Multi-Agent Search Client

Patent metadata

Field	Value
Publication number	US-2025231942-A1
Application number	US-202418413732-A
Country	US
Kind code	A1
Filing date	Jan 16, 2024
Priority date	Jan 16, 2024
Publication date	Jul 17, 2025
Grant date	—

Abstract

Techniques are disclosed relating to a multi-agent search client. In an example method, a search agent application receives a query. The search agent application determines, using a search agent, one or more agents to each execute a portion of the query. The search agent application outputs, to a domain agent, a portion of the query. The search agent application determines, using the domain agent, an execution step based on the portion of the query and information about an executor. The executor generates instructions based on the execution step that are configured to generate a response to the execution step and outputs a command to execute the instructions. The executor then outputs, to the first domain agent, the response to the execution step. The search agent application outputs, using the search agent, a response to the query based on the response to the execution step.

First claim

1 . A method, comprising:
receiving, by a search agent application from a client device, a query; determining, by the search agent application using a search agent and based on the query, one or more domain agents to each execute at least a portion of the query; outputting, by the search agent application to a first domain agent, the at least the portion of the query; outputting, to a first executor, by the search agent application using the first domain agent, a first execution step based on the at least the portion of the query, wherein the first execution step is determined based on information about the first executor, wherein the first executor:
generates one or more first instructions based on the first execution step, wherein the first instructions are configured to generate a response to the first execution step;
outputs, for each of the first instructions, a command to execute the first instruction; and
outputs, to the first domain agent, based on the outputs to the commands corresponding to the first instructions, the response to the first execution step;
outputting, by the search agent application using the first domain agent, the response to the first execution step to the search agent; and outputting, by the search agent application using the search agent, to the client device, a response to the query based on the response to the first execution step.

2 . The method of claim 1 , further comprising:
outputting, to the first executor, by the search agent application using the first domain agent, a second execution step based on the at least the portion of the query and the response to the first execution step, wherein the second execution step is determined based on the information about the first executor, wherein the first executor:
determines one or more second instructions based on the second execution step, wherein the second instructions are configured to generate a response to the second execution step;
outputs, for each of the second instructions, a command to execute the second instruction; and
outputs, to the first domain agent, based on the outputs to the commands corresponding to the second instructions, the response to the second execution step; and
outputting, by the search agent application using the first domain agent, the response to the second execution step to the search agent, wherein the response to the query is further based on the response to the second execution step.

3 . The method of claim 1 , further comprising:
outputting, to the first executor, by the search agent application using a second domain agent, a second execution step based on the at least the portion of the query, wherein the second execution step is determined based on the information about the first executor, wherein the first executor:
generates one or more second instructions based on the second execution step, wherein the second instructions are configured to generate a response to the second execution step;
outputs, for each of the second instructions, a command to execute the second instruction; and
outputs, based on the outputs to the commands corresponding to the second instructions, the response to the second execution step; and
outputting, by the search agent application using the first domain agent, the response to the second execution step to the search agent, wherein the response to the query is further based on the response to the second execution step.

4 . The method of claim 3 , further comprising:
outputting, by the search agent application using the search agent, to the client device, the response to the query based on the response to the first execution step and the response to the second execution step comprises:
outputting, by the search agent application using the first domain agent, to an organizer component, the response to the first execution step and the response to the second execution step; and
generating, by the search agent application using the organizer component and based on the response to the first execution step and the response to the second execution step, information for generating the response to the query, wherein the response to the query is further based on the information for generating the response to the query.

5 . The method of claim 1 , further comprising:
outputting, to a second executor, by the search agent application using the first domain agent, a second execution step based on the at least the portion of the query, wherein the second execution step is determined based on information about the second executor, wherein the second executor:
generates one or more second instructions based on the second execution step, wherein the second instructions are configured to generate a response to the second execution step;
outputs, for each of the second instructions, a command to execute the second instruction; and
outputs, to the first domain agent, based on the outputs to the commands corresponding to the second instructions, the response to the second execution step; and
outputting, by the search agent application using the first domain agent, the response to the second execution step to the search agent, wherein the response to the query is further based on the response to the second execution step.

6 . The method of claim 1 , wherein the search agent is a first large language model (LLM) and the first domain agent is a second LLM.

7 . The method of claim 6 , wherein the first executor is a third LLM.

8 . The method of claim 6 , wherein the information about the first executor includes one or more operations available for performance by the first executor.

9 . The method of claim 8 , wherein the one or more operations available for performance by the first executor comprise application programming interface (API) requests.

10 . The method of claim 1 , wherein:
the one or more domain agents are each mapped to a domain; and determining, by the search agent application using a search agent and based on the query, the one or more domain agents to each execute the at least the portion of the query comprises determining a domain of the at least the portion of the query, the domain of the at least the portion of the query corresponding to a domain mapped to a domain agent of the one or more domain agents.

11 . The method of claim 10 , wherein the domains mapped to the one or more domain agents comprise video, chat, documents, meetings, or language.

12 . The method of claim 1 , wherein the query comprises human-readable text and a plurality of portions, each portion of the plurality of portions corresponding to one or more domains from among a plurality of domains.

13 . The method of claim 1 , further comprising:
receiving, by the search agent application using the first domain agent, from the first executor, a request for additional information about the first execution step based on the at least the portion of the query; and receiving, by the search agent application using the first executor, from the first domain agent, the additional information about the first execution step based on the at least the portion of the query.

14 . The method of claim 1 , further comprising:
receiving, by the search agent application using the first executor, from the first domain agent, a request for additional information about the response to the first execution step; and receiving, by the search agent application using the first domain agent, from the first executor, the additional information about the first execution step based on the response to the first execution step.

15 . A non-transitory computer-readable storage medium

Assignees

• Zoom Video Communications Inc

Inventors

• Dong Qiumin
• Han Lin
• Li Chen
• Lu Ying
• Ni Kai
• Tian Wang
• Wu da-yi
• Zeng Linxiao

Classifications

• G06F16/248

  Presentation of query results · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06F16/483

  using metadata automatically derived from the content · CPC title

• G06F16/3344

  using natural language analysis · CPC title

• G06F16/2455 Primary

  Query execution · CPC title