Conversational digital assistant
The present disclosure relates to computer-implemented methods, software, and systems that can use a language model trained for conversational interaction using natural language prompts. A prompt is received to request a resource from a corporate network environment. First input can be provided to a language model to determine an interface of an application from a set of applications associated with the corporate network environment. In response to determining the interface by the language model, a requirement description of the interface is obtained. Second input is provided to the language model that comprises the obtained requirement description of the interface and the received prompt, and is used as context for generating a network request definition for a request to be directed to the interface. The request is sent to the interface based on the network request definition provided by the language model to obtain the requested resource.
1 . A computer-implemented method executed by a digital assistant, comprising:
receiving, as a prompt from a user, a user request for obtaining a resource from a corporate network environment;
providing a first input to a language model for processing by the language model to identify a first interface of a first application, from one or more interfaces of each of a set of applications available in the corporate network environment, that matches the user request for obtaining the resource, wherein the first input includes the received prompt and a textual description of each of the one or more interfaces of each of the set of applications;
in response to identifying the first interface that matches the user request for obtaining the resource by using the language model, obtaining a requirement description of the first interface based on executing, by using a name of the first interface, a search at an online resource provider that stores technical documentation for the one or more interfaces;
providing a second input to the language model for processing by the language model to generate a network request definition for a call to be directed to the first interface, the second input comprising the requirement description of the first interface obtained at a result of executing the search at the online resource provider and the prompt received from the user, wherein the second input to the language model is to serve as context for generating the network request definition for the call to be directed to the first interface;
sending the call to the first interface of the first application based on the network request definition generated by the language model;
in response to the sending of the call to the first interface, obtaining the requested resource from the first application; and
providing the requested resource obtained by sending the call to the first interface of the first application to the user in response to the received prompt, wherein the first interface of the first application is identified as matching the received prompt by using the language model from the one or more interfaces of each of the set of applications available in the corporate network environment.
2 . The method of claim 1 , comprising:
providing third input to the language model to request generation of a human readable output, wherein, the third input comprises the requested resources and the received prompt.
3 . The method of claim 2 , wherein the third input comprises a definition of a format of a user interface supported by the digital assistant to be used for generating the human readable output.
4 . The method of claim 2 , comprising:
in response to receiving the generated human readable output from the language model, providing the generated human readable output at a user interface of a display device coupled to the digital assistant, wherein the generated human readable output is provided as interactive content.
5 . The method of claim 1 , wherein obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface from a storage separate from the language model.
6 . The method of claim 1 , wherein:
obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface; and
the method comprises maintaining a mapping between interfaces provided by the set of applications associated with the corporate network environment and resources providing content usable for obtaining requirement descriptions of the interfaces.
7 . The method of claim 1 , wherein obtaining the requirement description of the first interface comprises:
obtaining technical definition of the first interface, wherein the technical definition is a textual description of the first interface including (i) a list of possible attributes and (ii) a respective format for defining a request directed to the first interface, and wherein the generated network request definition is based on the format as provided with the second input including the textual description of the first interface.
8 . The method of claim 1 , wherein the requirement description of the first interface comprises example source code defining example requests directed at the first interface, and wherein obtaining the requirement description of the first interface comprises:
executing, by using the name of the first interface, a search at a database that stores source code examples defining example requests directed at the one or more interfaces.
9 . A non-transitory, computer-readable medium storing computer-readable instructions executable by a computer and configured to perform operations executed by a digital assistant, the operations comprising:
receiving, as a prompt from a user, a user request for obtaining a resource from a corporate network environment;
providing a first input to a language model for processing by the language model to identify a first interface of a first application, from one or more interfaces of each of a set of applications available in the corporate network environment, that matches the user request for obtaining the resource, wherein the first input includes the received prompt and a textual description of each of the one or more interfaces of each of the set of applications;
in response to identifying the first interface that matches the user request for obtaining the resource by using the language model, obtaining a requirement description of the first interface based on executing, by using a name of the first interface, a search at an online resource provider that stores technical documentation for the one or more interfaces;
providing a second input to the language model for processing by the language model to generate a network request definition for a call to be directed to the first interface, the second input comprising the requirement description of the first interface obtained at a result of executing the search at the online resource provider and the prompt received from the user, wherein the second input to the language model is to serve as context for generating the network request definition for the call to be directed to the first interface;
sending the call to the first interface of the first application based on the network request definition generated by the language model;
in response to the sending of the call to the first interface, obtaining the requested resource from the first application; and
providing the requested resource obtained by sending the call to the first interface of the first application to the user in response to the received prompt, wherein the first interface of the first application is identified as matching the received prompt by using the language model from the one or more interfaces of each of the set of applications available in the corporate network environment.
10 . The non-transitory, computer-readable medium of claim 9 , comprising instructions which when executed by the computer perform operations comprising:
providing third input to the language model to request generation of a human readable output, wherein, the third input comprises the requested resources and the received prompt.
11 . The non-transitory, computer-readable medium of claim 10 , wherein the third input comprises a definition of a format of a user interface supported by the digital assistant to be used for generating the human readable output.
12 . The non-transitory, computer-readable medium of claim 10 , comprising instructions which when executed by the computer perform operations comprising:
in response to receiving the generated human readable output from the language model, providing the generated human readable output at a user interface of a display device coupled to the digital assistant, wherein the generated human readable output is provided as interactive content.
13 . The non-transitory, computer-readable medium of claim 9 , wherein obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface from a storage separate from the language model.
14 . The non-transitory, computer-readable medium of claim 9 , wherein obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface; and
wherein the non-transitory, computer-readable medium comprises instructions which when executed by the computer perform operations comprising:
maintaining a mapping between interfaces provided by the set of applications associated with the corporate network environment and resources providing content usable for obtaining requirement descriptions of the interfaces.
15 . A digital assistant system comprising
a computing device; and
a computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations, the operations comprising:
receiving, as a prompt from a user, a user request for obtaining a resource from a corporate network environment;
providing a first input to a language model for processing by the language model to identify a first interface of a first application, from one or more interfaces of each of a set of applications available in the corporate network environment, that matches the user request for obtaining the resource, wherein the first input includes the received prompt and a textual description of each of the one or more interfaces of each of the set of applications;
in response to identifying the first interface that matches the user request for obtaining the resource by using the language model, obtaining a requirement description of the first interface based on executing, by using a name of the first interface, a search at an online resource provider that stores technical documentation for the one or more interfaces;
providing a second input to the language model for processing by the language model to generate a network request definition for a call to be directed to the first interface, the second input comprising the requirement description of the first interface obtained at a result of executing the search at the online resource provider and the prompt received from the user, wherein the second input to the language model is to serve as context for generating the network request definition for the call to be directed to the first interface;
sending the call to the first interface of the first application based on the network request definition generated by the language model;
in response to the sending of the call to the first interface, obtaining the requested resource from the first application; and
providing the requested resource obtained by sending the call to the first interface of the first application to the user in response to the received prompt, wherein the first interface of the first application is identified as matching the received prompt by using the language model from the one or more interfaces of each of the set of applications available in the corporate network environment.
16 . The system of claim 15 , wherein the computer-readable storage device has instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations comprising:
providing third input to the language model to request generation of a human readable output, wherein, the third input comprises the requested resources and the received prompt.
17 . The system of claim 16 , wherein the third input comprises a definition of a format of a user interface supported by the digital assistant to be used for generating the human readable output.
18 . The system of claim 16 , wherein the computer-readable storage device has instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations comprising:
in response to receiving the generated human readable output from the language model, providing the generated human readable output at a user interface of a display device coupled to the digital assistant, wherein the generated human readable output is provided as interactive content.
19 . The system of claim 15 , wherein obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface from a storage separate from the language model.
20 . The system of claim 15 , wherein obtaining the requirement description of the first interface comprises obtaining text content of a specified resource mapped to the first interface; and
wherein the computer-readable storage device has instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations comprising:
maintaining a mapping between interfaces provided by the set of applications associated with the corporate network environment and resources providing content usable for obtaining requirement descriptions of the interfaces.