Completions Download API for GenSearch Responses

Last updated
Save as PDF

Your site can be configured to store generative queries and corresponding responses. This capability serves to help site administrators identify specific queries and responses for periodic auditing purposes, and to identify content gaps to improve generative responses. The report is a CSV download accessible through your browser or through the API.

Completions report overview

The Completions Report API endpoint gives you data about the queries and responses site visitors are generating using Expert LLM tools on your site. This information shows:

How users ask questions
The responses generative AI returned for those questions
The pages the responses were pulled from to answer the query.

Each row in the report represents a Completions event.

API endpoint: GET {site-url}/@api/deki/llm/completion/report?month=YYYY-MM

Admins can call the report from the browser by modifying their URL as follows: {site-url}/@api/deki/llm/completion/report?month=YYYY-MM.

Report data

This report contains the following data:

The report is updated daily at 12:00 AM PST, and contains data from the time the report has been enabled onward. Data is available one day after the query was submitted.

Field Name	Description	Type	Example
PATH_ANCESTORS (For report dates after December 2024)	The values of any path ancestors that were used	STRING	Manage Admin/Release_Notes
QUERY_DATE	The date and time the query was asked	UTC timezone Format: MM/DD/YYYY HH:MM	6/12/2024 18:54
QUERY	The question the user asked GenAI	STRING	how do i get a server token
RATING (For report dates after December 2024)	Whether the response was rated "helpful" or "unhelpful"	STRING	Unhelpful responses are "negative" Helpful responses are "positive" Unrated responses are blank
RESPONSE	The response provided to answer the user's query	STRING	" Here are the steps to get a server token: 1. Navigate to Site tools > Dashboard > Integrations > API Token Management. 2. In the Add an API token section, select the Server radio button. 3. Enter a Name for the token that is easy to track and manage. 4. Click Generate API token. 5. Record the API token information and copy your API token secret to a safe location. 6. Do not close the API Token created window without copying and recording your secret. Your API Token key will not function without your secret, and your secret will no longer be accessible once the window is closed. Both the key and the secret are necessary for integrating with the Expert API, so do not lose them."
SOURCE_IDS	The page IDs of the Kernels that answered the query Report dates after December 2024 will show page URLs instead of page IDs.	STRING	For report dates on / before December 2024: deki:page:16733,deki:page:1701,deki:page:19053 To view the page, take the page ID and add `/@go/page/{page id}` to your site URL. Example: {your-site-URL}/@go/page/16733 For report dates after December 2024: https://mysite.com/Page1, https://mysite.com/Page5, https://mysite.com/Page37
TAGS (For report dates after December 2024)	The value of any tags that have been applied as a filter.	STRING	customer
TOTAL_SOURCES (For report dates after December 2024)	The number of source IDs returned	NUMBER	10
USER_ID	The user who performed the query	NUMBER	840

Paginate responses for large downloads

Depending on your browser or operating system, the completion CSV download may contain so many lines that Excel cannot ingest them all. To solve for this, paginate the report rows to get smaller bits of information.

To get a paginated XML or JSON response:

(Optional) Check the size of the download

API endpoint: GET {site-url}/@api/deki/llm/completion/report?month=YYYY-MM&summary=true

Response:

SIZE The size of the download in bytes

ROW_COUNT
The total rows in the completion report + 1 (Reserved heading row)
Paginate responses

API endpoint: GET {site-url}/@api/deki/llm/completion/report?month=YYYY-MM&page=X&limit=Y

Response:

PAGE Page to navigate to. If left blank, the entire file will be returned.

LIMIT
Number of records to display on each page. Defaults to 100. Page parameter is required to use limit.

Example: After checking the total row count, and seeing 1,000,000 rows, you choose a limit for each pages response to be 100,000. You will consistently use limit=100000 and call the endpoint 10 times as follows:

{site-url}/@api/deki/llm/completion/report?month=YYYY-MM&page={1-10}&limit=100000

Each page result is an XML formatted response. To get a JSON formatted response, append &dream.out.format=json to the request.

SIZE	The size of the download in bytes
ROW_COUNT	The total rows in the completion report + 1 (Reserved heading row)

PAGE	Page to navigate to. If left blank, the entire file will be returned.
LIMIT	Number of records to display on each page. Defaults to 100. Page parameter is required to use limit.