Self-evolution memory-based intelligent agent query processing method and device, equipment and medium

Abstract

This disclosure relates to a query processing method, apparatus, device, and medium based on a self-evolving memory agent, belonging to the technical field of data processing. The method includes: responding to a current query request output by a first user, invoking a data processing model under a target configuration node within a set memory configuration period, and extracting a target memory feature set matching the current query request; wherein the feature elements of the target memory feature set are associated with the target configuration node; invoking a preset data detection model, using the target configuration node and the memory features as input conditions, to obtain a current working memory sequence; wherein the current working memory sequence is used to characterize the solution result of the current query request.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and more specifically, to a query processing method, apparatus, device, and medium based on a self-evolving memory intelligent agent. Background Technology

[0002] In scenarios involving complex task solving, information analysis, and decision support, intelligent agents typically need to handle a series of interconnected consecutive queries or task requests. Insights, rule constraints, and exceptional experiences gained from early tasks directly impact the reasoning strategies and execution efficiency of subsequent tasks. Therefore, general-purpose intelligent agents not only need to achieve accurate reasoning and execution for single tasks but also need to continuously accumulate and reuse historical experience during continuous task processing to achieve cumulative reasoning and capability iteration across tasks. Currently, existing intelligent agents, such as DS-Agent, DS-STAR, and MLE-STAR, focus on multi-step reasoning and code generation, possessing strong capabilities in decomposing and solving single queries, but lack cross-query knowledge accumulation and reuse mechanisms, resulting in lower accuracy. Summary of the Invention

[0003] One objective of this disclosure is to provide a novel technical solution based on query processing of self-evolving memory agents.

[0004] According to a first aspect of this disclosure, a query processing method based on a self-evolving memory agent is provided, the method comprising: In response to the current query request output by the first user, the data processing model under the target configuration node in the set memory configuration period is invoked to extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node; A preset data detection model is invoked, using the target configuration node and the memory features as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

[0005] Optionally, the memory configuration cycle includes an initial cognitive configuration node; before responding to a query request output by the first user, invoking a preset data processing model, and extracting memory features matching the query request, the method further includes: In the cognitive configuration node, environmental parameters of the current environment where the first user is located are extracted to generate initial cognitive memory entries; Based on the preset constraints and the initial cognitive memory entries, set the model processing configuration of the data processing model; In response to the current query request output by the first user, the data processing model under the target configuration node in the set memory configuration period is invoked to extract the target memory feature set matching the current query request, including: In response to the current query request output by the first user, the data processing model under the model processing configuration is invoked to extract the initial cognitive memory that matches the current query request as the target memory feature set.

[0006] Optionally, the data processing model generates a historical working memory sequence based on historical query requests output by the second user; the memory configuration period includes a fusion cognitive configuration node; after setting the model processing configuration of the data processing model according to preset constraints and the initial cognitive memory entries, the method further includes: Based on the historical query requests and the historical working memory sequence, an update sequence for a given memory type is constructed; The update sequence is used to first update the generating model of the data processing model, and then update the merging model of the data processing model to obtain the updated data processing model.

[0007] Optionally, the defined memory type includes an experience memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the core experience information of the data processing model is extracted and used as an update sequence; wherein, the core experience information includes problem description, execution status, inference results, key insight chains, execution results and detection tags.

[0008] Optionally, the defined memory type includes a heuristic memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the heuristic core information of the data processing model is extracted and used as the update sequence; wherein, the heuristic core information includes cluster description, cluster-level chain results, general insight chain, applicable conditions, pending hypotheses and cluster detection labels.

[0009] Optionally, the defined memory type includes a cognitive memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the cognitive core information of the data processing model is extracted and used as an update sequence; wherein, the cognitive core information includes a task domain environment insight chain.

[0010] Optionally, before updating the generator model of the data processing model and then updating the merged model of the data processing model through the update sequence to obtain the updated data processing model, the method further includes: The data processing model uses a validation model to filter target update sequences that meet the set validation conditions from the update sequences. The step of updating the data processing model by first updating the generator model and then updating the merged model to obtain the updated data processing model includes: The target update sequence is used to first update the generating model of the data processing model, and then update the merging model of the data processing model to obtain the updated data processing model.

[0011] According to a second aspect of this disclosure, a query processing apparatus based on a self-evolving memory agent is also provided, the apparatus comprising: The response module is used to respond to the current query request output by the first user, call the data processing model under the target configuration node in the set memory configuration period, and extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node; The calling module is used to call a preset data detection model, using the target configuration node and the memory features as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

[0012] According to a third aspect of this disclosure, an electronic device is also provided, including a memory and a processor, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method according to a first aspect of this disclosure.

[0013] According to a fourth aspect of this disclosure, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the method described according to a first aspect of this disclosure.

[0014] According to a fifth aspect of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described according to a first aspect of this disclosure.

[0015] One beneficial effect of this disclosure is that the query processing method based on self-evolving memory intelligent agents provided by the present invention can realize a cross-query knowledge accumulation and reuse mechanism through data processing models and data detection models, so as to improve the accuracy of query results for query requests.

[0016] Other features and advantages of the embodiments of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the present disclosure and, together with their description, serve to explain the principles of the embodiments of the present disclosure.

[0018] Figure 1 This is a flowchart illustrating a query processing method based on a self-evolving memory agent according to one embodiment; Figure 2 This is a schematic diagram of the structure of a data processing model according to one embodiment; Figure 3 This is a block diagram of a query processing device based on a self-evolving memory agent according to one embodiment; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to one embodiment. Detailed Implementation

[0019] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the parts and steps set forth in these embodiments do not limit the scope of the invention.

[0020] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0021] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0022] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0023] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0024] <Method Implementation> Figure 1 This is a flowchart illustrating a query processing method based on a self-evolving memory intelligent agent according to one embodiment. The implementing entity is a smart terminal, which can be a mobile phone, tablet, or personal computer, etc., and is not limited thereto.

[0025] like Figure 1 As shown, the query processing method based on self-evolving memory intelligent agents in this embodiment may include the following steps S110 to S120: Step S110: In response to the current query request output by the first user, the data processing model under the target configuration node in the set memory configuration period is invoked to extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node.

[0026] In this embodiment, the data processing model can be an existing general network model or an optimized network model; no limitation is made here. Specifically, the data processing model can be as follows: Figure 2 The Plan Agent shown.

[0027] In this embodiment, the data processing model generates an executable process sequence based on the query request q and the retrieved cognitive context Mq, as shown in the following expression: P:(q,Mq)→{Step1,…,Stepn} In this process sequence, each process node Step contains four major fields: "target description, required tools, dependency memory, and expected intermediate results". Taking process node s1 as an example, Step1: Use the environment state awareness interface to obtain the real-time state of the task domain. The task_id of the environment state awareness interface is the unique index in the dependency cognitive memory. Expected result: No missing cognitive memory perception.

[0028] In this embodiment, the data processing model prioritizes reusing verified knowledge in the cognitive context Mq to avoid redundant exploration. For example, if the heuristic memory Hπ(q) contains a preprocessing process that "the format validity of the environmental region state must be verified before performing regionalized task scheduling", the data processing model can directly incorporate this preprocessing process.

[0029] In this embodiment, if the data processing model discovers that key information in the cognitive context is missing during the planning process, it will automatically insert an "exploratory process" to supplement the necessary information before continuing to run.

[0030] In some embodiments, the memory configuration cycle includes an initial cognitive configuration node; prior to step S110, the method further includes the following steps S210 and S220: Step S210: In the cognitive configuration node, extract the environmental parameters of the current environment where the first user is located, and generate an initial cognitive memory entry.

[0031] In this embodiment, before the data processing model processes any query request, the cognitive memory initialization of the target task domain environment D is completed through the Task DomainEnvironment Inception Agent, generating the initial cognitive memory CD(0), which provides basic global task domain environment knowledge for all subsequent queries.

[0032] In this embodiment, the data processing model generates a structured initial cognitive entry C0 by parsing the task domain environment directory structure, file format, pattern statistics, inter-table relationships, and task manual.

[0033] Step S220: Set the model processing configuration of the data processing model according to the preset constraints and the initial cognitive memory entries.

[0034] In this embodiment, the initial cognitive entry C0 of the data processing model is based on a single insight and is stored incrementally. This initial cognitive entry C0 may include two core interactive modules: the target task domain environment includes an action execution interface and an environment state perception interface. The task_id identifier of the environment state perception interface serves as a unique index.

[0035] In this embodiment, the data processing model can be configured with technical constraints, which may include: not making unfounded semantic inferences; and explicitly labeling uncertain relationships.

[0036] Based on this, step S110 may include the following step S230: Step S230: In response to the current query request output by the first user, the data processing model under the model processing configuration is invoked to extract the initial cognitive memory that matches the current query request as the target memory feature set.

[0037] In this embodiment, after the data processing model is initialized, it can quickly process query requests.

[0038] In some embodiments, the data processing model generates a historical working memory sequence based on historical query requests output by a second user; the memory configuration period includes a fusion cognitive configuration node; after step S220, the method further includes the following steps S310 and S320: Step S310: Based on the historical query request and the historical working memory sequence, construct an update sequence for the specified memory type.

[0039] In this embodiment, the memory type can include experiential memory consolidation type, heuristic memory consolidation type, and cognitive memory consolidation type.

[0040] Step S320: Through the update sequence, first update the generation model of the data processing model, then update the merging model of the data processing model, to obtain the updated data processing model.

[0041] In this embodiment, the data processing model's generation model includes an experience generation model corresponding to the experience memory consolidation type, a heuristic generation model corresponding to the heuristic memory consolidation type, and a cognitive generation model corresponding to the cognitive memory consolidation type. The data processing model's merging model includes an experience merging model corresponding to the experience memory consolidation type, a heuristic merging model corresponding to the heuristic memory consolidation type, and a cognitive merging model corresponding to the cognitive memory consolidation type.

[0042] In this embodiment, by setting up a generation model and a merging model, the reasoning ability of the subsequent data detection model can be effectively improved, and the accuracy of the current working memory sequence output by the data detection model can be improved.

[0043] In some embodiments, the memory type is defined as an experience memory consolidation type; step S310 may include the following step S410: Step S410: Based on the historical query request and the historical working memory sequence, extract the core experience information of the data processing model and use it as an update sequence; wherein, the core experience information includes problem description, execution status, inference result, key insight chain, execution result and detection label.

[0044] In this embodiment, Experience memory is the finest-grained memory layer. It uses a single historical query request q as an index to record the core experience information of a single reasoning session. The expression for this core experience information is as follows:

[0045] in, This indicates a problem description (e.g., "Calculate the planning time for path nodes 1234 of a Robot-X robot with task_type=S"). Indicates the execution status (success / fail, marked by an external signal in supervised mode, and success by default in unsupervised mode). Indicates the result of reasoning. This indicates a key insight chain (such as calculating planning time by extracting rules from the environmental rule base, rather than by statistically analyzing historical data from historical execution logs). This represents the execution result that characterizes the core execution product. This indicates the detection label (such as "task_type=S, hypothetical query").

[0046] Experience generation model Extracting core information from working memory sequences to generate candidate experiential memories If the query request has no existing candidate experience memory. Then directly insert the candidate experience memory. If existing candidate experience memories exist. Then, the empirical merging model Integration complete.

[0047] In some embodiments, the memory type is defined as a heuristic memory consolidation type; step S310 may include the following step S510: Step S510: Based on the historical query request and the historical working memory sequence, extract the heuristic core information of the data processing model and use it as an update sequence; wherein, the heuristic core information includes cluster description, cluster-level chain results, general insight chain, applicable conditions, pending hypotheses and cluster detection labels.

[0048] In this embodiment, heuristic memory is a mid-level memory at the clustering level. Using query cluster π as an index, it extracts general problem-solving strategies between semantically similar queries to obtain heuristic core information. The expression of this heuristic core information is as follows:

[0049] in, Indicates cluster description, This represents the cluster-level chain result that summarizes the cluster-level process. Indicates the general insight chain, Indicates the applicable conditions. Indicates an unresolved hypothesis. The cluster detection label is represented by π(q). The semantic embedding of the query request q is calculated using a pre-trained text embedding model (such as Qwen's text-embedding-v4). A similarity threshold (such as 0.90 / 0.95) and a minimum cluster size (top_k≥4) are set, and the query request q is assigned to the corresponding query cluster π(q).

[0050] Heuristic generative models A general strategy is extracted from the working memory sequence and the intra-cluster nearest neighbor query N(q) to generate candidate heuristic memories. Heuristic merging of intelligent agents Complete the fusion and perform deduplication: remove duplicate policy entries; stabilization: promote policies that appear ≥3 times to "default recommended policies"; conditionalization: clarify the applicable scenarios for divergent policies.

[0051] In some embodiments, the memory type is defined as a cognitive memory consolidation type; step S310 may include the following step S610: Step S610: Based on the historical query request and the historical working memory sequence, extract the cognitive core information of the data processing model and use it as an update sequence; wherein, the cognitive core information includes the task domain environment insight chain.

[0052] In this embodiment, cognitive memory (Cognition) is a high-level memory at the task domain environment level. It captures globally stable knowledge for the target task domain ring D to obtain cognitive core information, the expression of which is as follows:

[0053] Each of them This forms an independent task domain environment-level insight, contributing to the formation of a task domain environment insight chain.

[0054] Cognitive Generative Intelligent Agent Extract cross-task general and well-supported task-domain contextual insights from working memory sequences to generate candidate cognitive memories. Cognitive merging model The update was completed using a conservative update strategy: only insights that are not covered by existing knowledge and are applicable across tasks were added; the update method was append-only, and existing entries were not deleted; the update frequency was strictly limited, with a maximum of one new knowledge entry added per query request to avoid memory expansion.

[0055] In this embodiment, the historical working memory sequence is transformed into a three-layered persistent memory consisting of "experience-heuristic-cognition." Each layer is completed collaboratively by a "Generator" and a "Consolidator," enabling the retrieval, insertion, and fusion of new memories. Explicit conflict detection ensures the consistency of the memories. The consolidation process of the three layers of memory progresses progressively, abstracting from single-query experience to query cluster heuristics, and then refining it into task domain environment cognition.

[0056] In some embodiments, prior to step S320, the method further includes the following step S710: Step S710: Using the verification model in the data processing model, filter the target update sequences that meet the set verification conditions in the update sequences.

[0057] Based on this, step S320 may include the following step S720: Step S720: Using the target update sequence, first update the generating model of the data processing model, then update the merging model of the data processing model, to obtain the updated data processing model.

[0058] In this embodiment, the verification agent verifies the update sequence after each process is completed. The effectiveness assessment is performed using the following expression:

[0059] The evaluation dimensions, or verification conditions, for this verification agent can be as follows: The verification phase evaluates the agent across four dimensions: logical consistency, memory compatibility, and task relevance. Specifically: Execution legality checks if the agent's actions are legal and error-free; logical consistency verifies if intermediate results conform to common sense; memory compatibility checks if the current result conflicts with existing memories; and task relevance ensures the execution result is relevant to the query target. Conflict and failure handling: If the evaluation is a failure (e.g., logical error), bounded retries (maximum 3 by default) are triggered under the guidance of the Reflection Tool, prioritizing the reuse of corrective solutions from historical experience during retries; if the evaluation is a conflict (e.g., contradictory assumptions, inconsistent intermediate results), the current plan is terminated, and the cause of the conflict and suggested solutions are output; only working memories evaluated as success will proceed to the subsequent memory consolidation phase.

[0060] Step S120: Invoke the preset data detection model, using the target configuration node and the memory feature as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

[0061] In this embodiment, the data detection model can be an existing general network model or an optimized network model; no limitation is made here. Specifically, the data detection model can be as follows: Figure 2 The Execute Agent shown.

[0062] The data detection model executes step-by-step based on the target memory feature set, dynamically updating the structured working memory to obtain the current working memory sequence. The specific expression is as follows: Et:(st,wt 1)→wt Among them, wt is a structured object containing "intermediate results, inference logic, and status flags", which is constrained by three major control tools: Planning Tool: records the local goals and dependencies of the current process; Thinking Tool: extracts key inferences in the analysis; Reflection Tool: records exceptions in the execution and correction plans.

[0063] This data inspection model is instantiated through structured prompts and supports the invocation of a general task execution toolset (environment interaction, action scheduling, status verification, result generation, etc.). The model can check whether a valid result already exists in the historical experience Eq; if so, it is directly reused to avoid duplicate execution. For example, if the historical experience Eq already stores the calculation result for "task execution status of region A" and the data has not been updated, that result is directly invoked, skipping redundant calculations.

[0064] In this embodiment, the "agent context reasoning loop" and "hierarchical memory consolidation" are deeply coupled through the inference-time self-evolution scheduling module, realizing the inference-time self-evolution of the agent in the agent task domain. This requires no model parameter updates or external supervision signals, and continuously improves reasoning capabilities solely by processing consecutive query sequences. The specific process is as follows: During the initialization phase, the task domain environment initial data processing model is called to generate the initial cognitive memory CD(0), while the experience memory E and heuristic memory H are initialized to empty.

[0065] During the query sequence processing phase, for each query in the target task domain environment D, each query request q is processed round by round: a. Memory retrieval: Retrieve the current cognitive context Mq=(Eq, Hπ(q), CD), where Eq is the experiential memory of query q (empty if none), Hπ(q) is the heuristic memory of query cluster π(q) (empty if none), and CD is the cognitive memory of the current task domain environment.

[0066] b. Inference Loop: The data detection model context inference loop generates a validated working memory sequence Wtt=1T to complete the solution of the current query request.

[0067] c. Memory consolidation: The generation and merging models of experience, heuristics, and cognitive memory are called in sequence to update the memory layer corresponding to the data processing model.

[0068] d. Ability evolution: As the number of query requests processed increases, the cognitive context Mq is continuously enriched. Early reasoning (such as the first 10 queries) mainly relies on the initial cognitive memory CD(0); mid-term reasoning (such as 10–50 queries) begins to reuse accumulated experience memory; late-term reasoning (such as after 50 queries) is dominated by heuristic memory and cognitive memory, achieving the self-evolutionary effect of reducing repetitive reasoning, decreasing error rate, and improving reasoning efficiency.

[0069] <Equipment Example 1> Figure 3 This is a schematic diagram of a query processing device based on a self-evolving memory intelligent agent according to one embodiment. Figure 3 As shown, the query processing device 300 based on self-evolving memory intelligent agents may include: The response module 310 is used to respond to the current query request output by the first user, call the data processing model under the target configuration node in the set memory configuration period, and extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node; The calling module 320 is used to call a preset data detection model, using the target configuration node and the memory features as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

[0070] In some embodiments, the device further includes a generation module, configured to extract environmental parameters of the current environment of the first user in the cognitive configuration node, generate an initial cognitive memory entry, and set the model processing configuration of the data processing model according to preset constraints and the initial cognitive memory entry; The response module 310 is also used to respond to the current query request output by the first user, call the data processing model under the model processing configuration, and extract the initial cognitive memory that matches the current query request as the target memory feature set.

[0071] In some embodiments, the apparatus further includes an update module for constructing an update sequence of a specified memory type based on the historical query request and the historical working memory sequence; using the update sequence, first updating the generation model of the data processing model, and then updating the merging model of the data processing model to obtain the updated data processing model.

[0072] In some embodiments, the update module is further configured to extract the core experience information of the data processing model based on the historical query request and the historical working memory sequence, and use it as an update sequence; wherein the core experience information includes problem description, execution status, inference result, key insight chain, execution result and detection label.

[0073] In some embodiments, the update module is further configured to extract heuristic core information of the data processing model based on the historical query request and the historical working memory sequence, and use it as an update sequence; wherein the heuristic core information includes cluster description, cluster-level chain results, general insight chain, applicable conditions, pending hypotheses and cluster detection labels.

[0074] In some embodiments, the update module is further configured to extract the cognitive core information of the data processing model based on the historical query request and the historical working memory sequence, and use it as an update sequence; wherein the cognitive core information includes a task domain environment insight chain.

[0075] In some embodiments, the apparatus further includes a verification module for filtering target update sequences in the update sequence that meet set verification conditions through a verification model in the data processing model; The update module is also used to update the generating model of the data processing model first, and then update the merging model of the data processing model, through the target update sequence, to obtain the updated data processing model.

[0076] <Equipment Example 2> Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to another embodiment.

[0077] like Figure 4 As shown, the electronic device 400 includes a processor 410 and a memory 420, the memory 420 for storing an executable computer program, and the processor 410 for executing methods as described in any of the above method embodiments under the control of the computer program.

[0078] The modules of the self-evolving memory intelligent agent query processing device 300 described above can be implemented by the processor 410 in this embodiment executing the computer program stored in the memory 420, or they can be implemented by other structures, which are not limited here.

[0079] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for causing a processor to implement any of the methods in the foregoing embodiments of this disclosure. A computer-readable storage medium may be a tangible device capable of holding and storing instructions used by an instruction execution device. For example, a computer-readable storage medium may include an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), compact disc-read-only memory (CD-ROM), digital versatile disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any combination thereof. Computer-readable storage media as used herein is not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires. The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include one or more of copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to computer-readable storage media in the respective computing / processing device.The computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source or object programs written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Smalltalk, C++, etc.) and conventional procedural programming languages ​​(such as the "C" language or similar programming languages). The computer-readable program instructions may execute entirely on a first user's computer, partially on a first user's computer, as a standalone software package, partially on a first user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the first user's computer via any type of network (e.g., a local area network or a wide area network), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays, or programmable logic arrays, is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the embodiments of this disclosure. Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions. These computer-readable program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. It should be noted that implementation in hardware, implementation in software, and implementation using a combination of software and hardware are all equivalent. Various embodiments of the present disclosure have been described above; the above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of this disclosure is defined by the appended claims.

Claims

1. A query processing method based on a self-evolving memory intelligent agent, characterized in that, The method includes: In response to the current query request output by the first user, the data processing model under the target configuration node in the set memory configuration period is invoked to extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node; A preset data detection model is invoked, using the target configuration node and the memory features as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

2. The method according to claim 1, characterized in that, The memory configuration cycle includes an initial cognitive configuration node; before responding to a query request output by the first user, invoking a preset data processing model, and extracting memory features matching the query request, the method further includes: In the cognitive configuration node, environmental parameters of the current environment where the first user is located are extracted to generate initial cognitive memory entries; Based on the preset constraints and the initial cognitive memory entries, set the model processing configuration of the data processing model; In response to the current query request output by the first user, the data processing model under the target configuration node in the set memory configuration period is invoked to extract the target memory feature set matching the current query request, including: In response to the current query request output by the first user, the data processing model under the model processing configuration is invoked to extract the initial cognitive memory that matches the current query request as the target memory feature set.

3. The method according to claim 1, characterized in that, The data processing model generates a historical working memory sequence based on historical query requests output by the second user; the memory configuration period includes a fusion cognitive configuration node; after setting the model processing configuration of the data processing model according to preset constraints and the initial cognitive memory entries, the method further includes: Based on the historical query requests and the historical working memory sequence, an update sequence for a given memory type is constructed; The update sequence is used to first update the generating model of the data processing model, and then update the merging model of the data processing model to obtain the updated data processing model.

4. The method according to claim 3, characterized in that, The defined memory type includes an experience memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the core experience information of the data processing model is extracted and used as an update sequence; wherein, the core experience information includes problem description, execution status, inference results, key insight chains, execution results and detection tags.

5. The method according to claim 3, characterized in that, The defined memory type includes heuristic memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the heuristic core information of the data processing model is extracted and used as the update sequence; wherein, the heuristic core information includes cluster description, cluster-level chain results, general insight chain, applicable conditions, pending hypotheses and cluster detection labels.

6. The method according to claim 3, characterized in that, The defined memory type includes cognitive memory consolidation type; the step of constructing an update sequence for the defined memory type based on the historical query request and the historical working memory sequence includes: Based on the historical query requests and the historical working memory sequence, the cognitive core information of the data processing model is extracted and used as an update sequence; wherein, the cognitive core information includes a task domain environment insight chain.

7. The method according to claim 3, characterized in that, Before updating the generator model of the data processing model and then updating the merged model of the data processing model through the update sequence to obtain the updated data processing model, the method further includes: The data processing model uses a validation model to filter target update sequences that meet the set validation conditions from the update sequences. The step of updating the data processing model by first updating the generator model and then updating the merged model to obtain the updated data processing model includes: The target update sequence is used to first update the generating model of the data processing model, and then update the merging model of the data processing model to obtain the updated data processing model.

8. A query processing device based on a self-evolving memory intelligent agent, characterized in that, The device includes: The response module is used to respond to the current query request output by the first user, call the data processing model under the target configuration node in the set memory configuration period, and extract the target memory feature set that matches the current query request; wherein, the feature elements of the target memory feature set are associated with the target configuration node; The calling module is used to call a preset data detection model, using the target configuration node and the memory features as input conditions, to obtain the current working memory sequence; wherein, the current working memory sequence is used to characterize the solution result of the current query request.

9. An electronic device, characterized in that, The system includes a memory and a processor, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images