CN-122019083-A - Interactive system and interactive method of multiple artificial intelligence applications based on browser

CN122019083ACN 122019083 ACN122019083 ACN 122019083ACN-122019083-A

Abstract

The invention relates to the field of intelligent application interaction, and discloses an interaction system and an interaction method of a plurality of artificial intelligent applications based on a browser, wherein the interaction system comprises a plurality of intelligent application instances in the same user session period for session-level identification and functional role disassembly to form interaction participation relation mapping; the method comprises the steps of carrying out time alignment and semantic rearrangement on multisource interaction events from user interface operation, page content change and application internal response to generate an interaction instruction stream, continuously tracking the read-write behavior of each intelligent application on shared context elements around the interaction instruction stream to construct a context correlation structure, identifying instruction multiplexing nodes and conflict occupied areas according to the context correlation structure, simultaneously establishing an instruction distribution and recovery mechanism facing to application examples on a browser side, and carrying out dynamic correction on interaction scheduling sequences and information injection opportunities.

Inventors

LI HONGLI
TONG TIANLE
YANG JIAWEI
GUO SHU
DING WEIRONG

Assignees

贵州算家计算服务有限公司

Dates

Publication Date: 20260512
Application Date: 20260112

Claims (10)

1. An interaction method of a plurality of artificial intelligence applications based on a browser is characterized by comprising the following steps: Performing session level identification and functional role disassembly on a plurality of intelligent application instances in the same user session period, constructing an application capability description set according to application interaction types, input and output constraints and a calling triggering mode, and forming interaction participation relation mapping for a current task scene; based on interaction participation relation mapping, performing time alignment and semantic rearrangement on multi-source interaction events from user interface operation, page content change and application internal response, and generating an interaction instruction stream with execution precedence relation by introducing session context span identification and instruction dependency mark; continuously tracking the read-write behavior of each intelligent application on shared context elements around an interactive instruction stream, and constructing a context association structure reflecting a cross-application information transfer relationship by combining an instruction propagation path and a state backtracking record; identifying an instruction multiplexing node and a conflict occupation area according to a context association structure, merging and cutting execution conditions of the interactive instructions, and simultaneously establishing an instruction distribution and recovery mechanism facing an application instance at a browser side; And dynamically correcting the interactive scheduling sequence and the information injection time based on the context correlation structure and the instruction distribution and recovery mechanism.
2. The interactive method for a plurality of artificial intelligence applications based on a browser according to claim 1, wherein the process of forming the interactive participation relation map for the current task scene is as follows: Acquiring identification information, calling interface description and running state information of each loaded intelligent application instance in a current browser session; Analyzing the input parameter types, the output result formats and the calling triggering conditions of each intelligent application instance, and extracting to form application capability feature vectors; performing functional role classification and intra-session grouping on each intelligent application instance based on the application capability feature vector; and screening and sequencing the functional role groups by combining the current user task type and the matched interaction mode in the history session, and generating interaction participation relation mapping for the current task scene.
3. The method of claim 2, wherein the steps of time alignment and semantic rearrangement of multi-source interaction events from user interface operations, page content changes and application internal responses are: based on the interaction participation relation mapping, collecting user operation events, DOM change events and application internal callback events corresponding to the interaction participation relation mapping from a browser event bus; Adding uniform time stamps to all collected events, and executing time alignment processing of cross-event sources; And carrying out semantic classification and reordering on the events based on the event source types and the associated application capabilities, and outputting a multisource interaction event sequence which is sorted according to the time sequence and the semantic relation.
4. The interactive method for a plurality of artificial intelligence applications based on a browser according to claim 3, wherein the process of generating the interactive instruction stream with the execution precedence relationship comprises the following steps: the multisource interaction event sequence which is arranged according to the time sequence and the semantic relation is used as a processing object; distributing corresponding session context span identifiers to each interaction event in the multi-source interaction event sequence; Analyzing the input-output dependency relationship between the interaction events in the multi-source interaction event sequence, and generating a corresponding instruction dependency mark; and performing dependency sequencing on the multi-source interaction event sequence according to the session context span identification and the instruction dependency mark to form an interaction instruction stream with an execution precedence relationship.
5. The method for interacting a plurality of artificial intelligence applications based on a browser according to claim 4, wherein the process of continuously tracking the read-write behavior of each intelligent application on the shared context element is as follows: taking the interactive instruction stream as an execution driving sequence, and establishing an access monitoring hook for the shared context element at the browser side; Recording the reading, writing and covering operations of the corresponding intelligent application on the shared context elements when the interactive instructions in the interactive instruction stream are executed; and associating each read-write operation with the interaction instruction of the triggering operation to generate a read-write behavior log sequence containing an application identifier, an instruction identifier, a context element identifier, an access type and access time.
6. The method for interacting a plurality of artificial intelligence applications based on a browser according to claim 5, wherein the process of constructing a context correlation structure reflecting a cross-application information transfer relationship is as follows: constructing propagation paths of the context elements among different applications according to the read-write behavior log sequence; Carrying out snapshot record on the context states before and after the interactive instruction is executed to form a state backtracking record; And constructing a context association structure reflecting the cross-application information transfer relation based on the propagation path and the state backtracking record.
7. The method for interacting a plurality of artificial intelligence applications based on a browser according to claim 6, wherein the process of identifying the command multiplexing node and the conflict occupation area according to the context correlation structure comprises the following steps: Detecting multiplexing conditions that a plurality of interactive instructions point to the same context element in a context association structure, and identifying and forming an instruction multiplexing node; Performing time overlapping and access type conflict analysis on a plurality of writing edges pointing to the same context element node in the heterogeneous graph model; when the competition writing or coverage conflict judging condition is met, marking the corresponding sub-graph area as a conflict occupation area, and outputting an instruction multiplexing node set and a conflict occupation area set which comprise node identification, an associated instruction set and a conflict type label.
8. The interactive method for a plurality of artificial intelligence applications based on a browser according to claim 7, wherein the process of simultaneously establishing an instruction distribution and recovery mechanism for application instances on a browser side comprises the following steps: based on the instruction multiplexing node set and the conflict occupation area set, constructing a corresponding instruction distribution queue, a delay waiting queue and an execution recovery queue for each application instance; Reading the identification of the instruction multiplexing node to which the interactive instruction to be executed belongs and the marking information of whether the interactive instruction is positioned in the conflict occupation area or not; When the interactive instruction does not belong to any conflict occupation area, writing the interactive instruction into an instruction distribution queue of the corresponding application instance; When the interactive instruction belongs to the conflict occupation area, writing the interactive instruction into a delay waiting queue, and occupying constraint identification and retry triggering condition identification for additional resources; when the browser side detects an application instance state change event or a resource release event, re-executing conflict matching and execution condition judgment on the interactive instructions in the delay waiting queue, and migrating the interactive instructions meeting the execution conditions to the instruction distribution queue; and recording the execution result state, the resource release information and the conflict identification update information of the interactive instruction which completes execution, and writing the interactive instruction into an execution recovery queue.
9. The interactive method for a plurality of artificial intelligence applications based on the browser according to claim 8, wherein the process of dynamically correcting the interactive scheduling sequence and the information injection timing is as follows: Monitoring the association strength change of the context elements in the context association structure and the risk state change of the conflict occupation area in real time; When the change of the association strength or the conflict risk is detected, recalculating the scheduling priority of the interactive instructions in the instruction distribution queue and the delay waiting queue, and reordering the interactive instructions in the instruction distribution queue; controlling the migration of the interactive instruction between the delay waiting queue and the instruction distributing queue according to the conflict matching result and the execution condition judging result; before the interactive instruction enters the instruction distribution queue, determining whether to inject a delay identifier or an early injection identifier for the additional information of the interactive instruction according to the current state of the corresponding context element and the conflict identifier state; and executing the interactive instruction according to the updated instruction distribution queue sequence and the information injection mark.
10. An interactive system for a plurality of artificial intelligence applications based on a browser, applied to a method according to any one of claims 1 to 9, comprising: The session identification module is used for identifying a plurality of intelligent application instances in the same user session period and disassembling functional roles, and generating interaction participation relation mapping for the current task scene; The instruction generation module is used for carrying out time alignment and semantic rearrangement on the multisource interaction event based on the interaction participation relation mapping to generate an interaction instruction stream with an execution sequence relation; the context construction module is used for tracking the read-write behavior of each intelligent application on the shared context element around the interactive instruction stream and constructing a context association structure reflecting the cross-application information transfer relationship; the conflict mediation module is used for identifying the instruction multiplexing node and the conflict occupation area according to the context association structure, merging and clipping the execution conditions of the corresponding interaction instruction, and establishing an instruction distribution and recovery mechanism oriented to the application instance; and the dynamic scheduling module is used for dynamically correcting the interactive scheduling sequence and the information injection opportunity based on the context association structure and the mediation result.

Description

Interactive system and interactive method of multiple artificial intelligence applications based on browser Technical Field The invention relates to the field of intelligent application interaction, in particular to an interaction system and an interaction method of a plurality of artificial intelligent applications based on a browser. Background With the wide application of artificial intelligence technology in the fields of information retrieval, content generation, data analysis, auxiliary decision making and the like, more and more artificial intelligence applications are deployed in browser environments in the form of web services or browser plug-ins, in actual use scenes, users often need to call a plurality of artificial intelligence applications with different functions or sources at the same time in the same browser session, for example, while writing content by using text generation type applications, the data are sorted and verified by means of data analysis type applications, or the conversational application and search enhancement type application are used for information cross verification at the same time, however, the existing artificial intelligence applications based on the browser usually run in mutually independent pages, tag pages or plug-in examples, lack of unified interaction mechanism and context association capability between the applications, when users frequently switch among a plurality of artificial intelligence applications, not only need to repeatedly input the same or highly similar instructions and data, but also need to manually maintain task states and processing sequences among different applications, thus easily causing the problems of information splitting, operation redundancy and low interaction efficiency, and especially when a plurality of artificial intelligence applications are required to finish the same task, for example, the problems of the task is more important than the task, the task is amplified or the practical experience is further influenced on the aspects of the intelligent applications, and the practical experience is more important than the technical proposal. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an interaction system and an interaction method of a plurality of artificial intelligence applications based on a browser, which have the advantage of improving the overall interaction efficiency of the collaborative use of the plurality of artificial intelligence applications and solve the problems in the background art. In order to achieve the purpose of improving the overall interaction efficiency of collaborative use of multiple artificial intelligence applications, the invention provides the following technical scheme that the interaction method of the multiple artificial intelligence applications based on the browser comprises the following steps: Performing session level identification and functional role disassembly on a plurality of intelligent application instances in the same user session period, constructing an application capability description set according to application interaction types, input and output constraints and a calling triggering mode, and forming interaction participation relation mapping for a current task scene; based on interaction participation relation mapping, performing time alignment and semantic rearrangement on multi-source interaction events from user interface operation, page content change and application internal response, and generating an interaction instruction stream with execution precedence relation by introducing session context span identification and instruction dependency mark; continuously tracking the read-write behavior of each intelligent application on shared context elements around an interactive instruction stream, and constructing a context association structure reflecting a cross-application information transfer relationship by combining an instruction propagation path and a state backtracking record; identifying an instruction multiplexing node and a conflict occupation area according to a context association structure, merging and cutting execution conditions of the interactive instructions, and simultaneously establishing an instruction distribution and recovery mechanism facing an application instance at a browser side; And dynamically correcting the interactive scheduling sequence and the information injection time based on the context correlation structure and the instruction distribution and recovery mechanism. Preferably, the process of forming the interaction participation relation mapping facing the current task scene is as follows: Acquiring identification information, calling interface description and running state information of each loaded intelligent application instance in a current browser session; Analyzing the input parameter types, the output result formats and the calling triggering conditions of each intelligent application in