Create a collection (HTTP)

For details about additional languages, see:

Creates a new collection in a database.

Ready to write code? See the examples for this method to get started.

Result

Creates a collection with the specified parameters.

If the command succeeds, the response indicates the success.

Example successful response:

{
  "status": {
    "ok": 1
  }
}

You cannot edit a collection’s definition after you create the collection.

Signature

Use the createCollection command.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
--header "Token: APPLICATION_TOKEN" \
--header "Content-Type: application/json" \
--data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": OPTIONS
  }
}'

Parameters

Name Type Summary

name

string

The name of the new collection.

Collection names must follow these rules:

  • Can contain letters, numbers, and underscores

  • Cannot exceed 48 characters

  • Must be unique within the keyspace

options

object

Optional. The options for this operation. See Properties of options for more details.
Properties of options
Name Type Summary

defaultId

object

Optional. Specifies the default ID type for documents in the collection. This is used when you insert a document without an _id field.

Can be one of:

  • {"type": "objectId"}: Each autogenerated _id value is an objectId as provided by the bson library.

  • {"type": "uuidv7"}: Each autogenerated _id value is a version 7 UUID. This is designed as a replacement for version 1 time UUID, and it is recommended for use in new systems.

  • {"type": "uuidv6"}: Each autogenerated _id value is a version 6 UUID. This is field-compatible with version 1 time UUIDs, and it supports lexicographical sorting.

  • {"type": "uuid"}: Each autogenerated _id value is a version 4 UUID. This type is analogous to the uuid type and functions in Apache Cassandra®.

For examples, see Create a collection and specify the default ID format.

For more information, see Document IDs (HTTP).

Default: Each autogenerated _id value is a string form of a version 4 UUID

vector

object

Optional. The vector configuration for the collection. This includes things like the vector dimension, similarity metric, and source model. This also includes settings for server-side embedding generation if you want your collection to have vectorize enabled.

Required for vector search and hybrid search.

For examples, see Create a collection that can automatically generate vector embeddings and Create a collection that can store vector embeddings.

The vector object contains the following properties:

  • dimension (int): The dimension for vector embeddings in the collection. This should match the dimension of the vector that your embedding model produces. Optional if you specify a vector.service.modelName value that has a default dimension value.

  • metric (string): The similarity metric to use for vector search. Can be one of: cosine (default), dot_product, euclidean.

  • sourceModel (string): Optional. The model used to generate the vector embeddings. This enables certain vector optimizations on the index. Can be one of: ada002, bert, cohere-v3, gecko, nv-qa-4, openai-v3-large, openai-v3-small, other.

  • service (object): Optional. The configuration for a vectorize embedding provider integration. This lets your collection use vectorize to automatically generate embeddings. See Create a collection that can automatically generate vector embeddings, use findEmbeddingProviders, or see the documentation for your embedding provider integration to determine what values to specify.

    The service object contains the following properties:

    • provider (string): The name of the vectorize embedding provider.

    • modelName (string): A valid model name for the specified vectorize embedding provider.

    • authentication (string): Optional depending on your provider. Use credentials stored in Astra DB KMS to authenticate with your vectorize embedding provider. In options.vector.service.authentication.providerKey, provide the credential’s API Key name as given in Astra DB KMS. Alternatively, you can omit the authentication object, and then provide the authentication key in an x-embedding-api-key header instead. If you use header authentication, you must provide the x-embedding-api-key header with every command that requires vectorize for this collection, including inserting data and vector search with vectorize. If you use an AWS embedding provider, you must use the x-embedding-access-id and x-embedding-secret-id headers to pass your access ID and secret ID instead of the x-embedding-api-key header.

    • parameters (object): Optional depending on your provider. Additional parameters required for your embedding provider

lexical

object

Optional. The lexical search configuration for the collection.

The lexical object has the following properties:

For examples, see Create a collection that supports lexicographical matching.

Default: An object with an enabled value of true and an analyzer value of "standard", which corresponds to the standard Apache Lucene™ analyzer.

rerank

object

Optional. The reranker configuration for the collection.

The rerank object has the following properties:

  • enabled (boolean): Whether to enable reranking for the collection. Use this to disable reranking and hybrid search for the collection. Required to support hybrid search.

  • service (object): Describes the provider and model name for a reranker model. The service object contains the following properties:

    • provider (string): The name of the reranking provider. Only Nvidia is supported.

    • modelName (string): The name of a reranking model supported by the reranking provider. Only nvidia/llama-3.2-nv-rerankqa-1b-v2 is supported.

For examples, see Create a collection that supports hybrid search.

Default: An object with an enabled value of true and a service value corresponding to the NVIDIA llama-3.2-nv-rerankqa-1b-v2 reranking model. This means that reranking is enabled by default.

indexing

object

Optional. Configures selective indexing for data inserted to the collection.

The indexing object must contain one of:

  • allow (array): The properties to index. Must contain at least one property. "allow": ["*"] indexes all properties, which is the same as the default behavior.

  • deny (array): The properties to not index. Must contain at least one property. "deny": ["*"] means that no properties are indexed.

You must use & to escape any . or & in field names in the indexing clause. You cannot use & to escape any other characters. Dot notation, which is used to reference nested fields, should not be escaped. For more information, see Work with . and & in field names (HTTP).

For examples, see Create a collection and specify which fields to index and Create a collection and specify which fields shouldn’t be indexed.

Default: All fields of all documents.

Examples

The following examples demonstrate how to create a collection.

Create a collection that is not vector-enabled

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {}
  }
}'

Create a collection that can store vector embeddings

Collections that are vector-enabled can store vector embeddings in the reserved $vector field and work with vector search.

For optimal vector search results, you should specify the dimension, metric, and source model of your vector embeddings. All vector embeddings in a collection should be generated by the same model with the same dimensions. The source model can be one of: ada002, bert, cohere-v3, gecko, nv-qa-4, openai-v3-large, openai-v3-small, other.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": 1024,
        "metric": "cosine",
        "source_model": "nv-qa-4"
      }
    }
  }
}'

Create a collection that can automatically generate vector embeddings

If you want to automatically generate vector embeddings, create a vector-enabled collection and configure an embedding provider integration for the collection.

The configuration depends on the embedding provider.

Configure Azure OpenAI as the embedding provider

For more detailed instructions, see Integrate Azure OpenAI as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "azureOpenAI",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          },
          "parameters": {
            "resourceName": "RESOURCE_NAME",
            "deploymentId": "DEPLOYMENT_ID"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Azure OpenAI API key that you want to use. Must be the name of an existing Azure OpenAI API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002.

    For Azure OpenAI, you must select the model that matches the one deployed to your DEPLOYMENT_ID in Azure.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

  • RESOURCE_NAME: The name of your Azure OpenAI Service resource, as defined in the resource’s Instance details. For more information, see the Azure OpenAI documentation.

  • DEPLOYMENT_ID: Your Azure OpenAI resource’s Deployment name. For more information, see the Azure OpenAI documentation.

Configure Hugging Face (Dedicated) as the embedding provider

For more detailed instructions, see Integrate Hugging Face Dedicated as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "metric": "SIMILARITY_METRIC",
        "dimension": MODEL_DIMENSIONS,
        "service": {
          "provider": "huggingfaceDedicated",
          "modelName": "endpoint-defined-model",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          },
          "parameters": {
            "endpointName": "ENDPOINT_NAME",
            "regionName": "REGION",
            "cloudName": "CLOUD_PROVIDER"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Hugging Face Dedicated user access token that you want to use. Must be the name of an existing Hugging Face Dedicated user access token in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: endpoint-defined-model.

    For Hugging Face Dedicated, you must deploy the model as a text embeddings inference (TEI) container.

    You must set MODEL_NAME to endpoint-defined-model because this integration uses the model specified in your dedicated endpoint configuration.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

  • ENDPOINT_NAME: The programmatically-generated name of your Hugging Face Dedicated endpoint. This is the first part of the endpoint URL. For example, if your endpoint URL is https://mtp1x7muf6qyn3yh.us-east-2.aws.endpoints.huggingface.cloud, the endpoint name is mtp1x7muf6qyn3yh.

  • REGION: The cloud provider region your Hugging Face Dedicated endpoint is deployed to. For example, us-east-2.

  • CLOUD_PROVIDER: The cloud provider your Hugging Face Dedicated endpoint is deployed to. For example, aws.

Configure Hugging Face (Serverless) as the embedding provider

For more detailed instructions, see Integrate Hugging Face Serverless as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "huggingface",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Hugging Face Serverless user access token that you want to use. Must be the name of an existing Hugging Face Serverless user access token in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: sentence-transformers/all-MiniLM-L6-v2, intfloat/multilingual-e5-large, intfloat/multilingual-e5-large-instruct, BAAI/bge-small-en-v1.5, BAAI/bge-base-en-v1.5, BAAI/bge-large-en-v1.5.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

Configure Jina AI as the embedding provider

For more detailed instructions, see Integrate Jina AI as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "jinaAI",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Jina AI API key that you want to use. Must be the name of an existing Jina AI API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: jina-embeddings-v2-base-en, jina-embeddings-v2-base-de, jina-embeddings-v2-base-es, jina-embeddings-v2-base-code, jina-embeddings-v2-base-zh.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

Configure Mistral AI as the embedding provider

For more detailed instructions, see Integrate Mistral AI as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "mistral",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Mistral AI API key that you want to use. Must be the name of an existing Mistral AI API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: mistral-embed.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

Configure NVIDIA as the embedding provider

For more detailed instructions, see Integrate NVIDIA as an embedding provider. Your database must be in a supported region.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "metric": "cosine",
        "service": {
          "provider": "nvidia",
          "modelName": "nvidia/nv-embedqa-e5-v5"
        }
      }
    }
  }
}'

Configure OpenAI as the embedding provider

For more detailed instructions, see Integrate OpenAI as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "openai",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          },
          "parameters": {
            "organizationId": "ORGANIZATION_ID",
            "projectId": "PROJECT_ID"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the OpenAI API key that you want to use. Must be the name of an existing OpenAI API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

  • ORGANIZATION_ID: Optional. The ID of the OpenAI organization that owns the API key. Only required if your OpenAI account belongs to multiple organizations or if you are using a legacy user API key to access projects. For more information about organization IDs, see the OpenAI API reference.

  • PROJECT_ID: Optional. The ID of the OpenAI project that owns the API key. This cannot use the default project. Only required if your OpenAI account belongs to multiple organizations or if you are using a legacy user API key to access projects. For more information about project IDs, see the OpenAI API reference.

Configure Upstage as the embedding provider

For more detailed instructions, see Integrate Upstage as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "upstageAI",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Upstage API key that you want to use. Must be the name of an existing Upstage API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: solar-embedding-1-large.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

Configure Voyage AI as the embedding provider

For more detailed instructions, see Integrate Voyage AI as an embedding provider.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "vector": {
        "dimension": MODEL_DIMENSIONS,
        "metric": "SIMILARITY_METRIC",
        "service": {
          "provider": "voyageAI",
          "modelName": "MODEL_NAME",
          "authentication": {
            "providerKey": "API_KEY_NAME"
          }
        }
      }
    }
  }
}'

Replace the following:

  • COLLECTION_NAME: The name for your collection.

  • SIMILARITY_METRIC: The method you want to use to calculate vector similarity scores. The available metrics are Cosine (default), Dot Product, and Euclidean.

  • API_KEY_NAME: The name of the Voyage AI API key that you want to use. Must be the name of an existing Voyage AI API key in the Astra Portal. For more information, see Embedding provider authentication.

    Alternatively, you can omit this parameter and instead provide the authentication key in an x-embedding-api-key header. Header authentication overrides the API_KEY_NAME parameter if you set both. If you use the header instead of specifying the API_KEY_NAME parameter, you must include the header in every command that uses vectorize, including writes and vector search.

  • MODEL_NAME: The model that you want to use to generate embeddings. The available models are: voyage-2, voyage-code-2, voyage-finance-2, voyage-large-2, voyage-large-2-instruct, voyage-law-2, voyage-multilingual-2.

  • MODEL_DIMENSIONS: The number of dimensions that you want the generated vectors to have. Your chosen embedding model must support the specified number of dimensions.

    If you omit the dimension, Astra DB can use a default dimension value. However, some models don’t have default dimensions. You can use the Data API to find supported embedding providers and their configuration parameters, including dimensions ranges and default dimensions.

Create a collection that supports hybrid search

If you want to perform hybrid search on your collection, you must create a collection that has vector, lexical, and rerank enabled.

Lexical and rerank are enabled by default, but you can optionally configure the lexical analyzer and the reranker model.

For configuration details about the lexical analyzer, see Find data with CQL analyzers. The following example uses a configuration suitable for English text.

For configuration details about the reranker model, inspect the available reranker models. Only the NVIDIA llama-3.2-nv-rerankqa-1b-v2 reranking model reranker model is supported.

For configuration details about vector, see Create a collection that can store vector embeddings and Create a collection that can automatically generate vector embeddings.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "lexical": {
        "analyzer": {
            "tokenizer": {
                "name": "standard",
                "args": {}
            },
            "filters": [
                {
                   "name": "lowercase"
                },
                {
                   "name": "stop"
                },
                {
                   "name": "porterstem"
                },
                {
                   "name": "asciifolding"
                }
            ],
            "charFilters": []
        },
        "enabled": true
      },
      "rerank": {
        "enabled": true,
        "service": {
          "modelName": "nvidia/llama-3.2-nv-rerankqa-1b-v2",
          "provider": "nvidia"
        }
      },
      "vector": {
        "dimension": 1024,
        "metric": "cosine",
        "service": {
          "provider": "nvidia",
          "modelName": "nvidia/nv-embedqa-e5-v5"
        }
      }
    }
  }
}'

Create a collection that supports lexicographical matching

If you want to use lexicographical matching to find documents in your collection, you must create a collection that has lexical enabled.

Lexical is enabled by default when you create a collection, but you can optionally configure the lexical analyzer.

For configuration details about the lexical analyzer, see Find data with CQL analyzers. The following example uses a configuration suitable for English text.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "lexical": {
        "analyzer": {
            "tokenizer": {
                "name": "standard",
                "args": {}
            },
            "filters": [
                {
                   "name": "lowercase"
                },
                {
                   "name": "stop"
                },
                {
                   "name": "porterstem"
                },
                {
                   "name": "asciifolding"
                }
            ],
            "charFilters": []
        },
        "enabled": true
      }
    }
  }
}'

Create a collection and specify the default ID format

For more information about the default ID format, see Document IDs (HTTP). For allowed values, see the Parameters.

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "defaultId": {
        "type": "uuidv7"
      }
    }
  }
}'

Create a collection and specify which fields to index

For more information about selective indexing, see Indexes in collections (HTTP).

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "indexing": {
        "allow": ["city", "country"]
      }
    }
  }
}'

Create a collection and specify which fields shouldn’t be indexed

For more information about selective indexing, see Indexes in collections (HTTP).

curl -sS -L -X POST "API_ENDPOINT/api/json/v1/KEYSPACE_NAME" \
  --header "Token: APPLICATION_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "createCollection": {
    "name": "COLLECTION_NAME",
    "options": {
      "indexing": {
        "deny": ["city", "country"]
      }
    }
  }
}'

Was this helpful?

Give Feedback

How can we improve the documentation?

© Copyright IBM Corporation 2026 | Privacy policy | Terms of use |  Manage Privacy Choices

Apache, Apache Cassandra, Cassandra, Apache Tomcat, Tomcat, Apache Lucene, Apache Solr, Apache Hadoop, Hadoop, Apache Pulsar, Pulsar, Apache Spark, Spark, Apache TinkerPop, TinkerPop, Apache Kafka and Kafka are either registered trademarks or trademarks of the Apache Software Foundation or its subsidiaries in Canada, the United States and/or other countries. Kubernetes is the registered trademark of the Linux Foundation.

General Inquiries: Contact IBM