Make your first API call with Google Docs

Assuming you've already has gone through the step 1 and step 2, let's now make your first API call with Google Docs.

Start with simple data-retrieving

Start by retrieving all GoogleDoc files from your source.

The response is a JSON array that includes the latest 100 Google Docs. Each item in the array is a GoogleDoc object containing default fields. You can adjust the returned fields using the select method.

Perform a semantic search

Next, perform a semantic search on these documents by adding .search.about to the command. To streamline the payload, use the select method to specify exactly what fields should be included in the response.

The response is smaller, containing only title and autoSummary due to the select command. It also includes an _additional field which provides information on the search, including certainty and distance which indicate the relevancy and accuracy of the result.

Time for Some Generative Tasks

Now that we understand semantic search and data retrieval, let's enhance the existing code by adding generative features.

This can be achieved by appending the generate command to the chain. Generally, we have two methods for the generate command: single and group. For more details about the differences between these methods, refer to the generative search page. For this guide, we'll stick to the grouping method. This method is used when we have a set of documents that we want to group. Then, we ask the AI model to generate something based on the combined data/content of these documents.

So far, we've had a search function that returns a set of search results. Now, we want to group those results and apply a generative prompt. Here's how you can achieve this using _additional{generate} in GraphQL or the .generate.fromMany command in the JS client:

In this code, we've incorporated two new methods. The limit method confines the results and context for the generative query to the top 2 results. The generate method enables the generative feature. It takes two arguments: the prompt and the properties from the GoogleDoc object. We're instructing the API to consider only the content inside the ["title", "autoSummary", "autoTopics"] fields. More information about the generative function is available here.