Get Assistance from Capella iQ

  • Capella Columnar
  • how-to
    +
    Capella iQ is a service that leverages a large language model to help you write queries.
    Remember that Capella iQ harnesses the power of third-party large language models (LLM) to improve your queries. Do not provide sensitive data in the prompt such as personally identifiable information or any confidential information that you would not want shared. Because technology built on LLMs can also suffer from hallucinations from time to time, be sure to review the output.

    Capella iQ is your partner in Capella Columnar, allowing you to work faster and assist you directly in the Capella UI. It uses a large language model (LLM) to generate queries based on plain language prompts.

    For example, if you loaded the travel-sample dataset, you could use a prompt like Show me the flights flying from JFK to SFO to get Capella iQ to generate a SQL++ for Capella columnar query using the airline and route collections. Capella iQ also produces responses based on context - you could refine the query with another prompt, such as Only show those with the call sign United, without repeating your first request.

    Prerequisites

    • If you’re the Organization Owner, accept Capella iQ’s supplemental terms and turn it on for your organization. For more information, see Configure Capella iQ.

    • (Optional) Create a data source for Couchbase sample data. For the first example on this page, you need the travel-sample dataset. See Access Data.

    Open Capella iQ

    Use Capella iQ from the Capella Columnar workbench. For more information about the workbench, see Query and Explore with the Workbench.

    From the Workbench, click the iQ tab.

    Generate SQL++ for Capella columnar Queries

    You can generate SQL++ for Capella columnar queries with conversational questions using the sample data or your own data.

    1. At the top of the iQ pane, use the lists to choose a database, scope, and at least one collection for your iQ queries.

      After you choose the collection, iQ suggests some sample queries to start working with your data.

    2. Run a suggested SQL++ for Capella columnar query or provide a prompt for a custom query:

      • Click Run for a suggestion to have iQ return a related SQL++ for Capella columnar query.

      • To generate a custom query, type your prompt in the iQ message field. After you review the query that Capella iQ returns, click Run.

    3. Modify the query using follow-up prompts.

    Always review every query that iQ returns before running it.

    Example: Generate a SQL++ for Capella columnar Query with travel-sample Sample Data

    If you have loaded the travel-sample dataset into Capella Columnar, you can try the following with Couchbase iQ to generate SQL++ for Capella columnar queries:

    1. At the top of the iQ pane, use the lists to select the following:

      • The travel-sample database

      • The inventory scope

      • The airline and route collections

    2. Enter the following prompt:

      Show me airlines that fly from JFK to SFO.
    3. Click Run and look at the results.

    4. Enter another prompt to modify and iterate on iQ’s first query:

      This should be the number of flights by each airline.
    5. Click Chart to view the results of the query in chart format, instead.

    You can continue iterating on the query and working with the query results inside the workbench.

    Visualize Query Results with iQ Insights

    After running a Capella iQ query, use iQ Insights to generate a variety of relevant insights into your query results. With the power of AI, iQ Insights generates relevant questions, descriptions, and visualizations, helping you gain a better understanding of your data.

    To gain insights on your query results, select the iQ Insights tab in the query results pane and click the Generate Now button.

    For more information, see Explore Results with iQ Insights.

    Create Database Objects and Generate Sample Documents

    You can also prompt iQ to create new database objects, such as databases, scopes, and collections. You can choose to turn these collections into standalone collections, and populate them with generated sample documents.

    Example: Create Database Objects, Populate with Generated Data, and Use a JOIN

    Try the following to create new database objects and generate sample data:

    1. In the iQ pane, enter the following prompt to create a new database:

      Create a new database called "finance".
    2. Click Run to create the database.

    3. With the new finance database selected in the lists at the top of the iQ pane, enter another prompt to create a new scope:

      Create a scope called "data".
    4. Click Run to create the scope.

    5. With the new data scope also selected in the lists, enter another prompt to create a new collection:

      Create a collection called "customerPersonalData".
    6. Click Run to create the collection.

    7. Enter another prompt to generate sample data:

      Generate 10 documents to be inserted into customerPersonalData with the fields name, age, gender, address, and zipcode.
    8. Click Run to insert the generated documents.

    9. Enter another prompt to create another new collection:

      Create a collection called "customerAccountData".
    10. Click Run to create the collection.

    11. Enter another prompt to generate more sample documents:

      Generate 10 documents to be inserted into customerAccountData. Use the same name values as in customerPersonalData. Add the fields checking account number, checking account balance, savings account number, and savings account balance.
    12. Click Run.

    13. With the customerPersonalData and customerAccountData collections selected in the lists, enter a prompt for a SQL++ for Capella columnar query that joins data from the 2 collections:

      Give me the name, age, checking account balance and savings account balance for all customers.
    14. Click Run and look at the results.

    Keep iterating on queries with iQ, or explore your own queries and results with the workbench.