Records
This section describes the API calls used to create, read, update and delete database records. The page builds upon (and assumes you've read) the API Usage Guide. Please ensure you're familiar with the API calling conventions before reading on.
Search for records in a database (by meaning and concepts)
The database to search for records.
The globally unique identifier of the database.
What you would like to search for, in natural language. Typically supplied by an end user through a search bar.
A number between 0 and 1 that indicates how much "creative liberty" a model should have in the results it surfaces. Values closer to 0 attempt to faithfully match the language used in the query, whereas values closer to 1 will maximally search for records that are on point conceptually regardless of the language used.
The maximum number of results to return in this page.
A token used to paginate the search results. Cursors returned in previous responses can be used in subsequent requests.
The number of results returned.
Whether there are more results on the next page
A token that can be used to request the next page of results. It's only useful if has_more is true.
A list of results, grouped by record and sorted by score.
A relative indicator of the AI model's confidence that the record associated with this result is a good match for the query.
The record associated with this search result.
The globally unique identifier of the record.
The database associated with this record.
The globally unique identifier of the database.
The content of the record, in the form of arbitrary JSON.
A list of relevant passages from the record, sorted by score.
A relative indicator of the AI model's confidence that this passage is a good match for the query.
The index associated with this match.
The globally unique identifier of the index.
The extracted, normalized text from the record. This value is useful for presentation but not highlighting as it may be subtly different from the original text in terms of spacing and minor details.
A list of exact, non-normalized passages drawn from the record. This is useful for in-document highlighting. However each individual item may be too specific to display without context - prefer to use the passage field for presentation.
List records in a database
The database to search for records.
The globally unique identifier of the database.
The maximum number of results to return in this page.
A token used to paginate the search results. Cursors returned in previous responses can be used in subsequent requests.
The number of results returned.
Whether there are more results on the next page
A token that can be used to request the next page of results. It's only useful if has_more is true.
A list of results.
The globally unique identifier of the record.
The database associated with this record.
The globally unique identifier of the database.
The content of the record, in the form of arbitrary JSON.
Fetch a record
The globally unique identifier of the record to fetch.
The globally unique identifier of the record.
The database associated with this record.
The globally unique identifier of the database.
The content of the record, in the form of arbitrary JSON.
Create a record
The database to hold this record.
The globally unique identifier of the database.
The content of the record, in the form of arbitrary JSON.
A globally unique identifier of the newly created record.
Update a record
A globally unique identifier of the record to update.
The content of the record, in the form of arbitrary JSON.
Always null if successful.
Delete a record
The globally unique identifier of the record to delete.
Always null if successful.