Astra DB Serverless quickstart for collections

network_check Beginner

query_builder 15 min

If your data is fully structured and you want to use a fixed schema, see the quickstart for tables instead.

This quickstart demonstrates how to create a collection, insert data to the collection, generate vector embeddings, and perform a vector search to find similar data.

The Next steps section discusses how to insert other types of data, use a different embedding model, insert data with pre-generated vector embeddings, or skip embedding generation.

To learn more about vector databases and vector search, see What are vector databases? and What is Vector Search.

Create a database and store your credentials

Sign in or create an Astra account.
In the Astra Portal navigation menu, click Databases, and then click Create Database.
For this quickstart, select the following:
- Serverless (Vector) as the deployment type
- Amazon Web Services as the provider
- us-east-2 as the region
Click Create Database.

Wait for your database to initialize and reach Active status. This can take several minutes.
Under Database Details, copy your database’s API endpoint.
Under Database Details, click Generate Token, then copy the token.

For this quickstart, store the endpoint and token in environment variables:

Linux or macOS
Windows

export API_ENDPOINT=API_ENDPOINT
export APPLICATION_TOKEN=APPLICATION_TOKEN

set API_ENDPOINT=API_ENDPOINT

set APPLICATION_TOKEN=APPLICATION_TOKEN

Install a client

Install one of the Astra DB Data API clients to facilitate interactions with the Data API.

Python
TypeScript
Java

Update to Python version 3.8 or later if needed.
Update to pip version 23.0 or later if needed.
Install the latest version of the astrapy package .
```
pip install "astrapy>=2.0,<3.0"
```

Update to Node version 18 or later if needed.
Update to TypeScript version 5 or later if needed. This is unnecessary if you are using JavaScript instead of TypeScript.
Install the latest version of the @datastax/astra-db-ts package .

For example:
```
npm install @datastax/astra-db-ts
```

Maven
Gradle

Update to Java version 17 or later if needed. DataStax recommends Java 21.
Update to Maven version 3.9 or later if needed.

Add a dependency to the latest version of the astra-db-java package .

pom.xml

<dependencies>
  <dependency>
    <groupId>com.datastax.astra</groupId>
    <artifactId>astra-db-java</artifactId>
    <version>VERSION</version>
  </dependency>
</dependencies>

Update to Java version 17 or later if needed. DataStax recommends Java 21.
Update to Gradle version 11 or later if needed.
Add a dependency to the latest version of the astra-db-java package .
build.gradle(.kts)
```
dependencies {
    implementation 'com.datastax.astra:astra-db-java:VERSION'
}
```

Connect to your database

The following function will connect to your database.

Copy the file into your project. You don’t need to execute the function now; the subsequent code examples will import and use this function.

Python
TypeScript
Java

quickstart_connect.py

import os
from astrapy import DataAPIClient, Database


def connect_to_database() -> Database:
    """
    Connects to a DataStax Astra database.
    This function retrieves the database endpoint and application token from the
    environment variables `API_ENDPOINT` and `APPLICATION_TOKEN`.

    Returns:
        Database: An instance of the connected database.

    Raises:
        RuntimeError: If the environment variables `API_ENDPOINT` or
        `APPLICATION_TOKEN` are not defined.
    """
    endpoint = os.environ.get("API_ENDPOINT")  (1)
    token = os.environ.get("APPLICATION_TOKEN")

    if not token or not endpoint:
        raise RuntimeError(
            "Environment variables API_ENDPOINT and APPLICATION_TOKEN must be defined"
        )

    # Create an instance of the `DataAPIClient` class
    client = DataAPIClient()

    # Get the database specified by your endpoint and provide the token
    database = client.get_database(endpoint, token=token)

    print(f"Connected to database {database.info().name}")

    return database

1	Store your database’s endpoint and application token in environment variables named `API_ENDPOINT` and `APPLICATION_TOKEN`, as instructed in Create a database and store your credentials.

quickstart-connect.ts

import { DataAPIClient, Db } from "@datastax/astra-db-ts";

/**
 * Connects to a DataStax Astra database.
 * This function retrieves the database endpoint and application token from the
 * environment variables `API_ENDPOINT` and `APPLICATION_TOKEN`.
 *
 * @returns An instance of the connected database.
 * @throws Will throw an error if the environment variables
 * `API_ENDPOINT` or `APPLICATION_TOKEN` are not defined.
 */
export function connectToDatabase(): Db {
  const { API_ENDPOINT: endpoint, APPLICATION_TOKEN: token } =
    process.env; (1)

  if (!token || !endpoint) {
    throw new Error(
      "Environment variables API_ENDPOINT and APPLICATION_TOKEN must be defined.",
    );
  }

  // Create an instance of the `DataAPIClient` class
  const client = new DataAPIClient();

  // Get the database specified by your endpoint and provide the token
  const database = client.db(endpoint, { token });

  console.log(`Connected to database ${database.id}`);

  return database;
}

1	Store your database’s endpoint and application token in environment variables named `API_ENDPOINT` and `APPLICATION_TOKEN`, as instructed in Create a database and store your credentials.

src/main/java/com/quickstart/QuickstartConnect.java

package com.quickstart;

import com.datastax.astra.client.DataAPIClient;
import com.datastax.astra.client.databases.Database;

public class QuickstartConnect {

  /**
   * Connects to a DataStax Astra database. This function retrieves the database endpoint and
   * application token from the environment variables `API_ENDPOINT` and
   * `APPLICATION_TOKEN`.
   *
   * @return an instance of the connected database
   * @throws IllegalStateException if the environment variables `API_ENDPOINT` or
   *     `APPLICATION_TOKEN` are not defined
   */
  public static Database connectToDatabase() {
    String endpoint = System.getenv("API_ENDPOINT"); (1)
    String token = System.getenv("APPLICATION_TOKEN");

    if (endpoint == null || token == null) {
      throw new IllegalStateException(
          "Environment variables API_ENDPOINT and APPLICATION_TOKEN must be defined");
    }

    // Create an instance of `DataAPIClient` with your token.
    DataAPIClient client = new DataAPIClient(token);

    // Get the database specified by your endpoint.
    Database database = client.getDatabase(endpoint);

    System.out.println("Connected to database.");

    return database;
  }
}

1	Store your database’s endpoint and application token in environment variables named `API_ENDPOINT` and `APPLICATION_TOKEN`, as instructed in Create a database and store your credentials.

Create a collection

The following script will create an empty collection in your database.

Copy the script into your project.
If needed, update the import path to the "connect to database" function from the previous section.
Execute the script.

For information about executing scripts, refer to the documentation for the language that your script uses.

Once the script completes, you should see a printed message confirming the collection creation.

Python
TypeScript
Java

quickstart_create_collection.py

from quickstart_connect import connect_to_database  (1)
from astrapy.constants import VectorMetric
from astrapy.info import (
    CollectionDefinition,
    CollectionVectorOptions,
    VectorServiceOptions,
)


def main() -> None:
    database = connect_to_database()

    collection = database.create_collection(
        "quickstart_collection",  (2)
        definition=CollectionDefinition(
            vector=CollectionVectorOptions(
                metric=VectorMetric.COSINE,
                service=VectorServiceOptions(
                    provider="nvidia",  (3)
                    model_name="NV-Embed-QA",
                ),
            )
        ),
    )

    print(f"Created collection {collection.full_name}")


if __name__ == "__main__":
    main()

1	This is the `connect_to_database` function from the previous section. Update the import path if necessary. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	This collection will use the Astra-hosted NVIDIA embedding model to generate vector embeddings. This is currently only supported in certain regions. Ensure that your database is in the Amazon Web Services us-east-2 region, as instructed in Create a database and store your credentials.

This example creates an untyped collection, but you can define a client-side type for your collection to help statically catch errors. For examples, see Create a collection and Typing Collections and Tables.

quickstart-create-collection.ts

import { connectToDatabase } from "./quickstart-connect"; (1)

(async function () {
  const database = connectToDatabase();

  const collection = await database.createCollection(
    "quickstart_collection", (2)
    {
      vector: {
        service: {
          provider: "nvidia", (3)
          modelName: "NV-Embed-QA",
        },
      },
    },
  );

  console.log(
    `Created collection ${collection.keyspace}.${collection.name}`,
  );
})();

1	This is the `connectToDatabase` function from the previous section. Update the import path if necessary. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	This collection will use the Astra-hosted NVIDIA embedding model to generate vector embeddings. This is currently only supported in certain regions. Ensure that your database is in the Amazon Web Services us-east-2 region, as instructed in Create a database and store your credentials.

src/main/java/com/quickstart/QuickstartCreateCollectionDemo.java

package com.quickstart;

import com.datastax.astra.client.collections.Collection;
import com.datastax.astra.client.collections.definition.CollectionDefinition;
import com.datastax.astra.client.collections.definition.documents.Document;
import com.datastax.astra.client.core.vector.SimilarityMetric;
import com.datastax.astra.client.databases.Database;

public class QuickstartCreateCollectionDemo {

  public static void main(String[] args) {
    Database database = QuickstartConnect.connectToDatabase(); (1)

    Collection<Document> collection =
        database.createCollection(
            "quickstart_collection", (2)
            new CollectionDefinition()
                .vectorSimilarity(SimilarityMetric.COSINE)
                .vectorize("nvidia", "NV-Embed-QA")); (3)

    System.out.println("Created collection " + collection.getCollectionName());
  }
}

1	This is the `connectToDatabase` function from the previous section. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	This collection will use the Astra-hosted NVIDIA embedding model to generate vector embeddings. This is currently only supported in certain regions. Ensure that your database is in the Amazon Web Services us-east-2 region, as instructed in Create a database and store your credentials.

Insert data to your collection

The following script will insert data from a JSON file to your collection.

Copy the script into your project.
Download the quickstart_dataset.json sample dataset (76 kB). This dataset is a JSON array describing library books.
Replace PATH_TO_DATA_FILE in the script with the path to the dataset.
If needed, update the import path to the "connect to database" function from the previous section.
Execute the script.

For information about executing scripts, refer to the documentation for the language that your script uses.

Once the script completes, you should see a printed message confirming the insertion of 100 documents.

Python
TypeScript
Java

quickstart_insert_to_collection.py

import json
from quickstart_connect import connect_to_database  (1)
from astrapy.data_types import DataAPIDate


def main() -> None:
    database = connect_to_database()

    collection = database.get_collection("quickstart_collection")  (2)

    data_file_path = "PATH_TO_DATA_FILE"  (3)

    # Read the JSON file and parse it into a JSON array
    with open(data_file_path, "r", encoding="utf8") as file:
        json_data = json.load(file)

    # Assemble the documents to insert:
    # - Convert the date string into a DataAPIDate
    # - Add a $vectorize field (4)
    documents = [
        {
            **data,
            "due_date": (
                DataAPIDate.from_string(data["due_date"])
                if data.get("due_date")
                else None
            ),
            "$vectorize": (
                f"summary: {data['summary']} | genres: {', '.join(data['genres'])}"
            ),
        }
        for data in json_data
    ]

    # Insert the data
    inserted = collection.insert_many(documents)

    print(f"Inserted {len(inserted.inserted_ids)} documents.")


if __name__ == "__main__":
    main()

1	This is the `connect_to_database` function from the previous section. Update the import path if necessary. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	Replace `PATH_TO_DATA_FILE` with the path to the JSON data file.
4	When you insert data to a collection that can automatically generate embeddings, you can specify a `$vectorize` value for the data. The `$vectorize` value will be used to generate vector embeddings. `$vectorize` can be any string and should include the parts of the data that you want to be considered when you search for similar data with a vector search.

quickstart-insert-to-collection.ts

import { connectToDatabase } from "./quickstart-connect"; (1)
import fs from "fs";

(async function () {
  const database = connectToDatabase();

  const collection = database.collection("quickstart_collection"); (2)

  const dataFilePath = "PATH_TO_DATA_FILE"; (3)

  // Read the JSON file and parse it into a JSON array
  const rawData = fs.readFileSync(dataFilePath, "utf8");
  const jsonData = JSON.parse(rawData);

  // Assemble the documents to insert:
  // - Convert the date string into a Date
  // - Add a $vectorize field (4)
  const documents = jsonData.map((data: any) => ({
    ...data,
    due_date: data.due_date ? new Date(data.due_date) : null,
    $vectorize: `summary: ${data["summary"]} | genres: ${data["genres"].join(", ")}`,
  }));

  // Insert the data
  const inserted = await collection.insertMany(documents);

  console.log(`Inserted ${inserted.insertedCount} documents.`);
})();

1	This is the `connectToDatabase` function from the previous section. Update the import path if necessary. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	Replace `PATH_TO_DATA_FILE` with the path to the JSON data file.
4	When you insert data to a collection that can automatically generate embeddings, you can specify a `$vectorize` value for the data. The `$vectorize` value will be used to generate vector embeddings. `$vectorize` can be any string and should include the parts of the data that you want to be considered when you search for similar data with a vector search.

src/main/java/com/quickstart/QuickstartInsertToCollectionDemo.java

package com.quickstart;

import com.datastax.astra.client.collections.Collection;
import com.datastax.astra.client.collections.commands.results.CollectionInsertManyResult;
import com.datastax.astra.client.collections.definition.documents.Document;
import com.datastax.astra.client.databases.Database;
import com.fasterxml.jackson.databind.JsonNode;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.fasterxml.jackson.databind.node.ArrayNode;
import com.fasterxml.jackson.databind.node.ObjectNode;
import java.io.IOException;
import java.nio.charset.StandardCharsets;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.util.ArrayList;
import java.util.List;
import java.util.stream.StreamSupport;

public class QuickstartInsertToCollectionDemo {

  public static void main(String[] args) throws IOException {
    Database database = QuickstartConnect.connectToDatabase(); (1)

    Collection<Document> collection = database.getCollection("quickstart_collection"); (2)

    // Initialize Jackson ObjectMapper
    ObjectMapper objectMapper = new ObjectMapper();

    // Read the JSON file and parse it into a JSON array (ArrayNode). (3)
    String rawData =
        Files.readString(Paths.get("PATH_TO_DATA_FILE"), StandardCharsets.UTF_8);

    ArrayNode jsonData = (ArrayNode) objectMapper.readTree(rawData);

    // Convert the data to a list of Documents, and
    // add a $vectorize field to each piece of data. (4)
    List<Document> documents = new ArrayList<>();

    for (JsonNode data : jsonData) {
      if (data instanceof ObjectNode obj) {
        String summary = obj.path("summary").asText("");
        String genres =
            String.join(
                ", ",
                StreamSupport.stream(obj.withArray("genres").spliterator(), false)
                    .map(JsonNode::asText)
                    .toList());
        String summary_genres = String.format("summary: %s | genres: %s", summary, genres);
        obj.put("$vectorize", summary_genres);
      }

      documents.add(Document.parse(data.toString()));
    }

    CollectionInsertManyResult result = collection.insertMany(documents);

    System.out.println("Inserted " + result.getInsertedIds().size() + " documents.");
  }
}

1	This is the `connectToDatabase` function from the previous section. To use the function, ensure you stored your database’s endpoint and application token in environment variables as instructed in Create a database and store your credentials.
2	This script creates a collection named `quickstart_collection`. If you want to use a different name, change the name before running the script.
3	Replace `PATH_TO_DATA_FILE` with the path to the JSON data file.
4	When you insert data to a collection that can automatically generate embeddings, you can specify a `$vectorize` value for the data. The `$vectorize` value will be used to generate vector embeddings. `$vectorize` can be any string and should include the parts of the data that you want to be considered when you search for similar data with a vector search.

Find data in your database

After you insert data to your database, you can search the data. In addition to traditional database filtering, you can perform a vector search to find data that is most similar to a search string.

The following script performs three searches on the sample data that you loaded in Insert data to your collection.

Python
TypeScript
Java

quickstart_find.py

from quickstart_connect import connect_to_database  (1)

def main() -> None:
    database = connect_to_database()

    collection = database.get_collection("quickstart_collection")  (2)

    # Find documents that match a filter
    print("\nFinding books with rating greater than 4.7...")

    rating_cursor = collection.find({"rating": {"$gt": 4.7}})

    for document in rating_cursor:
        print(f"{document['title']} is rated {document['rating']}")

    # Perform a vector search to find the closest match to a search string
    print("\nUsing vector search to find a single scary novel...")

    single_vector_match = collection.find_one(sort={"$vectorize": "A scary novel"})

    print(f"{single_vector_match['title']} is a scary novel")

    # Combine a filter, vector search, and projection to find the 3 books with
    # more than 400 pages that are the closest matches to a search string,
    # and just return the title and author
    print("\nUsing filters and vector search to find 3 books with more than 400 pages that are set in the arctic, returning just the title and author...")

    vector_cursor = collection.find(
        {"number_of_pages": {"$gt": 400}},
        sort={"$vectorize": "A book set in the arctic"},
        limit=3,
        projection={"title": True, "author": True}
    )

    for document in vector_cursor:
        print(document)


if __name__ == "__main__":
    main()

1	This is the `connect_to_database` function from the previous section. Update the import path if necessary.
2	If you changed the collection name in the previous script, change it in this script as well.

quickstart-find.ts

import { connectToDatabase } from "./quickstart-connect"; (1)

(async function () {
  const database = connectToDatabase();

  const collection = database.collection("quickstart_collection"); (2)

  // Find documents that match a filter
  console.log("\nFinding books with rating greater than 4.7...");

  const ratingCursor = collection.find(
    { rating: { $gt: 4.7 } },
    { limit: 10 },
  );

  for await (const document of ratingCursor) {
    console.log(`${document.title} is rated ${document.rating}`);
  }

  // Perform a vector search to find the closest match to a search string
  console.log("\nUsing vector search to find a single scary novel...");

  const singleVectorMatch = await collection.findOne(
    {},
    { sort: { $vectorize: "A scary novel" } },
  );

  console.log(`${singleVectorMatch?.title} is a scary novel`);

  // Combine a filter, vector search, and projection to find the 3 books with
  // more than 400 pages that are the closest matches to a search string,
  // and just return the title and author
  console.log("\nUsing filters and vector search to find 3 books with more than 400 pages that are set in the arctic, returning just the title and author...");

  const vectorCursor = collection.find(
    { number_of_pages: { $gt: 400 } },
    {
      sort: { $vectorize: "A book set in the arctic" },
      limit: 3,
      projection: { title: true, author: true },
    },
  );

  for await (const document of vectorCursor) {
    console.log(document);
  }
})();

1	This is the `connectToDatabase` function from the previous section. Update the import path if necessary.
2	If you changed the collection name in the previous script, change it in this script as well.

src/main/java/com/quickstart/QuickstartFindDemo.java

package com.quickstart;

import static com.datastax.astra.client.core.query.Projection.include;

import com.datastax.astra.client.collections.Collection;
import com.datastax.astra.client.databases.Database;
import com.datastax.astra.client.collections.definition.documents.Document;
import com.datastax.astra.client.core.query.Filter;
import com.datastax.astra.client.core.query.Filters;
import com.datastax.astra.client.core.query.Sort;
import com.datastax.astra.client.collections.commands.options.CollectionFindOneOptions;
import com.datastax.astra.client.collections.commands.options.CollectionFindOptions;

public class QuickstartFindDemo {

  public static void main(String[] args) {

    Database database = QuickstartConnect.connectToDatabase(); (1)

    Collection<Document> collection = database.getCollection(
      "quickstart_collection" (2)
    );

    // Find documents that match a filter
    System.out.println("\nFinding books with rating greater than 4.7...");

    Filter filter = Filters.gt("rating", 4.7);

    CollectionFindOptions options = new CollectionFindOptions().limit(10);

    collection
        .find(filter, options)
        .forEach(
            document -> {
              System.out.println(
                  document.getString("title") + " is rated " + document.get("rating"));
            });

    // Perform a vector search to find the closest match to a search string
    System.out.println("\nUsing vector search to find a single scary novel...");

    CollectionFindOneOptions options2 = new CollectionFindOneOptions().sort(Sort.vectorize("A scary novel"));

    collection
        .findOne(options2)
        .ifPresent(
            document -> {
              System.out.println(document.getString("title") + " is a scary novel");
            });

    // Combine a filter, vector search, and projection to find the 3 books with
    // more than 400 pages that are the closest matches to a search string,
    // and just return the title and author
    System.out.println("\nUsing filters and vector search to find 3 books with more than 400 pages that are set in the arctic, returning just the title and author...");

    Filter filter3 = Filters.gt("number_of_pages", 400);

    CollectionFindOptions options3 =
        new CollectionFindOptions()
            .limit(3)
            .sort(Sort.vectorize("A book set in the arctic"))
            .projection(include("title", "author"));

    collection
        .find(filter3, options3)
        .forEach(
            document -> {
              System.out.println(document);
            });
  }
}

1	This is the `connectToDatabase` function from the previous section.
2	If you changed the collection name in the previous script, change it in this script as well.

Next steps

For more practice, you can continue building with the database that you created here. For example, try inserting more data to the collection, or try different searches. The Data API reference provides code examples for various operations.

Upload data from different sources

This quickstart demonstrated how to insert data from a JSON file, but you can insert data from many sources, including CSV and PDF files.

If you can convert your data into JSON, you can use the example from this quickstart.
If your data is in unstructured files, such as PDF files, you can use the Astra Portal or write a script to use the Unstructured.io integration to insert your data.

This quickstart also demonstrated how to insert data to a collection, which uses a flexible schema. If your data is structured and you want to use a fixed schema, you can use a table instead of a collection. See the quickstart for tables.

Use a different method to generate vector embeddings

This quickstart used the Astra-hosted NVIDIA embedding model to generate vector embeddings. You can also use other embedding models, or you can insert data with pre-generated vector embeddings (or without vector embeddings) and skip embedding.

To use a different embedding model, see Auto-generate embeddings with vectorize.
To insert pre-embedded data, you need to specify the vector dimensions and similarity metric instead of specifying the embedding provider. See the API documentation for collections.
To skip embedding, you can create a collection without any vector options. See the API documentation for collections.

Perform more complex searches

This quickstart demonstrated how to find data using filters and vector search. To learn more about the searches you can perform, see Find a document, Find documents, and Find data with vector search.

Use different database settings

For this quickstart, you need a Serverless (Vector) database in the Amazon Web Services us-east-2 region, which is required for the Astra-hosted NVIDIA embedding model integration.

For production databases, you might use different database settings.

For more information, see Astra DB Serverless database regions and maintenance schedules and Create an Astra DB Serverless database.

For more examples, see the integration guides, code examples, and tutorials.

Astra DB Serverless quickstart for collections

Create a database and store your credentials

Install a client

Connect to your database

Create a collection

Insert data to your collection

Find data in your database

Next steps

Was this helpful?

Give Feedback