
+Named vectors can be added to a collection that already has named vectors, and to a [collection created without a vector](#create-a-collection-without-a-vector). You cannot add one to a [single vector collection](/weaviate/config-refs/collections.mdx#single-vector-collections).
 
 <Tabs className="code" groupId="languages">
   <TabItem value="py" label="Python">