Intelligent agent automatic generation and scheduling system based on artificial intelligence large language model

Abstract

The invention provides an intelligent agent automatic generation and scheduling system based on an artificial intelligence large language model. The system comprises an intelligent agent configuration module driven by the large language model, a workflow engine and automation module and a front-end and rear-end integration module. The agent configuration module driven by the large language model realizes automatic configuration of an agent and a workflow based on the large language model, and comprises a large language model interaction layer and an agent configuration module; the workflow engine and automation module is responsible for execution logic of workflow, supports multi-task parallel processing and comprises a workflow engine module and a task scheduler; the front-end and rear-end integration module realizes cooperative work of a JavaScript front end and a Python Django rear end by designing a unified data interface, the front end is responsible for interacting with a user and displaying an execution state of a workflow, and the rear end provides service logic processing and data persistence.

Description

Technical Field

[0001] The present invention relates to an intelligent agent, and in particular to an intelligent agent automatic generation and scheduling system based on an artificial intelligence large language model. Background Art

[0002] Artificial intelligence (AI) technology has made significant progress in recent years, especially in the development and application of agents and workflow systems. AI agents are systems that can perform tasks and make decisions autonomously, and usually have the ability to perceive, reason, learn, and act. With breakthroughs in deep learning, natural language processing, and reinforcement learning, agents have shown great potential in more and more application scenarios.

[0003] Modern AI agents are widely used in a variety of tasks, from personal assistants to complex decision support systems. For example, virtual assistants (such as Siri, Alexa, etc.) can process user requests, retrieve information, or control smart devices; agents in self-driving cars can make decisions and navigate in real time. In addition, AI agents also play a key role in finance, healthcare, industrial automation and other fields, improving efficiency and accuracy through data analysis and prediction.

[0004] At the same time, advances in workflow systems have enabled automation and optimization of various tasks. Workflow systems are often used to manage and coordinate the execution of multi-step tasks, including resource allocation, task scheduling, and progress tracking. Traditional workflow systems are based on preset rules or processes and have difficulty coping with complex and changing environments. Today, the combination of AI agents and workflow systems makes these processes more intelligent and flexible. AI agents can improve the efficiency and adaptability of workflow systems by dynamically adjusting the order of task execution, optimizing resource utilization, and making decisions based on real-time data.

[0005] The current trend is to seamlessly integrate AI agents with automated workflows to support complex multi-tasking, especially in scenarios such as large-scale data processing, real-time feedback, and cross-domain collaboration. Agent-driven workflow systems not only improve the operational efficiency of enterprises, but also promote digital transformation, and gradually become an important tool for various industries to enhance their competitiveness.

[0006] Prior Art Related to the Present Invention: The technologies related to current AI agents and workflow systems cover a wide range of areas, from basic machine learning frameworks to highly complex automation and decision support systems. These technologies complement each other and promote the advancement of AI agents and workflow systems. The following are some related technologies:

[0007] 1. Deep learning and machine learning technology

[0008] Deep learning and traditional machine learning algorithms are the foundation of AI agents. Deep learning relies on neural networks, especially architectures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which can extract features, patterns and regularities from massive amounts of data. For agents, deep learning technology has been widely used in natural language processing (NLP), image recognition, speech recognition and reinforcement learning.

[0009] TensorFlow and PyTorch are commonly used open source deep learning frameworks that support everything from simple model training to large-scale production deployment. The Transformer architecture and similar technologies, such as BERT and GPT, provide more efficient performance for intelligent dialogue systems and language processing tasks.

[0010] 2. Reinforcement Learning (RL)

[0011] Reinforcement learning is a machine learning method in which an agent learns how to optimize its behavior through trial and error by interacting with the environment. Reinforcement learning technology has great potential in areas such as automated decision-making, robot control, and game AI.

[0012] Deep Reinforcement Learning (DRL) combines deep learning with reinforcement learning, enabling intelligent agents to learn complex behavioral strategies in high-dimensional spaces. It is widely used in autonomous driving, robotics, and complex system management.

[0013] 3. Natural Language Processing (NLP)

[0014] NLP is a key technology for AI agents to understand and generate human language. Technological breakthroughs in recent years, such as Transformer-based language models such as BERT and GPT, have enabled AI agents to perform tasks such as semantic understanding, dialogue generation, and information retrieval.

[0015] Pre-trained language models can be used in systems such as intelligent assistants and customer service robots, enabling these systems to interact with language more naturally. Sentiment analysis and text summarization techniques are widely used in document processing and user feedback analysis in intelligent analytical workflow systems.

[0016] 4. Multi-Agent Systems (MAS)

[0017] Multi-agent systems are a technology that studies how multiple AI agents work together, compete, or interact in the same environment. It is applicable to large-scale distributed systems, especially in application scenarios such as drone swarms, automated warehouse management, and traffic scheduling.

[0018] Distributed AI systems allow different agents to share information and coordinate actions to complete complex tasks. Game theory and cooperative learning are key technologies in multi-agent systems, enabling agents to cooperate or compete effectively.

[0019] 5. Workflow Automation and Process Mining

[0020] Workflow systems rely on automation technology to coordinate and manage tasks. Modern workflow systems, combined with AI technology, are able to optimize complex task execution through dynamic scheduling and adaptive processes. Process mining technology can extract actual processes from historical data, analyze their efficiency, and provide suggestions for future process optimization.

[0021] Robotic Process Automation (RPA) is one of the most widely used workflow automation technologies, and is widely used in finance, customer service, data management and other fields. RPA uses a rule engine to automate repetitive tasks, and when combined with AI, it can handle more complex tasks and adapt to changes. Process mining is a technology that extracts useful information from the log data of an enterprise information system to discover, optimize and monitor business processes, helping enterprises improve process efficiency. Summary of the invention

[0022] The present invention provides an agent automatic generation and scheduling system based on artificial intelligence large language model to solve the following problems:

[0023] 1. Complexity and efficiency of agent and workflow configuration: Existing agent configuration and workflow systems often rely on a lot of manual intervention, and the configuration process is complex, error-prone and difficult to expand. With the dynamic changes in business needs, how to enable agents and workflows to quickly adapt to these changes has become a major challenge. The present invention aims to reduce manual intervention and improve the agility and scalability of the system by automatically generating and configuring agents and workflows using a large language model (LLM).

[0024] 2. Cross-platform integration and compatibility issues: In the prior art, different front-end and back-end technology stacks (such as JavaScript front-end and Python back-end) are usually difficult to integrate seamlessly. Especially in the workflow system, the data interaction and synchronization efficiency between the front-end and the back-end is not high, which affects the overall performance of the system. The present invention achieves seamless docking between the JavaScript front-end and the Python Django back-end by building a unified workflow engine and supporting engineering facilities, solves the problem of cross-platform integration, and ensures the efficiency and stability of the system.

[0025] 3. Workflow dynamic adjustment and optimization issues: Most existing workflow systems are based on static configuration, which is difficult to optimize and adjust dynamically in real time. When business needs change, reconfiguring the workflow often takes a lot of time and resources. The present invention uses the automatic configuration capability of the large language model to dynamically adjust and optimize according to real-time data and context, ensuring the flexibility and intelligence of the workflow, thereby effectively improving the response speed and processing efficiency of the system.

[0026] 4. Generality and scalability of engineering facilities: Most existing workflow systems lack sufficient generality and scalability, and are often difficult to cope with complex and diverse application scenarios. The present invention provides a complete set of engineering facilities that not only supports standard intelligent agents and workflow configurations, but can also be flexibly expanded according to different business needs, thereby enhancing the generality of the system and being applicable to a wide range of application scenarios.

[0027] The technical solution is as follows:

[0028] An intelligent agent automatic generation and scheduling system based on an artificial intelligence large language model, including an intelligent agent configuration module driven by a large language model, a workflow engine and automation module, and a front-end and back-end integration module;

[0029] The agent configuration module driven by the large language model realizes automatic configuration of the agent and workflow based on the large language model, including a large language model interaction layer and an agent configuration module;

[0030] The workflow engine and automation module are responsible for the execution logic of the workflow and support multi-task parallel processing, including a workflow engine module and a task scheduler;

[0031] The front-end and back-end integration module realizes the collaborative work of JavaScript front-end and Python Django back-end by designing a unified data interface. The front-end is responsible for interacting with users and displaying the execution status of the workflow, and the back-end provides business logic processing and data persistence, including JavaScript front-end, Python Django back-end and data interface layer for front-end and back-end communication.

[0032] Furthermore, the large language model interaction layer is an interface for the system to interact with an external intelligent language model, which is used to receive natural language instructions from users and generate corresponding agent configurations or workflow suggestions, including a natural language parser, a model call interface, and an output result optimizer;

[0033] The natural language parser decomposes and parses the natural language input by the user, and converts the unstructured language into structured instructions or configuration requirements;

[0034] The model calling interface enables access to a pre-trained large language model and generates agent configuration or workflow suggestions based on context data and user needs;

[0035] The output result optimizer is used to perform secondary processing on the results generated by the large language model so that the generated configuration file conforms to the format expected by the system and is compatible with subsequent modules.

[0036] Further, the agent configuration module includes a configuration generation engine, an agent orchestrator, and an adaptive optimization module;

[0037] The configuration generation engine automatically generates agent configuration files using a large language model by parsing business requirements or natural language instructions input by users, and dynamically adjusts workflow nodes;

[0038] The agent scheduler is used to assign the generated agent configuration files to different workflow tasks, realize dynamic scheduling and task allocation of agents, and ensure collaborative work between different tasks;

[0039] The adaptive optimization module automatically optimizes the configuration by analyzing the working status and performance data of the intelligent agent in real time, ensuring that the system can adaptively adjust according to changes in demand.

[0040] Further, the agent configuration generation engine includes a configuration template generator, a configuration file validator and a configuration file optimizer;

[0041] The configuration template generator selects and fills in a suitable configuration template based on the initial configuration generated by the large language model to ensure that the functions and behaviors of the agent meet expectations;

[0042] The configuration file validator performs syntax and logic verification on the generated configuration file to ensure the correctness and completeness of the configuration and compatibility with the agent framework;

[0043] The configuration file optimizer optimizes the agent configuration according to the actual performance of the agent in the workflow execution.

[0044] Furthermore, the workflow engine module is used to manage and schedule various tasks in the workflow. Based on event-driven and rule-driven engines, the workflow can be dynamically adjusted according to real-time data, including a workflow parser, event-driven and rule-driven engines, and a task queue manager;

[0045] The workflow parser is used to parse the workflow definition generated by the large language model, build an executable workflow task tree within the system, and define the execution order and dependency relationship of each task;

[0046] The event-driven and rule engine monitors various system events, adjusts the execution path of the workflow in real time, and dynamically selects the best task branch according to predefined rules.

[0047] The task queue manager enables the system to manage a large number of concurrent tasks and dynamically allocate computing resources based on task priorities and dependencies.

[0048] Furthermore, the task scheduler is used to manage the specific task scheduling in the workflow, supports asynchronous and concurrent execution, analyzes the dependencies of tasks, and includes a task status monitor, a scheduling optimizer, and a fault handling module;

[0049] The task status monitor is used to monitor the execution status of each workflow task and feed back the status to the front-end user interface and log system;

[0050] The scheduling optimizer dynamically adjusts the priority and resource allocation strategy of the task according to the real-time execution status of the task and the system load;

[0051] The fault handling module is used to handle abnormal situations in task execution and supports task retry and rollback.

[0052] Furthermore, the JavaScript front end is built based on the React or Vue.js framework, and realizes real-time communication with the back end through WebSocket or SSE; the Python Django back end is based on the Django framework, is responsible for processing business logic and storage and management of intelligent agent configuration, interacts with the front end through the RESTful API, and supports asynchronous task execution through ASGI.

[0053] Furthermore, the system also includes a security authentication module responsible for the security management of the data interface layer, including an identity authentication module, a rights management module, and a log tracking and auditing module;

[0054] The identity authentication module is used to verify the user's identity and ensure secure access to the system;

[0055] The authority management module allocates the user's operating authority in the system according to the user's identity and role;

[0056] The log tracking and auditing module is used to record all user operations and support security auditing and log analysis.

[0057] Furthermore, the system also includes a data storage and persistence module, which is used to manage the storage and persistence of all configurations, task execution data, workflow status, etc. in the system, including a configuration file storage module, a task execution log manager, and a data backup and recovery mechanism module;

[0058] The configuration file storage module is used to store the generated agent configuration file in a database;

[0059] The task execution log manager is used to record the execution log of each workflow task, including time, status, and output information;

[0060] The data backup and recovery mechanism module is used to provide regular backup of system data and perform data recovery when necessary.

[0061] According to the system for automatic generation and scheduling of intelligent agents based on artificial intelligence large language model according to claim 1, it is characterized in that: the system also includes an intelligent knowledge base and a rhetorical natural language communication module, the intelligent knowledge base is used to store and manage relevant knowledge in the system, including a knowledge base storage module, a natural language retrieval module and a self-learning and knowledge update module; the rhetorical natural language communication module introduces a rhetorical natural language interaction mechanism, including a rhetorical question generation module, a multi-round dialogue management module and a context retention and task association module;

[0062] The above-mentioned intelligent agent automatic generation and scheduling system based on artificial intelligence large language model significantly improves the intelligence, flexibility and efficiency of the system. The specific advantages are as follows:

[0063] 1. Intelligent automatic configuration

[0064] Limitations of existing technologies: Traditional intelligent agents and workflow systems often require manual configuration, and the configuration process is cumbersome and error-prone. Users must define every step and rule of the system in detail, which takes a lot of time and effort. Manually updating configurations is particularly challenging when dealing with complex tasks and dynamically changing business needs.

[0065] Advantages of the present invention: The present invention realizes full automation of agent and workflow configuration by introducing a large language model. Users only need to input requirements through natural language, and the system can automatically parse and generate configuration files without manually writing codes or configuring rules. This greatly reduces manual intervention, improves the accuracy and efficiency of configuration, and greatly reduces the possibility of errors. Especially when facing complex dynamic environments, the system can automatically adapt to changes, further improving the level of intelligence.

[0066] 2. Dynamic adjustment and real-time optimization

[0067] Limitations of existing technologies: Most workflow systems execute predefined processes after startup and rarely make dynamic adjustments based on real-time conditions. Such static configurations are difficult to cope with rapid changes in the business environment, resulting in low workflow execution efficiency and serious waste of resources.

[0068] Advantages of the present invention: The present invention supports dynamic adjustment of workflows through event-driven and rule engines in the workflow engine, as well as task schedulers. The system can monitor task execution in real time, and adjust paths or reallocate resources according to actual status to ensure the best execution path for the workflow. This dynamic optimization feature greatly improves the flexibility and responsiveness of the system, can cope with real-time changing business needs, and avoids inefficiency and resource waste in static systems.

[0069] 3. Cross-platform front-end and back-end integration

[0070] Limitations of existing technologies: Traditional workflow systems and agent configuration tools often have problems with front-end and back-end integration. Especially in cross-platform architectures, data transmission and synchronization between different technology stacks are inefficient, and integration costs are high, affecting the overall performance of the system.

[0071] Advantages of the invention: The invention designs a unified front-end and back-end communication interface, and realizes seamless integration of JavaScript front-end and Python Django back-end through a standardized data interface layer. The system adopts WebSocket or SSE technology to ensure that the front-end interface can obtain workflow status and task execution feedback in real time. This loose coupling design not only improves the efficiency of front-end and back-end communication, but also ensures the flexible scalability of the system, so that each component can be independently upgraded or replaced, reducing the maintenance cost of the system.

[0072] 4. High concurrency and task management optimization

[0073] Limitations of existing technologies: When faced with highly concurrent tasks, traditional workflow systems often have difficulty managing resources effectively, leading to task execution bottlenecks and performance degradation. At the same time, the lack of real-time monitoring and optimization of task status increases the risk of system failure or inefficient operation.

[0074] Advantages of the present invention: The present invention can efficiently manage concurrent tasks, dynamically adjust task priorities, and reasonably allocate system resources through the scheduling optimizer of the task scheduler. The task status monitor provides real-time feedback on the task execution status, and the fault handling module provides a powerful error recovery capability to ensure that the system can still run stably and efficiently in a high-concurrency environment. In addition, the dynamic adjustment and retry mechanism of tasks greatly improves the robustness of the system in complex environments, ensuring that tasks can be completed smoothly.

[0075] 5. Adaptive learning and optimization

[0076] Limitations of existing technologies: Existing workflow systems lack adaptive learning capabilities and usually rely on pre-set rules and processes, and are unable to self-optimize and adjust based on data collected during execution.

[0077] Advantages of the present invention: The present invention realizes the adaptive capability of the system through the configuration file optimizer and the adaptive optimization module. The system can automatically optimize the agent configuration and workflow execution path according to the actual task execution situation. This adaptive optimization mechanism based on real-time feedback ensures that the system can continuously improve its performance and efficiency according to changes in the environment. Compared with fixed preset rules, the adaptive characteristics of the present invention significantly enhance the flexibility and intelligence level of the system.

[0078] 6. Simplified user experience and scalability

[0079] Limitations of existing technologies: In traditional intelligent agents and workflow configuration, users need to have a high technical background and master complex configuration languages ​​or tools. For non-technical users, the usage threshold is high, and the system scalability is limited, making it difficult to quickly respond to new business needs.

[0080] Advantages of the present invention: Through the automatic configuration capability of the large language model, the present invention significantly reduces the technical threshold for users, and non-technical personnel can also generate agents and workflow configurations by describing requirements in natural language. This simplified user experience not only improves the ease of use of the system, but also speeds up the configuration and deployment process. At the same time, the system is designed with a flexible interface and an extensible workflow engine, which can quickly adapt and expand according to business changes to meet customized needs in different scenarios.

[0081] 7. Data persistence and fault recovery

[0082] Limitations of existing technologies: Traditional systems usually have deficiencies in data persistence and fault recovery when processing large-scale data. In particular, when encountering system failures or data corruption, data recovery is difficult, resulting in irreparable business losses.

[0083] Advantages of the invention: The data storage and persistence module of the invention provides powerful data management functions, including version control of agent configuration files, real-time recording of task execution logs, and data backup and recovery mechanisms. Even when a system failure occurs, data can be quickly restored to ensure business continuity and stability. The persistence mechanism not only enhances the security of the system, but also provides detailed historical data support for subsequent optimization of tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a schematic diagram of the framework of the automatic generation and scheduling system of intelligent agents based on the artificial intelligence large language model;

[0085] Figure 2 It is an application diagram of the automatic generation and scheduling system of intelligent agents based on the artificial intelligence large language model. DETAILED DESCRIPTION

[0086] like Figure 1 As shown in the figure, the intelligent agent automatic generation and scheduling system based on the artificial intelligence large language model includes an intelligent agent configuration module driven by a large language model, a workflow engine and automation module, and a front-end and back-end integration module. These three modules work together to form an efficient, dynamic and scalable intelligent workflow system. Each module in the architecture is designed to be highly configurable, support expansion, and adapt to the needs of different business scenarios.

[0087] A large language model driven agent configuration module

[0088] The large language model-driven agent configuration module realizes automatic configuration of agents and workflows based on a large language model (LLM), and automatically generates and optimizes agent configuration files and agent workflow flowcharts through natural language processing, context analysis, and data-driven reasoning capabilities. The module includes a large language model interaction layer and an agent configuration module.

[0089] 1The large language model interaction layer is the interface for the system to interact with external intelligent language models (such as Wenxinyiyan, GLM, local models, etc.), which is used to receive the user's natural language instructions and generate corresponding intelligent agent configurations or workflow suggestions, including a natural language parser, a model calling interface and an output result optimizer.

[0090] Natural language parser: This component is responsible for decomposing and parsing the natural language input by the user and converting unstructured language into structured instructions or configuration requirements.

[0091] Model calling interface: Through this interface, the system can access the pre-trained large language model and generate agent configuration or workflow recommendations based on contextual data and user needs. The interface supports remote calling or local model inference.

[0092] Output Result Optimizer: Performs secondary processing on the results generated by the large language model to ensure that the generated configuration file conforms to the format expected by the system and is compatible with subsequent modules.

[0093] 2 The intelligent agent configuration module includes a configuration generation engine, an intelligent agent orchestrator and an adaptive optimization module.

[0094] Configuration generation engine: This submodule automatically generates agent configuration files using a large language model by parsing business requirements or natural language instructions input by users, and dynamically adjusts workflow nodes.

[0095] Agent Orchestrator: Assigns the generated agent configuration files to different workflow tasks, realizes dynamic scheduling and task allocation of agents, and ensures collaborative work between different tasks.

[0096] Adaptive Optimization Module: By analyzing the working status and performance data of the intelligent agent in real time, it automatically optimizes the configuration to ensure that the system can adaptively adjust according to changes in demand.

[0097] 2.1 Furthermore, the agent configuration generation engine is responsible for generating the agent's configuration file based on the user's input requirements and the suggestions provided by the large language model, ensuring that the configuration file is structured, executable, and can be seamlessly integrated into the workflow system, including a configuration template generator, a configuration file validator, and a configuration file optimizer.

[0098] Configuration Template Generator: This submodule selects and fills in appropriate configuration templates based on the initial configuration generated by the large language model to ensure that the functions and behaviors of the agent meet expectations.

[0099] Configuration file validator: performs syntax and logic validation on the generated configuration file to ensure the correctness, completeness and compatibility with the agent framework.

[0100] Profile Optimizer: Optimizes agent configurations based on the agent’s actual performance in workflow execution to ensure system adaptability and performance improvement.

[0101] 2. Workflow Engine and Automation Module

[0102] The workflow engine and automation module are responsible for the core execution logic of the workflow system, ensuring that each task is completed efficiently within the specified order and supporting multi-task parallel processing, which includes a workflow engine module and a task scheduler.

[0103] 3 The workflow engine module is the core of the whole system, responsible for managing and scheduling each task in the workflow. The module has the ability to dynamically adjust and optimize the workflow to ensure efficient execution of tasks and reasonable allocation of resources. The workflow engine module is based on event-driven and rule-driven engines, and can dynamically adjust the workflow according to real-time data. It includes a workflow parser, event-driven and rule-driven engines, and a task queue manager.

[0104] Workflow parser: responsible for parsing the workflow definition generated by the large language model, building an executable workflow task tree within the system, and defining the execution order and dependencies of each task.

[0105] Event-driven and rule engine: By monitoring various system events (such as task completion, data changes, etc.), the execution path of the workflow is adjusted in real time, and the best task branch is dynamically selected according to predefined rules. The predefined rules can be updated according to real-time data, user feedback or system events.

[0106] Task Queue Manager: Through the task queue manager, the system can manage a large number of concurrent tasks and dynamically allocate computing resources based on task priorities and dependencies.

[0107] The task scheduler is responsible for scheduling workflow tasks, and is used to manage the specific task scheduling in the workflow, ensure the efficient execution of each task, support asynchronous and concurrent execution, and ensure the efficient use of resources by analyzing the dependency of tasks. It includes a task status monitor, a scheduling optimizer, and a fault handling module.

[0108] Task status monitor: responsible for monitoring the execution status of each workflow task (such as start, in progress, completed, failed, etc.), and feeding back the status to the front-end user interface and log system.

[0109] Scheduling optimizer: Dynamically adjusts task priorities and resource allocation strategies based on the real-time execution of tasks and system load to ensure the overall performance and response speed of the system.

[0110] Fault handling module: This module handles abnormal situations in task execution, such as task failure, timeout, etc., and supports task retry and rollback to ensure the stability of the system in the event of a fault.

[0111] Three front-end and back-end integration modules

[0112] The front-end and back-end integration module realizes efficient collaboration between the JavaScript front-end and the Python Django back-end by designing a unified data interface. The front-end is responsible for interacting with users and displaying the execution status of the workflow, while the back-end provides business logic processing and data persistence, including the JavaScript front-end, Python Django back-end and the data interface layer for front-end and back-end communication.

[0113] 5 JavaScript front-end: The front-end is built based on modern front-end frameworks (such as React or Vue.js), responsible for user interface interaction and data display. Real-time communication with the back-end is achieved through WebSocket or SSE (Server-Sent Events) technology to ensure real-time updates of workflow status.

[0114] 6Python Django backend: The backend is mainly based on the Django framework, responsible for processing business logic and storing and managing agent configurations. It interacts with the frontend through the RESTful API and supports asynchronous task execution through ASGI to ensure high concurrency and scalability of the system.

[0115] 7. Data interface layer: This layer provides a unified communication interface for the front-end and back-end to ensure data consistency and real-time performance. It realizes communication between the front-end and back-end through standardized interfaces, ensures loose coupling between the front-end and back-end, and facilitates subsequent expansion and maintenance. It mainly uses RESTful API, WebSocket or SSE technology, including data transmission interface.

[0116] Data transmission interface: Provides a unified interface for data transmission, supports real-time interaction between the front-end user interface and the back-end business logic, and handles updates to agent configuration and workflow execution status.

[0117] Four related modules

[0118] The association module is designed to make the operation of this system more perfect. In addition to the three modules of the intelligent agent configuration module driven by the large language model, the workflow engine and automation module, and the front-end and back-end integration module, other modules that can be used in conjunction include security authentication module, data storage and persistence module, intelligent knowledge base and rhetorical natural language communication module.

[0119] 8 Security authentication module: responsible for the security management of the data interface layer, ensuring the security of data transmission, using identity authentication, permission management and encryption mechanisms to protect sensitive data, and responsible for the identity authentication and permission allocation of different users in the system to ensure the security of system resources and data. The system adopts an identity-based access control mechanism (IBAC) to limit the operations of different users in the system through the division of roles and permissions, including identity authentication module, permission management module and log tracking and auditing module.

[0120] Identity authentication module: Verify user identity through various methods (such as username and password authentication, OAuth, SAML, etc.) to ensure secure access to the system.

[0121] Permission management module: Assign user operation permissions in the system according to their identities and roles. Users with different permissions can only access and operate data and functions related to their roles, preventing unauthorized operations.

[0122] Log tracking and auditing module: records all user operations, supports security auditing and log analysis, and facilitates system administrators to track potential security risks and issues.

[0123] 9Data storage and persistence module: responsible for the storage and persistence of all configurations, task execution data, workflow status, etc. in the management system. This module supports efficient data reading, writing, data backup and recovery, including configuration file storage module, task execution log manager and data backup and recovery mechanism module.

[0124] Configuration file storage module: Store the generated agent configuration files in the database to ensure fast retrieval, modification and version control. Supports storage in multiple databases (such as SQLite or PostgresQL, etc.).

[0125] Task Execution Log Manager: records the execution log of each workflow task, including time, status, output and other information, to facilitate debugging and task analysis.

[0126] Data backup and recovery mechanism module: This submodule provides regular backup of system data and performs data recovery when necessary to ensure the system's ability to recover in the event of data loss or damage.

[0127] 10 Intelligent knowledge base, used to store and manage all relevant knowledge in the system, including user's common problems, system configuration details, business process documents, etc. The knowledge base supports natural language-based retrieval and can provide real-time reference information for the system during agent configuration and workflow execution. It includes knowledge base storage module, natural language retrieval module and self-learning and knowledge update module.

[0128] Knowledge base storage module: responsible for storing structured and unstructured knowledge information, supporting data storage in multiple formats (such as documents, FAQs, configuration templates, etc.).

[0129] Natural language retrieval module: Through natural language processing technology, users can directly query relevant information in the knowledge base through questions or keywords. This module can intelligently recommend the most appropriate answer based on the user's question context.

[0130] Self-learning and knowledge updating module: The system can automatically update the information in the knowledge base according to user feedback and operations, ensuring the timeliness and accuracy of the knowledge base content.

[0131] 11. Rhetorical natural language communication module: This module realizes a more intelligent human-computer interaction experience by introducing a rhetorical natural language interaction mechanism. Different from the traditional question-answering system, this module can ask questions based on the questions raised by the user to obtain more context information, thereby generating more accurate and effective answers or operation suggestions. It includes a rhetorical question generation module, a multi-round dialogue management module, and a context preservation and task association module.

[0132] Rhetorical question generation module: When a user asks a question or request, the system can generate rhetorical questions based on the ambiguity or incompleteness of the question to further clarify the user's intention. For example, if the user asks "configure an agent", the system may ask "What type of agent do you want to configure? What are the specific requirements?"

[0133] Multi-round dialogue management module: supports multi-round dialogues, gradually refining the questions through the user's initial questions and counter-questions until the system can accurately understand the user's needs and provide appropriate solutions or operation steps.

[0134] Context preservation and task association module: This module can maintain the conversation context in multiple rounds of conversations, so that when users ask multiple related questions, the system can automatically understand the connection between the questions and respond accordingly. It can also associate the user's natural language needs with specific tasks or operation processes to help users complete operations more quickly.

[0135] The various modules of the present invention work together to ensure the automation, flexibility and efficiency of the system.

[0136] In the various modules of the present invention described above, some modules may be replaced by other solutions to achieve the same technical effect, which are specifically described as follows:

[0137] 1. Alternative 1: Automatic configuration generation of agents based on other machine learning models

[0138] Original technical solution: The present invention uses a large language model (LLM) as the core tool for automatic configuration generation of agents and workflows.

[0139] Alternatives: You can use machine learning models based on other types (such as decision trees, random forests, support vector machines, etc.) or rule engines to generate agent configurations. For example, rule engines combined with predefined business logic can achieve automatic configuration generation. Although it is not as flexible as large language models, it is still feasible in specific business scenarios. In addition, traditional knowledge graphs and template-based generation methods can also be used as an alternative technical means to complete the task of automated configuration generation.

[0140] By covering machine learning technologies that are not large language models and rule-based and template-based generation methods, the scope of patent protection is expanded to prevent others from using these traditional technologies to achieve the same configuration automation functions.

[0141] 2. Alternative 2: Workflow Adjustment Driven by Predefined Rules

[0142] Original technical solution: The present invention relies on an intelligent rule engine to optimize and dynamically adjust the workflow execution path in real time.

[0143] Alternative solution: Predefined static rules or conditional statements can be used to replace the intelligent rule engine to adjust the workflow. For example, in a simpler scenario, by defining all possible task paths and execution conditions in advance, the system can select the appropriate execution path based on the triggering conditions. Although this method does not have the ability to learn and optimize in real time, it can still work in scenarios where the business logic is not complex.

[0144] By introducing a workflow adjustment method based on predefined conditions, the dynamic adjustment of workflow is expanded to a wider range of solutions, including intelligent and non-intelligent dynamic adjustment technologies.

[0145] 3. Alternative 3: Front-end and back-end integration based on traditional API communication

[0146] Original technical solution: The present invention adopts WebSocket or SSE technology to realize real-time communication and status feedback between the front and back ends.

[0147] Alternatives: Traditional RESTful API or GraphQL can be used as an alternative for front-end and back-end communication. Although these technologies cannot provide real-time communication, they can periodically update the workflow status of the front-end through a polling mechanism. This method is suitable for scenarios with lower real-time requirements and can still achieve communication and data synchronization between the front-end and back-end.

[0148] By covering communication methods based on RESTful API, GraphQL and polling mechanism, the protection scope of front-end and back-end communications is expanded, preventing others from circumventing the present invention by using traditional communication technologies to implement similar functions.

[0149] 4. Alternative 4: Scheduling management based on distributed task queues

[0150] Original technical solution: The present invention uses an adaptive task scheduler to perform dynamic allocation and concurrency control of tasks.

[0151] Alternative solution: You can use traditional distributed task queue systems (such as Celery, RabbitMQ, etc.) to implement task scheduling management. These task queue systems distribute tasks to different working nodes for parallel processing. Although they do not have adaptive task optimization capabilities, they are still highly practical in scenarios with high concurrency and large-scale task processing.

[0152] By covering the scheduling management technology of distributed task queues, the protection of the patent is enhanced, ensuring that the traditional distributed task scheduling system also falls within the protection scope of the present invention.

[0153] 5. Alternative 5: Hybrid mode combining manual configuration and adaptive optimization

[0154] Original technical solution: The present invention realizes automatic optimization of configuration files through an adaptive optimization module of a large language model.

[0155] Alternative solution: A hybrid mode combining manual configuration and adaptive optimization can be adopted. In this solution, users can manually define some parameters of the agent configuration, and the system is only responsible for optimizing some key parameters. This approach reduces the complexity of system automation while maintaining a certain adaptive optimization capability under manual intervention.

[0156] By introducing a mode that combines manual configuration and adaptive optimization, the scope of optimization methods is expanded to cover configuration optimization techniques that combine manual and automatic configuration, preventing others from bypassing the fully automatic optimization method.

[0157] 6. Alternative 6: Cloud-based profile management

[0158] Original technical solution: The present invention uses a local database to perform version control and persistent management of configuration files.

[0159] Alternative solution: You can use cloud storage systems (such as AWS S3, Google Cloud Storage, etc.) to manage and persist configuration files. By storing configuration files in the cloud, the system can achieve high availability and backup management across regions. This approach has greater elasticity and flexibility in large-scale distributed systems, especially when processing massive amounts of data.

[0160] By introducing cloud storage-based file management methods, covering data management methods on different storage platforms (local and cloud), comprehensive protection of data storage and persistence technology is ensured.

[0161] 7. Alternative 7: Blockchain-based task execution log and data backup

[0162] Original technical solution: The present invention uses traditional database technology to store and back up task execution logs.

[0163] Alternative solution: Blockchain technology can be used to achieve decentralized storage and tamper-proof backup of task execution logs. By recording task execution logs on the blockchain, the system can ensure the transparency and security of log data while providing powerful auditing capabilities. This solution is particularly suitable for scenarios with high data security requirements, such as the financial and medical industries.

[0164] By introducing blockchain technology, the protection solution covers decentralized log management and data backup technology to prevent others from achieving the same logging function through blockchain technology.

[0165] Through the various modules of the present invention, the process of automatic generation and scheduling of intelligent agents can be realized, such as Figure 2 As shown, the overall steps of this system are as follows:

[0166] S1: This system receives natural language instructions input by users through the agent configuration module driven by the large language model, and generates corresponding agent configurations or workflow suggestions;

[0167] S2: Analyze whether there is an existing intelligent agent. If so, use the existing intelligent agent. If not, automatically create an intelligent agent.

[0168] S3: Decompose tasks through the workflow engine and automation module, and call other tools of the associated module;

[0169] S4: Solve relevant problems and provide feedback on work results.

[0170] The definitions of abbreviations and key terms used in the present invention are as follows:

[0171] API Application Programming Interface, application programming interface

[0172] NLP Natural Language Processing

[0173] LM Language Model

[0174] LLM Large Language Model

[0175] CPU Central Processing Unit

[0176] GPU Graphic Processing Unit, graphics processor

[0177] Agent

[0178] Workflow

[0179] In summary, the present invention significantly solves multiple technical problems in traditional systems by utilizing a large language model, automatic configuration of intelligent agents, and a dynamically optimized workflow engine. Compared with the prior art, the present invention has the following outstanding advantages:

[0180] (1) The configuration process is highly intelligent, which greatly reduces manual intervention and error rates.

[0181] (2) Dynamic adjustment and real-time optimization capabilities to ensure efficient execution of workflows.

[0182] (3) Cross-platform front-end and back-end integration enables seamless communication and improves system scalability and performance.

[0183] (4) Efficient task management and scheduling optimization significantly enhance the system's concurrent processing capabilities.

[0184] (5) Adaptive learning and optimization mechanisms to ensure that the system continuously improves performance.

[0185] (6) Simplified user experience, lower technical barriers, and support rapid deployment and expansion.

[0186] (7) Improved data persistence and fault recovery capabilities enhance the security and stability of the system.

[0187] The present invention has the following characteristics:

[0188] 1. Automatic configuration generation of intelligent agents based on large language models

[0189] Description of key technical points: The present invention uses the Large Language Model (LLM) as the core means of generating agent and workflow configurations, and automatically generates agent configuration files based on natural language input. Through the powerful natural language understanding and generation capabilities of the large language model, the system can parse user needs and automatically generate configuration files and task flows in a structured manner. This technology avoids the process of manually writing complex configurations, greatly improves configuration efficiency, and can dynamically adapt to changes in business needs.

[0190] 2. Dynamic adjustment and rule-driven optimization of intelligent workflow engine

[0191] Description of key technical points: The intelligent workflow engine in this invention has the ability to dynamically adjust the workflow execution path and optimize the workflow based on real-time data and rule engines. This technology can dynamically adjust the task execution order or branch selection according to the task status, system resource conditions or external events, thereby optimizing task scheduling and improving the efficiency of workflow execution.

[0192] 3. Optimization mechanism of task scheduling and concurrency control

[0193] Description of key technical points: The task management and scheduling module of the present invention realizes efficient concurrent task management and dynamic resource allocation through the scheduling optimizer. This technology can optimize the scheduling according to the priority, dependency and system load of the task, ensuring the efficient execution of the task and the rational use of system resources. In addition, the system supports dynamic retry and fault recovery of tasks, ensuring the stability and robustness of the system under high load.

[0194] 4. Adaptive optimization and continuous improvement of configuration files

[0195] Description of key technical points: The adaptive optimization module in the present invention can automatically optimize the configuration file according to the performance of the agent in the workflow execution. By analyzing the task execution data in real time, the system can adaptively adjust the configuration of the agent to improve the efficiency of task completion and reduce resource waste. This adaptive optimization mechanism enables the system to continuously learn and improve, adapting to complex business environments and changing needs.

Claims

1. An agent automatic generation and scheduling system based on artificial intelligence large language model, characterized by: It includes the agent configuration module driven by a large language model, the workflow engine and automation module, and the front-end and back-end integration module; The agent configuration module driven by the large language model realizes automatic configuration of the agent and workflow based on the large language model, including a large language model interaction layer and an agent configuration module; The workflow engine and automation module are responsible for the execution logic of the workflow and support multi-task parallel processing, including a workflow engine module and a task scheduler; The front-end and back-end integration module realizes the collaborative work of JavaScript front-end and Python Django back-end by designing a unified data interface. The front-end is responsible for interacting with users and displaying the execution status of the workflow, and the back-end provides business logic processing and data persistence, including JavaScript front-end, Python Django back-end and data interface layer for front-end and back-end communication.

2. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The large language model interaction layer is an interface for the system to interact with external intelligent language models, which is used to receive natural language instructions from users and generate corresponding agent configurations or workflow suggestions, including a natural language parser, a model call interface, and an output result optimizer; The natural language parser decomposes and parses the natural language input by the user, and converts the unstructured language into structured instructions or configuration requirements; The model calling interface enables access to a pre-trained large language model and generates agent configuration or workflow suggestions based on context data and user needs; The output result optimizer is used to perform secondary processing on the results generated by the large language model so that the generated configuration file conforms to the format expected by the system and is compatible with subsequent modules.

3. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The agent configuration module includes a configuration generation engine, an agent scheduler and an adaptive optimization module; The configuration generation engine automatically generates agent configuration files using a large language model by parsing business requirements or natural language instructions input by users, and dynamically adjusts workflow nodes; The agent scheduler is used to assign the generated agent configuration files to different workflow tasks, realize dynamic scheduling and task allocation of agents, and ensure collaborative work between different tasks; The adaptive optimization module automatically optimizes the configuration by analyzing the working status and performance data of the intelligent agent in real time, ensuring that the system can adaptively adjust according to changes in demand.

4. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 3 is characterized in that: The agent configuration generation engine includes a configuration template generator, a configuration file validator and a configuration file optimizer; The configuration template generator selects and fills in a suitable configuration template based on the initial configuration generated by the large language model to ensure that the functions and behaviors of the agent meet expectations; The configuration file validator performs syntax and logic verification on the generated configuration file to ensure the correctness and completeness of the configuration and compatibility with the agent framework; The configuration file optimizer optimizes the agent configuration according to the actual performance of the agent in the workflow execution.

5. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The workflow engine module is used to manage and schedule various tasks in the workflow. Based on event-driven and rule-driven engines, it can dynamically adjust the workflow according to real-time data, including a workflow parser, event-driven and rule-driven engines, and a task queue manager; The workflow parser is used to parse the workflow definition generated by the large language model, build an executable workflow task tree within the system, and define the execution order and dependency relationship of each task; The event-driven and rule engine monitors various system events, adjusts the execution path of the workflow in real time, and dynamically selects the best task branch according to predefined rules. The task queue manager enables the system to manage a large number of concurrent tasks and dynamically allocate computing resources based on task priorities and dependencies.

6. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The task scheduler is used to manage the specific task scheduling in the workflow, supports asynchronous and concurrent execution, analyzes the dependencies of tasks, and includes a task status monitor, a scheduling optimizer, and a fault handling module; The task status monitor is used to monitor the execution status of each workflow task and feed back the status to the front-end user interface and log system; The scheduling optimizer dynamically adjusts the priority and resource allocation strategy of the task according to the real-time execution status of the task and the system load; The fault handling module is used to handle abnormal situations in task execution and supports task retry and rollback.

7. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The JavaScript front-end is built based on the React or Vue.js framework, and realizes real-time communication with the back-end through WebSocket or SSE; the Python Django back-end is based on the Django framework, is responsible for processing business logic and storing and managing intelligent agent configuration, interacts with the front-end through RESTfulAPI, and supports asynchronous task execution through ASGI.

8. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The system also includes a security authentication module responsible for the security management of the data interface layer, including an identity authentication module, a rights management module, and a log tracking and auditing module; The identity authentication module is used to verify the user's identity and ensure secure access to the system; The authority management module allocates the user's operating authority in the system according to the user's identity and role; The log tracking and auditing module is used to record all user operations and support security auditing and log analysis.

9. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The system also includes a data storage and persistence module, which is used to store and persist all configurations, task execution data, workflow status, etc. in the system, including a configuration file storage module, a task execution log manager, and a data backup and recovery mechanism module; The configuration file storage module is used to store the generated agent configuration file in a database; The task execution log manager is used to record the execution log of each workflow task, including time, status, and output information; The data backup and recovery mechanism module is used to provide regular backup of system data and perform data recovery when necessary.

10. The system for automatically generating and scheduling intelligent agents based on an artificial intelligence large language model according to claim 1, characterized in that: The system also includes an intelligent knowledge base and a rhetorical natural language communication module. The intelligent knowledge base is used to store and manage relevant knowledge in the system, including a knowledge base storage module, a natural language retrieval module and a self-learning and knowledge update module; the rhetorical natural language communication module introduces a rhetorical natural language interaction mechanism, including a rhetorical question generation module, a multi-round dialogue management module and a context retention and task association module.

Images