Autonomous construction and optimization method of AI agent for voice robot

Abstract

The application discloses an AI Agent autonomous construction and optimization method for a voice robot, relates to the technical fields of artificial intelligence, large language model application and intelligent voice robot, and solves the problems of high artificial participation, long development cycle, fragmented links, slow iteration and poor cross-industry adaptability in the prior art, realizes autonomous construction and continuous optimization of the whole process of the intelligent voice robot, improves product quality and iteration efficiency, and is suitable for customer service, office, industrial scheduling, smart home and other scenes.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to the fields of large language model applications, AI agents, autonomous development and full lifecycle management of intelligent voice robots; specifically, it relates to a method for autonomous construction and optimization of AI agents for voice robots. Background Technology

[0002] With the rapid development of Large Language Models (LLM) and AI Agent technologies, building intelligent voice robots based on low-code development platforms such as Coze, Dify, and Flowise has become mainstream in the industry. These platforms provide basic functions such as prompt word editing, plugin selection, workflow building, knowledge base uploading, and online testing, which lowers the development threshold to a certain extent and enables intelligent voice robots to be widely used in customer service, office automation, smart home control, industrial equipment scheduling, and financial transactions.

[0003] However, existing AI Agent development platforms still have the following significant shortcomings when building intelligent voice robots:

[0004] High reliance on manual labor and low development efficiency: Core processes such as prompt writing, workflow design, tool configuration, and requirement analysis require a large amount of manual participation, resulting in long development cycles, high labor costs, and difficulty in meeting the needs of large-scale and rapid delivery.

[0005] The development process is fragmented and has a low degree of automation: development, testing, quality inspection, and optimization are independent of each other, requiring manual switching of tools, data organization, and parameter adjustment. Optimization strategies rely on the experience of developers, test coverage is insufficient, and product quality is unstable.

[0006] Delayed response to demand and high iteration costs: When customer requirements change, the entire process needs to be manually re-examined, configurations adjusted, and retested, resulting in long iteration cycles and slow market response.

[0007] Weak cross-industry adaptability: Different industries have different business logic, terminology systems, and interaction processes, making it difficult for existing platforms to quickly generate industry-specific intelligent voice robots, thus failing to balance universality and customization.

[0008] Currently, industry optimization efforts are mostly focused on upgrading single components or simplifying user interfaces, failing to develop a full lifecycle automation solution covering requirement analysis, development and construction, testing and quality inspection, optimization and iteration, and requirement changes. This approach cannot fundamentally solve the aforementioned shortcomings. Summary of the Invention

[0009] Therefore, the purpose of this invention is to propose a method and system for the autonomous construction and optimization of AI agents for voice robots. It aims to develop an intelligent voice robot construction method capable of automatic multimodal requirement parsing, multi-agent collaborative autonomous development, fully automated closed-loop iteration, and rapid response to requirement changes. This overcomes the shortcomings of high human involvement, time-consuming and labor-intensive prompt word writing and workflow setup, long development cycles, and high costs. It also addresses the problems of independent development, testing, quality inspection, and optimization stages, manual testing adjustments, reliance on experience for optimization, and insufficient test coverage, achieving fully automated integration. Furthermore, it eliminates the drawbacks of requiring manual adjustments to the entire process due to requirement changes, long iteration cycles, and slow responses, enabling rapid iteration. Finally, it overcomes technical bottlenecks such as difficulties in cross-industry adaptation and the inability to quickly generate industry-specific intelligent voice robots.

[0010] This invention provides a method for autonomous construction and optimization of AI agents for voice robots, including the following steps:

[0011] Step S1: Receive and preprocess multimodal raw business materials, perform modal normalization on multimodal data such as voice, text, images, and video, and output standardized unstructured business data;

[0012] Specifically, multimodal data refers to data sources that include various formats such as voice, text, images, and video. Modality normalization refers to the process of uniformly processing and standardizing multimodal data of different formats to form unstructured data that can be uniformly parsed.

[0013] Step S2: Perform intelligent parsing of unstructured business data to extract core business elements, complete conflict detection and standardization mapping, and output structured business requirement data;

[0014] Specifically, structured business requirement data refers to business requirements presented in a standardized and normalized format, which includes explicit information such as requirement objectives, functional requirements, and interaction rules.

[0015] Step S3: Based on the AI ​​Agent intelligent development and operation platform, automatically generate multi-Agent framework, interaction protocol, business workflow, and large model prompt words according to structured business needs, complete tool selection and customized plugin development, and build the initial version of the intelligent voice robot;

[0016] Specifically, an AI Agent (Artificial Intelligence Agent) refers to an artificial intelligence entity that can perceive its environment, autonomously perform tasks, and make decisions. A multi-Agent framework refers to a system architecture composed of multiple AI Agents with mutually cooperating functions, where each Agent completes joint tasks through pre-defined interaction protocols.

[0017] Step S4: Start the simulated dialogue agent to conduct automated interaction testing. The service quality and satisfaction inspection agent completes the quality assessment. If the standard is not met, the quality improvement agent generates optimization strategies and drives iteration. If the standard is met, a version that can be launched is output.

[0018] Step S5: The Agent receives and parses requirement changes through dynamic strategy iteration, automatically reconstructs the system architecture and execution process, and completes rapid iteration.

[0019] Furthermore, the modal regulation described in step S1 includes:

[0020] Perform noise reduction, voice enhancement, ASR transcription, and spoken language normalization on the speech data, and output sequential plain text;

[0021] Perform encoding cleaning, language normalization, and redundancy filtering on text data, and output unformatted standard text.

[0022] Perform tilt correction, OCR recognition, and visual semantic description on image data, and output business-related text content;

[0023] Perform audio-visual separation, keyframe extraction, separate normalization and temporal alignment on video data, and output coherent business content;

[0024] All modalities are ultimately mapped to plain text, denoised, and time-series unstructured data.

[0025] Furthermore, the intelligent parsing of business requirements in step S2 includes:

[0026] Unstructured data pre-cleaning and normalization, deep understanding of business semantics and intent recognition, extraction of core elements of business requirements, structured mapping and standardization of requirement elements, requirement conflict detection and intelligent verification, structured requirement output and iterative optimization.

[0027] The core elements include the business entity, core objectives, functional requirements, constraints, quantitative indicators, and related resources.

[0028] Furthermore, the AI ​​Agent intelligent development and operation platform mentioned in step S3 includes:

[0029] Task analysis and decomposition agent, workflow generation agent, prompt word generation agent, tool selection agent, programming agent, simulated dialogue agent, quality improvement agent, dynamic strategy iteration agent, service quality and satisfaction inspection agent.

[0030] Furthermore, the task analysis and decomposition are performed by the Agent:

[0031] Extract the core elements of the structured requirements, break them down into a hierarchy of total requirements, first-level sub-tasks, and second-level sub-tasks, match them with corresponding Agents, formulate a multi-Agent framework and interaction protocol, and clarify the calling order, data format, constraint verification, and exception handling mechanism.

[0032] Furthermore, the prompt word generation Agent executes:

[0033] Based on the type of business requirement, core tasks, constraints, and characteristics of the target model, it automatically generates dedicated prompts containing task instructions, constraints, and output specifications, adapting to the semantic understanding and execution boundaries of large models.

[0034] Furthermore, the tool selects an Agent to execute:

[0035] Based on the principles of functional adaptation, performance indicators, technology stack compatibility, and optimal cost-effectiveness, matching tools are selected from the tool library and sorted and output according to historical adaptation effect, resource consumption, and deployment cost.

[0036] The toolkit includes speech processing tools, text processing tools, multimodal processing tools, large model inference tools, and system optimization tools.

[0037] Furthermore, the programming agent executes:

[0038] It automatically completes plugin development, debugging, and deployment based on customized needs, supports Python, Docker, and RESTful interfaces, and outputs functional plugins that can be directly integrated, including plugins for terminology correction, implicit requirement completion, time sequence alignment, classification and clustering, and latency optimization.

[0039] Furthermore, the failure to meet the standards mentioned in step S4 is addressed by the quality improvement agent generating optimization strategies and driving iterations, including: simulating dialogue to generate test data, the quality inspection agent outputting defect reports, the quality improvement agent generating executable optimization strategies, feeding back to the task analysis and decomposition agent to reconstruct the architecture, regenerating the version and retesting, until the quality standards are met;

[0040] When the iteration fails to converge, a manual review and intervention mechanism is triggered.

[0041] Specifically, iterative non-convergence refers to a situation where, during the automated iterative optimization process, the quality inspection results consistently fail to meet the preset quality requirements, and there is no obvious trend of improvement.

[0042] Furthermore, the requirement change iteration described in step S5 includes:

[0043] The dynamic strategy iteration agent parses, organizes, and structures the changed content, and outputs standardized change requirements;

[0044] Task analysis and breakdown: The agent adjusts its architecture, workflow, and tool configuration based on changing requirements;

[0045] It automatically triggers a closed loop of development, testing, quality inspection, and optimization to complete the version update.

[0046] This invention also provides a system for the autonomous construction and optimization of AI agents for voice robots, used to execute the above-described method for the autonomous construction and optimization of AI agents for voice robots, including:

[0047] The multimodal data processing and intelligent parsing layer for business requirements is used to receive and organize multimodal business materials and output structured business requirements.

[0048] Preferably, the multimodal data processing and intelligent parsing layer for business requirements can be adapted to unstructured business materials in different industries. Various agent components can automatically generate customized multi-agent architectures, workflows, and tool libraries according to industry needs, enabling the rapid development of industry-specific intelligent voice robot agents and improving the system's versatility and adaptability.

[0049] The AI ​​Agent intelligent development and operation platform layer is used to automatically generate multi-Agent frameworks, workflows, prompts, and tool libraries to build the initial version of the intelligent voice robot;

[0050] An AI Agent full lifecycle automated management layer for testing, quality inspection, optimization iteration, and response to requirement changes.

[0051] The development, operation, and iteration of intelligent voice robots can be achieved through a unified AI Agent full lifecycle automated management system, eliminating the need to build independent management systems for different stages, thus saving computing resources and subsequent operation and maintenance costs.

[0052] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the above-described method for the autonomous construction and optimization of an AI Agent for voice robots.

[0053] The present invention also provides a computer device, the computer device including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the steps of the above-described method for autonomous construction and optimization of AI Agent for voice robots.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] This invention provides a method and system for autonomous construction and optimization of AI agents for voice robots. Through automatic multimodal requirement analysis and automated development of various agent components, it replaces manual prompt writing, workflow setup, and tool selection, significantly reducing human intervention, effectively lowering labor costs, and significantly shortening the development cycle of intelligent voice robots. It constructs an automated process for development, testing, quality inspection, and optimization, with each stage linked and connected through agent components. No manual intervention is required for testing and parameter adjustments; test coverage is guaranteed by the algorithm model, and optimization strategies are automatically generated by the quality improvement agent, improving the product quality and standardization of intelligent voice robots. Through dynamic strategy iteration agents, it enables rapid analysis and resolution of changing customer requirements. Combined with full-process automated development capabilities, it eliminates the need for manual readjustment of the entire process, significantly shortening the system iteration cycle and enabling rapid response to changes in customer needs, thus improving market responsiveness. It is applicable to scenarios involving autonomous construction, automated development, testing and quality inspection, iterative optimization, and cross-industry adaptation of AI agents for intelligent voice robots. Attached Figure Description

[0056] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0057] In the attached diagram:

[0058] Figure 1 This is a schematic diagram of the components of the AI ​​Agent autonomous construction and optimization system for voice robots according to an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the architecture of the multimodal data processing and intelligent parsing layer for business requirements in an embodiment of the present invention;

[0060] Figure 3 This is a schematic diagram of the component composition of the AI ​​Agent intelligent development and operation platform layer according to an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of the full lifecycle automated iteration process of the AI ​​Agent full lifecycle automated management layer in an embodiment of the present invention;

[0062] Figure 5 This is a flowchart illustrating the steps of the method for autonomous construction and optimization of AI Agent for voice robots according to an embodiment of the present invention.

[0063] Figure 6 This is a schematic diagram of the computer device configuration according to an embodiment of the present invention. Detailed Implementation

[0064] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of systems and products consistent with some aspects of this disclosure as detailed in the appended claims.

[0065] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0066] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0067] The embodiments of the present invention will be described in further detail below.

[0068] Example

[0069] This invention provides a method for autonomous construction and optimization of AI agents for voice robots, such as... Figure 5 As shown, it includes the following steps:

[0070] Step 1: Multimodal data processing and intelligent analysis of business requirements, including:

[0071] Modality shaping: Receives raw multimodal business materials (voice, text, images, video), and transforms them into plain text, denoised, and time-series unstructured business data through the modality shaping module; including:

[0072] Speech normalization: noise reduction, dereverberation, voice enhancement, ASR transcription (converting speech data into text data), speaker separation, and colloquialization correction;

[0073] Text normalization: UTF-8 encoding cleaning, traditional Chinese to simplified Chinese conversion, terminology normalization, and redundancy filtering;

[0074] Image standardization: correction, deblurring, OCR recognition, visual semantic description, and business information extraction;

[0075] Video smoothing: audio and video separation, keyframe extraction, separate smoothing of audio and video, and time-sequence alignment and fusion;

[0076] Cross-modal unification: semantic space mapping, adaptive noise filtering, and temporal and structural standardization.

[0077] Perform NLP processing (processing and analyzing human natural language) and requirements parsing on unstructured business data to output standardized structured requirements, including:

[0078] Pre-cleaning and normalization of unstructured data;

[0079] Deep understanding of business semantics and intent recognition;

[0080] Business entity, core objectives, functional requirements, constraints, quantitative indicators, and extraction of related resources;

[0081] Demand elements are mapped in a structured manner using JSON / XML templates;

[0082] Intelligent verification of logical conflicts, integrity, and compliance;

[0083] Structured requirements output and model iterative optimization.

[0084] Specifically, the parsing adopts a fusion scheme of large model, NLP, and knowledge graph, supporting zero / few-shot learning, rule fallback, graph reasoning, fine-grained extraction, clustering noise reduction, multimodal enhancement, and self-supervised optimization.

[0085] Step Two: Building the AI ​​Agent Intelligent Development and Operation Platform and Developing Functional AI Agents;

[0086] Step 3: The AI ​​Agent automatically generates an initial version of the intelligent voice robot, including:

[0087] Based on the AI ​​Agent intelligent development and operation platform, version building is completed through multi-agent collaboration:

[0088] Task analysis and agent decomposition: generating multi-agent frameworks and interaction protocols;

[0089] Workflow Generation Agent: Automatically builds business execution processes;

[0090] Prompt word generation Agent: Generates specialized prompt words adapted to large models;

[0091] Agent selection: Filter tools by functionality, performance, compatibility, and cost-effectiveness;

[0092] Programming Agent: Automatically develops and deploys customized plugins;

[0093] Integrate the framework, process, prompts, and toolkit to generate the initial version of the intelligent voice robot.

[0094] Step 4: Full lifecycle automated testing, quality inspection, and iterative optimization, including:

[0095] Simulates dialogue agents to simulate various types of user interactions and generate test data;

[0096] Service quality and satisfaction assessment agent automatically evaluates response accuracy, latency, satisfaction, and compliance;

[0097] If the standards are not met, quality improvement strategies will be implemented for Agent generation, including terminology correction, requirement completion, time sequence alignment, performance optimization, and format standardization.

[0098] The strategy is fed back to the architecture layer for reconstruction and retesting;

[0099] Iteration failure to converge triggers manual intervention.

[0100] Step 5: Dynamic and rapid iteration of changing customer needs, including:

[0101] The dynamic strategy iteration agent receives change information and completes parsing, normalization, and structuring.

[0102] Task analysis and breakdown: Agent architecture, workflow, and tool configuration adjustments;

[0103] It automatically executes a closed loop of development, testing, quality inspection, and optimization to complete rapid updates and deployments.

[0104] This invention also provides a system for the autonomous construction and optimization of AI agents for voice robots, such as... Figure 1 As shown, it includes a multimodal data processing and business requirement intelligent analysis layer, an AI Agent intelligent development and operation platform layer, and an AI Agent full lifecycle automated management layer, which are used to realize the full-process execution of the above-mentioned method for the autonomous construction and optimization of AI Agent for voice robots.

[0105] Among them, the multimodal data processing and business requirement intelligent parsing layer is used to receive and organize multimodal business materials and output structured business requirements;

[0106] like Figure 2 As shown, the multimodal data processing and intelligent business requirement parsing layer includes a multimodal data processing module and a business requirement intelligent parsing module;

[0107] The multimodal data processing module includes:

[0108] Speech preprocessing unit: used for noise reduction, de-reverberation, human voice enhancement, silence removal, and uniform sampling rate;

[0109] ASR transcription unit: supports multi-speaker separation and timestamp marking;

[0110] Spoken language standardization unit: used to remove modal particles, correct inversions, and complete omissions;

[0111] Text cleaning unit: used for encoding normalization, format removal, and terminology standardization;

[0112] Image processing unit: used for correction, deblurring, OCR, and visual description generation;

[0113] Video processing unit: used for audio-visual separation, keyframe extraction, temporal alignment, and semantic fusion;

[0114] Cross-modal fusion unit: used for semantic space mapping, information complementation, and noise filtering.

[0115] The business requirement intelligent analysis module includes:

[0116] Text preprocessing unit: used for cleaning, error correction, sentence segmentation, and normalization;

[0117] Intent recognition unit: classifies intent and locates scenarios based on large models and knowledge graphs;

[0118] Feature extraction unit: used for NER named entity recognition, relation extraction, and keyword extraction;

[0119] Structured mapping units: used for template matching, field filling, and format standardization;

[0120] Conflict detection unit: used for logic verification, integrity verification, and compliance verification;

[0121] Iterative optimization unit: used for evaluation metrics, human feedback, and model fine-tuning.

[0122] The AI ​​Agent intelligent development and operation platform layer is used to automatically generate multi-Agent frameworks, workflows, prompts, and tool libraries to build the initial version of the intelligent voice robot;

[0123] Specifically, the AI ​​Agent intelligent development and operation platform layer includes:

[0124] Task Analysis and Agent Decomposition: Input structured requirements, output multi-agent framework, sub-task list, and interaction protocol. The decomposition principles are: requirement matching, function adaptation, process coherence, clear division of labor, and interactivity.

[0125] Workflow Generation Agent: Automatically constructs nodes, branches, judgment conditions, and execution order based on business logic, and supports visualization and JSON export;

[0126] Prompt word generation agent: Automatically generates prompt words based on task type, industry domain, model characteristics, and constraint indicators. The structure includes: task instructions, constraints, execution boundaries, output format, and quality requirements.

[0127] Agent selection: The tool library is categorized into: speech processing, text processing, multimodal processing, large model inference, and system optimization; the selection criteria are: prioritized by functional coverage, performance compliance, technical compatibility, and cost-effectiveness.

[0128] Programming Agent: The process is: requirement breakdown → technology selection → code generation → debugging → deployment; Supported language: Python; Deployment environment: Docker, RESTful interface; Output: Plugins can be directly integrated, including calling documentation, parameter descriptions, and logging modules;

[0129] Simulated Dialogue Agent: Simulates user types: new users, familiar users, questioning users, professional users, and users with abnormal input; Generates dialogues: Covering normal processes, abnormal branches, boundary scenarios, and stress scenarios;

[0130] Service Quality and Satisfaction Inspection Agent: Evaluation dimensions include: response accuracy, semantic understanding rate, response latency, terminology accuracy, process completion rate, satisfaction, and compliance;

[0131] Quality Improvement Agent: Optimization types include: speech normalization optimization, requirement parsing optimization, timing alignment optimization, performance latency optimization, and format standardization optimization; Output: executable, quantifiable, and verifiable strategies.

[0132] Dynamic Strategy Iteration Agent: Handles requirement changes: parsing, standardization, structuring, conflict detection, and architecture update triggering.

[0133] Figure 3 The components of the AI ​​Agent intelligent development and operation platform layer in this embodiment are shown.

[0134] The AI ​​Agent full lifecycle automation management layer is used for testing, quality inspection, optimization iteration, and response to requirement changes. Figure 4 This embodiment illustrates the full lifecycle automated iteration process of the AI ​​Agent full lifecycle automated management layer.

[0135] Specifically, the AI ​​Agent full lifecycle automated management layer includes:

[0136] Test Management Unit: Used for automated test case generation, execution, and data logging;

[0137] Quality Inspection and Assessment Unit: Used for multi-dimensional scoring, defect location, and report generation;

[0138] Iterative control unit: used for policy distribution, architecture refactoring, version retesting, and convergence determination;

[0139] Manual intervention unit: used to trigger verification, correction and intervention when iteration fails to converge.

[0140] Application examples

[0141] The present invention can be applied to a variety of practical scenarios. The following are specific application examples of the present invention in two different scenarios:

[0142] Example 1: Self-construction of an intelligent voice customer service robot for the power industry, including the following steps:

[0143] Step S1: Multimodal data normalization, the processing procedure is as follows;

[0144] Inputs: Audio recordings of power business requirements, business documents, process screenshots, and training videos.

[0145] Speech processing: noise reduction, ASR transcription, removal of interjections, and correction of spoken language;

[0146] Text processing: Cleaning up formats and standardizing terminology (switchgear, zero-sequence protection, load current);

[0147] Image processing: OCR extracts process text and generates visual descriptions;

[0148] Video processing: audio-visual separation, keyframe extraction, and timing alignment;

[0149] Output: Standardized unstructured business text.

[0150] Step S2: Intelligent analysis of business requirements, the processing procedure is as follows;

[0151] Extracting elements:

[0152] Entities involved: Electricity users, customer service robots;

[0153] Core objectives: fault reporting, power consumption inquiry, and business consultation;

[0154] Functional requirements: voice interaction, terminology recognition, workflow guidance, and work order generation;

[0155] Constraints: Response delay ≤ 1s, terminology accuracy ≥ 97%;

[0156] Quantitative indicator: Consultation resolution rate ≥ 95%;

[0157] The requirement is to verify for conflicts and integrity, and output structured JSON.

[0158] Step S3: The AI ​​Agent builds its own version;

[0159] Task analysis and decomposition agent generation framework: voice interaction agent, terminology parsing agent, process scheduling agent, work order generation agent;

[0160] The workflow generation agent constructs the "voice access → intent recognition → process execution → response output → work order generation" process;

[0161] Agent generates prompts specifically for the power industry;

[0162] Tool selection: Agent filtering for power ASRs, terminology dictionary, semantic matching tool;

[0163] Develop a programming agent to correct electrical terminology errors;

[0164] Generate the initial version.

[0165] Step S4: Testing, quality control, and iteration, the specific process is as follows:

[0166] The simulated dialogue generated 1000 test cases, and quality control revealed the following:

[0167] The terminology recognition error rate was 8%, which is higher than the threshold.

[0168] Long speech is poorly segmented.

[0169] Quality Improvement Agent Output Strategy:

[0170] Load the power terminology hot word dictionary;

[0171] Optimize VAD endpoint detection;

[0172] Enable post-processing error correction model.

[0173] After retesting, the error rate dropped to 2.8% and the latency was 0.7s, meeting the standards for deployment.

[0174] Step S5: Requirement Change Iteration:

[0175] The customer added a "proactive push notification for power outage information" requirement. The dynamic strategy iterated and the Agent parsing was changed. The task analysis and decomposition Agent was used to adjust the architecture and add a push module. The iterative update was completed within 2 hours.

[0176] Example 2: Autonomous construction of an intelligent voice assistant for office scenarios, including the following steps:

[0177] Step S1: Multimodal data normalization, the processing procedure is as follows:

[0178] Input: Office requirement documents, meeting recordings, screenshots of the operation interface, and process videos.

[0179] Standardized office business documents are output after being organized.

[0180] Step S2: Requirements analysis, the process is as follows:

[0181] Extraction: Requirements such as file upload, meeting minutes, scheduling, approval reminders, and multimodal material management, with clear constraints: Batch upload support, format validation, and retry on failure.

[0182] Step S3: Agent Collaborative Development:

[0183] Generate a multi-agent framework, automatically generate prompts, workflows, and tool filters, develop batch upload and implicit requirement completion plugins for programming agents, and complete the initial version.

[0184] Step S4: Test Optimization:

[0185] Quality inspection revealed large timing alignment errors and inaccurate demand clustering.

[0186] The quality improvement agent optimized the time alignment strategy and adjusted the clustering threshold, achieving an accuracy of 99% after optimization.

[0187] Step S5: Change Iteration:

[0188] The system completed the iteration within one hour after adding the requirement to "create tasks via voice".

[0189] This invention's automatic parsing of multimodal data business requirements breaks through the traditional manual sorting mode, enabling the direct conversion of voice, text, images, and videos into structured development requirements with a parsing accuracy of ≥95%. It features multi-agent collaborative autonomous development, where multiple functional agents work together to automatically generate the entire process of framework, workflow, prompts, tools, and plugins without manual coding. It constructs an iterative process for development, testing, quality inspection, and optimization, with optimization strategies autonomously generated and executed by AI, achieving near 100% test coverage. It responds to changing requirements in real time, automatically parsing, refactoring, and iterating rapidly, reducing the cycle time from days to hours. Through industry knowledge graphs and plug-in design, it achieves lightweight cross-industry adaptation, enabling rapid adaptation to business scenarios across multiple domains.

[0190] This invention can be directly applied to intelligent voice robot development companies, artificial intelligence service providers, enterprise digital construction departments, government customer service platforms, financial intelligent interaction systems, industrial scheduling systems, smart home control systems, etc., enabling the large-scale, standardized, and highly efficient independent construction of intelligent voice robots, and possessing broad market promotion and application prospects.

[0191] This invention also provides a computer device. Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention; see the attached drawing. Figure 6 As shown, the computer device includes: an input device 23, an output device 24, a memory 22, and a processor 21; the memory 22 is used to store one or more programs; when the one or more programs are executed by the one or more processors 21, the one or more processors 21 implement the AI ​​Agent autonomous construction and optimization method for voice robots provided in the above embodiments; wherein the input device 23, the output device 24, the memory 22, and the processor 21 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.

[0192] The memory 22, as a read / write storage medium for a computing device, can be used to store software programs and computer-executable programs, such as the program instructions corresponding to the AI ​​Agent autonomous construction and optimization method for voice robots described in this embodiment of the invention. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the device. Furthermore, the memory 22 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 22 may further include memory remotely located relative to the processor 21, and these remote memories can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0193] Input device 23 can be used to receive input digital or character information, and generate key signal inputs related to user settings and function control of the device; output device 24 may include display devices such as a display screen.

[0194] The processor 21 executes various functional applications and data processing of the device by running software programs, instructions and modules stored in the memory 22, thereby realizing the above-mentioned method for autonomous construction and optimization of AI Agent for voice robots.

[0195] The computer equipment provided above can be used to execute the AIAgent autonomous construction and optimization method for voice robots provided in the above embodiments, and has corresponding functions and beneficial effects.

[0196] This invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the AI ​​Agent autonomous construction and optimization method for voice robots provided in the above embodiments. The storage medium can be any type of memory device or storage device, including: mounting media such as CD-ROM, floppy disk, or magnetic tape; computer system memory or random access memory such as DRAM, DDRRAM, SRAM, EDORAM, Rambus RAM, etc.; non-volatile memory such as flash memory, magnetic media (e.g., hard disk or optical storage); registers or other similar types of memory elements; the storage medium may also include other types of memory or combinations thereof; furthermore, the storage medium may reside in a first computer system in which the program is executed, or it may reside in a different second computer system connected to the first computer system via a network (such as the Internet); the second computer system can provide program instructions to the first computer for execution. The storage medium includes two or more storage media that can reside in different locations (e.g., in different computer systems connected via a network). The storage medium can store program instructions (e.g., specifically implemented as a computer program) executable by one or more processors.

[0197] Of course, the computer-executable instructions provided in the embodiments of the present invention are not limited to the method for autonomous construction and optimization of AI Agent for voice robots as described in the above embodiments, but can also execute related operations in the method for autonomous construction and optimization of AI Agent for voice robots provided in any embodiment of the present invention.

[0198] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0199] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for autonomous construction and optimization of AI agents for voice robots, characterized in that, Includes the following steps: S1. Receiving and preprocessing multimodal raw business materials, performing modal normalization on multimodal data such as voice, text, images, and video, and outputting standardized unstructured business data; S2. Perform intelligent parsing of unstructured business data to extract core business elements, complete conflict detection and standardization mapping, and output structured business requirement data. S3. Based on the AI ​​Agent intelligent development and operation platform, it automatically generates multi-Agent frameworks, interaction protocols, business workflows, and large model prompts according to structured business needs, completes tool selection and customized plugin development, and builds the initial version of the intelligent voice robot. S4. Start the simulated dialogue agent to conduct automated interaction testing. The service quality and satisfaction inspection agent completes the quality assessment. If the standard is not met, the quality improvement agent generates optimization strategies and drives iteration. If the standard is met, a version that can be launched is output. S5. Through dynamic strategy iteration, the Agent receives and parses requirement changes, automatically reconstructs the system architecture and execution process, and completes rapid iteration.

2. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, The modal regulation mentioned in step S1 includes: Perform noise reduction, voice enhancement, ASR transcription, and spoken language normalization on the speech data, and output sequential plain text; Perform encoding cleaning, language normalization, and redundancy filtering on text data, and output unformatted standard text. Perform tilt correction, OCR recognition, and visual semantic description on image data, and output business-related text content; Perform audio-visual separation, keyframe extraction, separate normalization and temporal alignment on video data, and output coherent business content; All modalities are ultimately mapped to plain text, denoised, and time-series unstructured data.

3. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, The intelligent parsing of business requirements in step S2 includes: Unstructured data pre-cleaning and normalization, deep understanding of business semantics and intent recognition, extraction of core elements of business requirements, structured mapping and standardization of requirement elements, requirement conflict detection and intelligent verification, structured requirement output and iterative optimization. The core elements include the business entity, core objectives, functional requirements, constraints, quantitative indicators, and related resources.

4. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, The AI ​​Agent intelligent development and operation platform mentioned in step S3 includes: Task analysis and decomposition agent, workflow generation agent, prompt word generation agent, tool selection agent, programming agent, simulated dialogue agent, quality improvement agent, dynamic strategy iteration agent, service quality and satisfaction inspection agent.

5. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 4, characterized in that, The task analysis and decomposition of Agent execution: Extract the core elements of the structured requirements, break them down into a hierarchy of total requirements, first-level sub-tasks, and second-level sub-tasks, match them with corresponding Agents, formulate a multi-Agent framework and interaction protocol, and clarify the calling order, data format, constraint verification, and exception handling mechanism.

6. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, The prompt word generates the Agent for execution: Based on the type of business requirement, core tasks, constraints, and characteristics of the target model, it automatically generates dedicated prompts containing task instructions, constraints, and output specifications, adapting to the semantic understanding and execution boundaries of large models.

7. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 4, characterized in that, The tool selects an Agent to execute: Based on the principles of functional adaptation, performance indicators, technology stack compatibility, and optimal cost-effectiveness, matching tools are selected from the tool library and sorted and output according to historical adaptation effect, resource consumption, and deployment cost. The toolkit includes speech processing tools, text processing tools, multimodal processing tools, large model inference tools, and system optimization tools.

8. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 4, characterized in that, The programming agent executes: It automatically completes plugin development, debugging, and deployment based on customized needs, supports Python, Docker, and RESTful interfaces, and outputs functional plugins that can be directly integrated, including plugins for terminology correction, implicit requirement completion, time sequence alignment, classification and clustering, and latency optimization.

9. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, If the quality is not met in step S4, the quality improvement agent will generate an optimization strategy and drive iteration, including: simulating dialogue to generate test data, the quality inspection agent outputting a defect report, the quality improvement agent generating an executable optimization strategy, feeding back to the task analysis and decomposition agent to reconstruct the architecture, regenerating the version and retesting, until the quality is met; When the iteration fails to converge, a manual review and intervention mechanism is triggered.

10. The method for autonomous construction and optimization of AI Agent for voice robots according to claim 1, characterized in that, The requirement change iteration mentioned in step S5 includes: The dynamic strategy iteration agent parses, organizes, and structures the changed content, and outputs standardized change requirements; Task analysis and breakdown: The agent adjusts its architecture, workflow, and tool configuration based on changing requirements; It automatically triggers a closed loop of development, testing, quality inspection, and optimization to complete the version update.

Images