Skip to main content

REST - /v1/schema

Overview

The schema endpoint is for creating, reading, updating and deleting collections (also called collections) and properties. The overall configuration is referred to as the schema.

Get the schema

The response will contain an array of objects.

Method and URL

GET /v1/schema

Example request

import weaviate

client = weaviate.connect_to_local()

try:
schema = client.collections.list_all(simple=False) # Use `simple=False` to get comprehensive information

print(schema)

finally:
client.close()
Example response
{
"classes": [
{
"class": "Category",
"description": "Category an article is a type off",
"moduleConfig": {
"text2vec-contextionary": {
"vectorizeClassName": false
}
},
"properties": [
{
"dataType": [
"text"
],
"description": "category name",
"indexFilterable": true,
"indexSearchable": true,
"moduleConfig": {
"text2vec-contextionary": {
"vectorizePropertyName": false
}
},
"name": "name"
}
],
"vectorIndexType": "hnsw",
"vectorizer": "none"
},
{
"class": "Publication",
"description": "A publication with an online source",
"moduleConfig": {
"text2vec-contextionary": {
"vectorizeClassName": false
}
},
"properties": [
{
"dataType": [
"text"
],
"description": "Name of the publication",
"name": "name"
},
{
"dataType": [
"geoCoordinates"
],
"description": "Geo location of the HQ",
"name": "headquartersGeoLocation"
},
{
"dataType": [
"Article"
],
"description": "The articles this publication has",
"name": "hasArticles"
},
{
"dataType": [
"Article"
],
"description": "Articles this author wrote",
"name": "wroteArticles"
}
],
"vectorIndexType": "hnsw",
"vectorizer": "none"
},
{
"class": "Author",
"description": "Normalised types",
"moduleConfig": {
"text2vec-contextionary": {
"vectorizeClassName": true
}
},
"properties": [
{
"dataType": [
"text"
],
"description": "Name of the author",
"name": "name"
},
{
"dataType": [
"Publication"
],
"description": "The publication this author writes for",
"name": "writesFor"
}
],
"vectorIndexType": "hnsw",
"vectorizer": "none"
},
{
"class": "Article",
"description": "Normalised types",
"moduleConfig": {
"text2vec-contextionary": {
"vectorizeClassName": false
}
},
"properties": [
{
"dataType": [
"text"
],
"description": "title of the article",
"indexFilterable": true,
"indexSearchable": true,
"moduleConfig": {
"text2vec-contextionary": {
"vectorizePropertyName": false
}
},
"name": "title"
},
{
"dataType": [
"text"
],
"description": "url of the article",
"indexFilterable": true,
"indexSearchable": false,
"moduleConfig": {
"text2vec-contextionary": {
"vectorizePropertyName": false
}
},
"name": "url"
},
{
"dataType": [
"text"
],
"description": "summary of the article",
"indexFilterable": true,
"indexSearchable": true,
"moduleConfig": {
"text2vec-contextionary": {
"vectorizePropertyName": false
}
},
"name": "summary"
},
{
"dataType": [
"date"
],
"description": "date of publication of the article",
"name": "publicationDate"
},
{
"dataType": [
"int"
],
"description": "Words in this article",
"name": "wordCount"
},
{
"dataType": [
"Author",
"Publication"
],
"description": "authors this article has",
"name": "hasAuthors"
},
{
"dataType": [
"Publication"
],
"description": "publication this article is in",
"name": "inPublication"
},
{
"dataType": [
"Category"
],
"description": "category this article is of",
"name": "ofCategory"
},
{
"dataType": [
"boolean"
],
"description": "whether the article is currently accessible through the url",
"name": "isAccessible"
}
],
"vectorIndexType": "hnsw",
"vectorizer": "none"
}
]
}

Create a collection

Create a new data object collection in the schema.

note

From v1.5.0 onwards, manual creation of a schema is optional. Learn more about Auto Schema.

Method and URL

POST /v1/schema

Parameters

For more details on these parameters, see the schema configuration reference page.

NameLocationTypeDescription
classbodystringThe name of the collection (a.k.a. class). Multiple words should be concatenated in CamelCase, e.g. ArticleAuthor.
descriptionbodystringDescription of the collection.
vectorizerbodystringVectorizer to use for data objects added to this collection. Default can be set via Weaviate environment variables.
vectorIndexTypebodystringVector index type. Defaults to hnsw.
vectorIndexConfigbodyobjectVector index type specific settings. See the vector index configuration page for more details.
moduleConfig > <module_name> > vectorizeClassNamebodybooleanInclude the collection name in vector calculation (default true). Learn more about semantic indexing in Weaviate.
propertiesbodyarrayAn array of property objects.
properties > namebodystringThe name of the property. Multiple words should be concatenated in camelCase, e.g. nameOfAuthor.
properties > dataTypebodyarraySee the available data types.
properties > descriptionbodystringDescription of the property.
properties > moduleConfig > <module_name> > skipbodybooleanIf true, the whole property will NOT be included in vectorization. Default is false, meaning that the object will be NOT be skipped.
properties > moduleConfig > <module_name> > vectorizePropertyNamebodybooleanWhether the name of the property is used in the calculation for the vector position of data objects. Default is true. Learn more about semantic indexing in Weaviate.
properties > indexFilterable (available from v1.19)bodybooleanShould the data stored in this property be indexed with the filterable, Roaring Bitmap index? Read more about indexing in Weaviate.
properties > indexSearchable (available from v1.19)bodybooleanShould the data stored in this property be indexed to allow BM25/hybrid-search index? Read more on how to configure indexing in Weaviate.
properties > tokenizationbodystringOnly for string/text props. Introduced in v1.12.0. Control how a field is tokenized in the inverted index. Defaults to "word", can be set to "field". Learn more about property tokenization.
invertedIndexConfig > bm25 > bbodyfloatb parameter for BM25 searches. Default: 0.75
invertedIndexConfig > bm25 > k1bodyfloatk1 parameter for BM25 searches. Default: 1.2
invertedIndexConfig > stopwordsbodyobjectConfigure stopword behavior. See more details here.
invertedIndexConfig > indexTimestampsbodybooleanMaintain an inverted index for each object by its internal timestamps, currently including creationTimeUnix and lastUpdateTimeUnix.
See more details here.
invertedIndexConfig > indexNullStatebodybooleanIndex null values to allow filtering.
invertedIndexConfig > indexPropertyLengthbodybooleanIndex property lengths to allow filtering.
shardingConfigbodyobjectSharding specific settings. See the schema configuration reference page for more details.
replicationConfig > factorbodyintThe replication factor, aka the number of copies in a replicated Weaviate setup.
multiTenancyConfig > enabledbodyBooleanWhether to enable multi-tenancy for this collection. (Defaults to false.)

Example request for creating a collection

You can specify as few parameters as the collection name.

import weaviate
import weaviate.classes as wvc

client = weaviate.connect_to_local()

try:
client.collections.create(
name="Article",
description="A written text, for example a news article or blog post",
properties=[
wvc.config.Property(
data_type=wvc.config.DataType.TEXT,
description="Title of the article",
name="title",
),
wvc.config.Property(
data_type=wvc.config.DataType.TEXT,
description="The content of the article",
name="content",
),
]
)

finally:
client.close()

Or you can manually specify any number of the possible parameters:

import weaviate
import weaviate.classes as wvc

client = weaviate.connect_to_local()

try:
client.collections.create(
name="Article",
description="A written text, for example a news article or blog post",
vector_index_config=wvc.config.Configure.VectorIndex.hnsw(
distance_metric=wvc.config.VectorDistances.COSINE,
ef=128,
max_connections=64
),
vectorizer_config=wvc.config.Configure.Vectorizer.text2vec_openai(
vectorize_collection_name=True,
),
properties=[
wvc.config.Property(
data_type=wvc.config.DataType.TEXT,
description="Title of the article",
name="title",
index_filterable=True,
index_searchable=True,
skip_vectorization=False,
vectorize_property_name=False,
),
wvc.config.Property(
data_type=wvc.config.DataType.TEXT,
description="The content of the article",
name="content",
index_filterable=True,
index_searchable=True,
skip_vectorization=False,
vectorize_property_name=False,
),
],
sharding_config=wvc.config.Configure.sharding(
virtual_per_physical=128,
desired_count=1,
desired_virtual_count=128,
)
)

finally:
client.close()

Get a single collection from the schema

Retrieves the configuration of a single collection in the schema.

Method and URL

GET /v1/schema/{collection_name}

Example request

import weaviate

client = weaviate.connect_to_local()

try:
articles = client.collections.get("Article")
articles_config = articles.config.get(simple=False)

print(articles_config)

finally:
client.close()

Delete a collection

Remove a collection (and all its data objects).

Method and URL

DELETE v1/schema/{collection_name}

URL parameters

NameLocationTypeDescription
{collection_name}pathstringThe name of the collection

Example request for deleting a collection

import weaviate

client = weaviate.connect_to_local()

try:
client.collections.delete("Article")

finally:
client.close()

Update a collection

Update settings of an existing collection. Use this endpoint to alter an existing collection in the schema.

Not all settings are mutable
  • Please note that not all settings are mutable.
  • To update any other (i.e. immutable) setting, you need to delete the collection, re-create it with the correct setting and then re-import the data.
The list of mutable settings
  • description
  • invertedIndexConfig
    • bm25
      • b
      • k1
    • cleanupIntervalSeconds
    • stopwords
      • additions
      • preset
      • removals
  • replicationConfig
    • factor
  • vectorIndexConfig
    • dynamicEfFactor
    • dynamicEfMin
    • dynamicEfMax
    • flatSearchCutoff
    • skip
    • vectorCacheMaxObjects
    • pq
      • bitCompression
      • centroids
      • enabled
      • segments
      • trainingLimit
      • encoder
        • type
        • distribution

This endpoint cannot be used to introduce additional properties. For this, use POST /v1/schema/{collection_name}/properties. A typical use case for this endpoint is to update configuration, such as vectorIndexConfig/dynamicEfFactor or vectorIndexConfig/pq/bitCompression. Note that even in mutable sections, such as vectorIndexConfig, some fields may be immutable.

You should attach a body to this PUT request with the entire new configuration of the collection.

Method and URL

PUT v1/schema/{collection_name}

Parameters

The URL must contain the following parameter:

NameLocationTypeDescription
{collection_name}pathstringThe name of the collection

Parameters in the PUT body:

NameLocationTypeDescription
classbodystringThe name of the collection (a.k.a. class). Multiple words should be concatenated in CamelCase, e.g. ArticleAuthor.
descriptionbodystringDescription of the collection.
vectorIndexConfigbodyobjectVector index type specific settings. See the vector index configuration page for more details.
invertedIndexConfig > bm25 > bbodyfloatb parameter for BM25 searches. Default: 0.75
invertedIndexConfig > bm25 > k1bodyfloatk1 parameter for BM25 searches. Default: 1.2
invertedIndexConfig > stopwordsbodyobjectConfigure stopword behavior. See more details here.
replicationConfig > factorbodyintThe replication factor, aka the number of copies in a replicated Weaviate setup.

Example request for updating a collection

import weaviate

client = weaviate.connect_to_local()

try:
# Retrieve the collection
articles = client.collections.get("Article")

# Update the collection
articles.config.update(
description="An updated collection description.",
inverted_index_config=wvc.config.Reconfigure.inverted_index(
bm25_k1=1.1,
bm25_b=0.8
),
vector_index_config=wvc.config.Reconfigure.VectorIndex.hnsw(
dynamic_ef_factor=10
)
)
# Coming soon :)

finally:
client.close()

Add a property

Limitations

Adding a property after importing objects can lead to limitations in inverted-index related behavior.


This is caused by the inverted index being built at import time. If you add a property after importing objects, the inverted index will not be updated. This means that the new property will not be indexed for existing objects. This can lead to unexpected behavior when querying.


To avoid this, you can either:

  • Add the property before importing objects.
  • Delete the collection, re-create it with the new property and then re-import the data.

We are working on a re-indexing API to allow you to re-index the data after adding a property. This will be available in a future release.

Method and URL

POST v1/schema/{collection_name}/properties

Parameters

NameLocationTypeDescription
dataTypebodyarrayAn available data type.
descriptionbodystringDescription of the property.
moduleConfig > <module_name> > skipbodybooleanIf true, the whole property will NOT be included in vectorization. Default is false, meaning that the object will be NOT be skipped.
moduleConfig > <module_name> > vectorizePropertyNamebodybooleanWhether the name of the property is used in the calculation for the vector position of data objects. Default is true. Learn more about how to configure indexing in Weaviate.
namebodystringThe name of the property. Multiple words should be concatenated in camelCase like nameOfAuthor.
indexFilterable (available from v1.19)bodybooleanShould the data stored in this property be indexed with the filterable, Roaring Bitmap index? Read more about indexing in Weaviate.
indexSearchable (available from v1.19)bodybooleanShould the data stored in this property be indexed to allow BM25/hybrid-search index? Read more about indexing in Weaviate.
indexInverted (deprecated)bodybooleanShould the data stored in this property be indexed? Learn more about indexing in Weaviate.

Example request for adding a property

import weaviate

client = weaviate.connect_to_local()

try:
articles = client.collections.get("Article")

articles.config.add_property(
wvc.config.Property(
name="onHomepage",
data_type=wvc.config.DataType.BOOL
)
)

finally:
client.close()

Inspect the shards of a collection

As described in Architecture > Storage, creation of a collection leads to creating an index which manages all the disk storage and vector indexing. An index itself can be comprised of multiple shards. If a collection index is used on multiple nodes of a multi-node Weaviate cluster there must be at least one shard per node.

You can view a list of all shards for a particular collection:

Method and URL

note

This API was added in v1.12.0.

GET v1/schema/{collection_name}/shards

Parameters

NameLocationTypeDescription
{collection_name}URL pathstringThe name of the collection

Example request viewing shards of a collection

import weaviate

client = weaviate.connect_to_local()

try:
articles = client.collections.get("Article")

article_shards = articles.config.get_shards()
print(article_shards)

finally:
client.close()

Update shard status

A shard may have been marked as read-only, for example because the disk was full. You can manually set a shard to READY again using the following API. There is also a convenience function in each client to set the status of all shards of a collection.

Method and URL

note

This API was added in v1.12.0

PUT v1/schema/{collection_name}/shards/{shard_name}

Parameters

NameLocationTypeDescription
{collection_name}URL pathstringThe name of the collection.
{shard_name}URL pathstringThe name/id of the shard.
statusbodystringThe status to update the shard to. One of READONLY, READY.

Example requests to update the status of a shard

import weaviate

client = weaviate.connect_to_local()

try:
# Coming soon :)

finally:
client.close()

Multi-tenancy

Multi-tenancy availability
  • Multi-tenancy available from version v1.20
  • (Experimental) Tenant activity status setting available from version v1.21

Tenants are used to separate data between different users or groups of users. They can be specified as follows:

Enable multi-tenancy

    import weaviate.classes as wvc

multi_collection = client.collections.create(
name="MultiTenancyCollection",
# Enable multi-tenancy on the new collection
multi_tenancy_config=wvc.config.Configure.multi_tenancy(True)
)

Add tenant(s)

Pass a payload with an array of tenant objects. The available fields are:

NameTypeDescription
namestring(Required) The name of the tenant.
activityStatusstring(Optional, experimental) The activity status of the tenant. Can be HOT (default) or COLD.

Example payload

[
{
"name": "TENANT_A"
},
{
"name": "TENANT_B",
"activityStatus": "COLD"
}
]
Allowable tenant names

A tenant name can only contain alphanumeric characters (a-z, A-Z, 0-9), underscore (_), and hyphen (-), with a length of 4 to 64 characters.

POST v1/schema/{collection_name}/tenants
    import weaviate.classes as wvc

# Add two tenants to the collection
multi_collection.tenants.create(
tenants=[
wvc.tenants.Tenant(name="tenantA"),
wvc.tenants.Tenant(name="tenantB"),
]
)

List tenants

GET v1/schema/{collection_name}/tenants
    multi_collection = client.collections.get("MultiTenancyCollection")
tenants = multi_collection.tenants.get()

print(tenants)

Remove tenants

Pass a payload with an array of tenant names in the form of ["TENANT_NAME1", "TENANT_NAME2"] to remove from the collection.

DELETE v1/schema/{collection_name}/tenants
    multi_collection = client.collections.get("MultiTenancyCollection")

# Remove a list of tenants - tenantX will be ignored.
multi_collection.tenants.remove(["tenantB", "tenantX"])

Update tenants

PUT v1/schema/{collection_name}/tenants

Pass a payload with an array of tenant objects. For updating tenants, both name and activityStatus are required.

Example payload

[
{
"name": "TENANT_A",
"activityStatus": "COLD"
},
{
"name": "TENANT_B",
"activityStatus": "HOT"
}
]
    multi_collection = client.collections.get("MultiTenancyCollection")
multi_collection.tenants.update(tenants=[
wvc.tenants.Tenant(
name="tenantA",
activity_status=weaviate.schema.TenantActivityStatus.COLD
)
])