Generative content service with multi-path content pipeline

1 . A computer-implemented method for providing a generative response to a user, the method comprising:

causing display of a graphical user interface of a frontend application of a content collaboration platform on a client device, the graphical user interface including a generative content interface;

in response to a natural language user input provided to a generative content interface:

performing a query analysis on the natural language user input to obtain a search feature set;

performing an intent analysis on the natural language user input to determine an intent confidence score with respect to each of a set of request classifiers, the set of request classifiers comprising a first request classifier associated with a request for an action, a second request classifier associated with a request for information, and a third request classifier associated with a request for a contact;

in accordance with a determination that a first intent confidence score associated with the first request classifier satisfies an intent confidence condition:

performing a first search of an issue tracking platform content store using the search feature set, including performing a search of a set of predefined forms and performing a search of a set of issue records, wherein respective issue records in the set of issue records are associated with respective predefined forms of the set of predefined forms; and

causing a first response to be presented, in the generative content interface, the first response including content associated with a result from the first search; and

in accordance with a determination that no respective intent confidence score satisfies the intent confidence condition:

performing a second search of the issue tracking platform content store using the search feature set;

performing a third search of a knowledge base content store using the search feature set;

performing a fourth search of a user contact content store using the search feature set; and

causing a second response to be presented, in the generative content interface, the second response including content associated with at least one result from the second search, the third search, or the fourth search.

2 . The method of claim 1 , wherein:

performing the search of the set of predefined forms includes searching the set of predefined forms, using the search feature set, to identify a first set of candidate predefined forms;

performing the search of the set of issue records includes:

searching the set of issue records, using the search feature set, to identify a set of candidate issue records; and

identifying, in an issue record of the set of candidate issue records, an identifier of a second set of candidate predefined forms; and

the method further includes receiving, from a response confidence analysis model, a ranking of at least a subset of the first set of candidate predefined forms and at least a subset of the second set of candidate predefined forms; and

the content associated with the result from the first search includes content associated with a highest ranking candidate predefined form.

3 . The method of claim 1 , wherein:

the natural language user input is provided by a user;

the result from the first search includes an identifier of a predefined form of the set of predefined forms;

the predefined form defines a user-fillable data structure having a set of input fields; and

the method further comprises:

retrieving user data associated with the user; and

populating an input field of the set of input fields of the predefined form with the user data.

4 . The method of claim 3 , wherein:

the user data is first user data;

the input field is a first input field; and

the first response further includes a prompt requesting second user data from the user, the second user data corresponding to a second input field of the set of input fields of the predefined form; and

the method further includes:

receiving, in the generative content interface, second user data from the user; and

populating the second input field of the set of input fields of the predefined form with the second user data.

5 . The method of claim 4 , further comprising, in response to a determination that the predefined form has been populated, automatically creating an issue record, in the issue tracking platform content store, that includes the predefined form.

6 . The method of claim 1 , further comprising:

providing, as input to a response confidence analysis model that is configured to predict a responsiveness of results from the first search, the second search, and the third search to the natural language user input:

results from the second search;

results from the third search; and

results from the fourth search;

receiving, from the response confidence analysis model, a ranking of at least a subset of the results from the second search, the third search, and the fourth search; and

causing a response to be presented, in the generative content interface, the response including content associated with a highest ranking search result.

7 . The method of claim 1 , wherein:

the result from the first search includes an identifier of a predefined form of the set of predefined forms; and

the first response includes:

a summary of the predefined form; and

a link to the predefined form.

8 . A computer-implemented method for providing a generative response using content selected from a set of candidate content stores, the method comprising:

causing display of a graphical user interface of a frontend application of a content collaboration platform on a client device, the graphical user interface including a generative content interface;

in response to a natural language user input provided to a generative content interface:

performing a query analysis on the natural language user input to obtain a search feature set;

performing an intent analysis on the natural language user input to determine an intent confidence score with respect to each of a set of request classifiers, the set of request classifiers comprising a first request classifier associated with a request for an action, a second request classifier associated with a request for information, and a third request classifier associated with a request for a contact;

in accordance with a determination that a first intent confidence score associated with the first request classifier satisfies an intent confidence condition, performing a first search of an issue tracking platform content store using the search feature set, performing the first search including performing a search of a set of predefined forms and performing a search of a set of issue records, wherein respective issue records in the set of issue records are associated with respective predefined forms of the set of predefined forms;

in accordance with a determination that a result of the first search satisfies a result confidence condition, causing a first response to be presented in the generative content interface, the first response including content from the issue tracking platform content store;

in accordance with a determination that the result of the first search fails to satisfy the result confidence condition:

performing a second search of a knowledge base content store using the search feature set;

performing a third search of a user contact content store using the search feature set; and

causing a second response to be presented in the generative content interface, the second response including content from at least one of the knowledge base content store or the user contact content store.

9 . The method of claim 8 , wherein:

the natural language user input is provided by a user;

the result from the first search includes an identifier of a predefined form of the set of predefined forms;

the predefined form defines a user-fillable data structure having a set of input fields; and

the method further comprises:

retrieving user data associated with the user; and

populating an input field of the set of input fields of the predefined form with the user data.

10 . The method of claim 9 , wherein:

the user data is first user data;

the input field is a first input field;

the first response further includes a prompt requesting second user data from the user, the second user data corresponding to a second input field of the set of input fields of the predefined form; and

the method further includes:

receiving, in the generative content interface, second user data from the user; and

populating the second input field of the set of input fields of the predefined form with the second user data.

11 . The method of claim 8 , wherein:

the search feature set is a keyword feature set including a set of keywords extracted from the natural language user input;

the query analysis is a first query analysis;

the method further includes performing a second query analysis on the natural language user input to obtain a semantic feature set; and

performing the first search further includes searching the issue tracking platform content store using the semantic feature set.

12 . The method of claim 8 , wherein:

the result from the first search includes an identifier of a predefined form of the set of predefined forms; and

the first response includes:

a summary of the predefined form; and

a link to the predefined form.

13 . A computer-implemented method for providing a generative response using content selected from a set of candidate content stores, the method comprising:

causing display of a graphical user interface of a frontend application of a content collaboration platform on a client device;

in response to a natural language user input provided to a generative content interface:

performing a query analysis on the natural language user input to obtain a search feature set;

performing an intent analysis on the natural language user input to determine an intent confidence score with respect to each of a set of request classifiers, the set of request classifiers comprising a first request classifier associated with a request for an action, a second request classifier associated with a request for information, and a third request classifier associated with a request for a contact;

performing a first search of an issue tracking platform content store of the set of conadidate content stores, using the search feature set, to identify a candidate form;

performing a second search of a knowledge base content store of the set of candidate content stores, using the search feature set, to identify a candidate knowledge base document;

performing a third search of user contact content store of the set of candidate content stores, using the search feature set, to identify a candidate user contact;

providing, as input to a response confidence analysis model that is configured to predict a responsiveness of results from the first search, the second search, and the third search to the natural language user input:

results from the first search and a first intent confidence score associated with the first request classifier;

results from the second search and a second intent confidence score associated with the second request classifier; and

results from the third search and a third intent confidence score associated with the third request classifier;

receiving, from the response confidence analysis model, a ranking of at least a subset of the results from the first search, the second search, and the third search, the ranking based at least in part the first, second, and third intent confidence scores; and

causing a response to be presented, in the generative content interface, the response including content associated with a highest ranking search result.

14 . The method of claim 13 , wherein performing the first search of the issue tracking platform content store comprises:

performing a search of a set of predefined forms, using the search feature set, to identify a first set of candidate predefined forms;

performing a search of a set of issue records, using the search feature set, to identify a set of candidate issue records; and

identifying, in an issue record of the set of candidate issue records, an identifier of a second set of candidate predefined forms.

15 . The method of claim 14 , further comprising:

receiving, from a response confidence analysis model, a ranking of at least a subset of the first set of candidate predefined forms and at least a subset of the second set of candidate predefined forms; and

the content associated with the highest ranking search result includes content associated with a highest ranking candidate predefined form.

16 . The method of claim 13 , wherein:

the natural language user input is provided by a user;

the results from the first search include an identifier of a predefined form of the set of predefined forms;

the predefined form defines a user-fillable data structure having a set of input fields; and

the method further comprises:

retrieving user data associated with the user; and

populating an input field of the set of input fields of the predefined form with the user data.

17 . A computer-implemented method for providing, in response to a user query, a generative response including an identifier of a predefined form, the predefined form having a user-fillable data structure, the method comprising:

causing display of a graphical user interface of a frontend application of a content collaboration platform on a client device;

in response to a natural language user input provided to a generative content interface:

analyzing the natural language user input to obtain a search feature set;

using the search feature set, searching a first content store of an issue tracking platform to identify a set of predefined forms responsive to the natural language user input, each predefined form of the set of predefined forms including a respective user-fillable data structure;

in response to a determination that a highest search confidence score associated with the set of predefined forms satisfies a confidence criteria, causing a first response to be provided in the generative content interface, the first response including an identifier of a first predefined form of the set of predefined forms associated with the highest search confidence score; and

in response to a determination that a highest search confidence score associated with the set of predefined forms fails to satisfy the confidence criteria:

using the search feature set, searching a second content store of the issue tracking platform to identify an issue record related to the search feature set;

identifying, in the issue record, an identifier of a second predefined form that was used in a completion of an issue associated with the issue record and that is responsive to the natural language user input; and

causing a second response to be provided in the generative content interface, the second response including an identifier of the second predefined form.

18 . The method of claim 17 , wherein:

the issue record includes a user-generated description of the issue associated with the issue record; and

searching the second content store to identify the issue record related to the search feature set comprises searching, with the search feature set, the user-generated description of the issue associated with the issue record.

19 . The method of claim 18 , wherein searching the second content store to identify the issue record related to the search feature set comprises performing a search of issue records associated with a completed issue status.

20 . The method of claim 17 , wherein:

the natural language user input is provided by a user; and

the method further comprises:

retrieving user data associated with the user;

populating a first input field of the first predefined form with the user data;

prompting the user for second user data;

receiving, in the generative content interface, the second user data;

populating a second input field of the first predefined form with the second user data; and

in response to a determination that the first predefined form has been populated, automatically creating a new issue record, in the issue tracking platform, that includes the first predefined form.