An agent-based no-code orchestration system

Abstract

This application provides a no-code orchestration system based on intelligent agents, belonging to the field of no-code programming and multi-agent collaboration. It addresses the problem that building multi-agent collaborative systems in related technologies requires specialized programming knowledge and has a high technical threshold. The system dynamically organizes intelligent agents through a group collaboration container building module that responds to no-code operations such as visual drag-and-drop. A dialogue-driven intelligent agent collaboration network module parses natural language commands and generates task orchestration plans. A visual rule-protocol conversion engine module converts graphically configured interaction rules into underlying protocols to schedule intelligent agent collaboration. This allows non-technical personnel to quickly build and command complex multi-agent collaborative systems in an intuitive no-code manner, effectively reducing the technical threshold and improving the development and deployment efficiency of intelligent collaborative applications.

Description

Technical Field

[0001] This application relates to the field of no-code programming and multi-agent system collaboration, and more particularly to an agent-based no-code orchestration system. Background Technology

[0002] No-code programming technology aims to replace traditional hand-written code with graphical interfaces and modular components, enabling non-professional developers to build applications or processes. Its applications have been widely covered in fields such as data management and business process automation.

[0003] Existing technologies mainly fall into two categories: one is traditional no-code platforms centered on form and workflow design, which mainly orchestrate static data objects and predefined business process nodes; the other is development tools for conversational agents (such as chatbots) built for single-point tasks, which allow users to trigger the response of a single agent through configuration.

[0004] However, existing technologies have significant drawbacks: traditional no-code platforms struggle to organize and orchestrate multiple agents with autonomous decision-making and dynamic interaction capabilities; while the implementation of professional agent collaborative systems (multi-agent systems) heavily relies on complex programming and communication protocol development, resulting in extremely high technical barriers. This prevents non-technical personnel from building and directing a collaborative work system composed of multiple agents in a low-barrier, visual manner. Summary of the Invention

[0005] This application provides a no-code orchestration system based on intelligent agents, which can realize the dynamic organization, task-driven and collaborative interaction of multiple intelligent agents through graphical drag and drop and natural language dialogue, and solves the technical threshold problem for non-technical personnel to build complex intelligent agent collaborative systems.

[0006] In a first aspect, this application provides a no-code orchestration system based on intelligent agents. The system includes: a group collaboration container construction module, a dialogue-driven intelligent agent collaboration network module, and a visual rule-protocol conversion engine module. The group collaboration container construction module is used to dynamically construct or modify a group collaboration container for accommodating multiple intelligent agents in response to user no-code operations. The no-code operations include dragging and dropping pre-packaged intelligent agent function modules into or out of the group collaboration container via a visual interface. The dialogue-driven intelligent agent collaboration network module, connected to the group collaboration container construction module, is used to receive natural language instructions input to the group collaboration container, and generate a task orchestration plan to be collaboratively executed by multiple intelligent agents within the group collaboration container by performing semantic understanding and task decomposition on the natural language instructions. The visual rule-protocol conversion engine module, connected to the group collaboration container construction module and the dialogue-driven intelligent agent collaboration network module, is used to provide a visual interface to receive interaction rules configured for the group collaboration container, and convert the interaction rules into a low-level protocol controlling the communication and scheduling order between intelligent agents within the group collaboration container, so as to drive the interaction of the intelligent agents based on the task orchestration plan and the low-level protocol.

[0007] By adopting the above technical solution, users can quickly create intelligent agent groups through intuitive drag-and-drop operations, issue complex task commands directly through natural language, and set collaboration rules through a graphical interface. The system automatically and seamlessly transforms these advanced, no-code user interaction intentions into underlying executable multi-agent task plans and communication control protocols, thereby achieving effective orchestration of dynamic, autonomous multi-agent collaborative systems while retaining the advantages of no-code and low barrier to entry.

[0008] Furthermore, the group collaboration container construction module includes: a visual interaction unit, used to provide a graphical interface, receive user drag-and-drop operations on the intelligent agent functional module, and attribute configuration operations on the intelligent agents within the group collaboration container, wherein the attribute configuration includes associating with a knowledge base or setting system prompt words; and an intelligent agent management unit, connected to the visual interaction unit, used to encapsulate and register the intelligent agent functional module, maintain its attributes, and monitor its operating status.

[0009] By adopting the above technical solution, the interactive interface is separated from the core management function of the intelligent agent, so that the user configuration operation can directly affect the runtime attributes of the intelligent agent module, realizing the dynamic management and flexible configuration of the intelligent agent as a reusable functional module.

[0010] Furthermore, the dialogue-driven agent collaboration network module includes: a task orchestration engine unit, used to perform intent recognition on the received natural language instructions, decompose tasks and match agents based on the recognized intent and a predefined agent capability library, and form a task orchestration plan containing task allocation logic; and a communication interface unit, used to send the task orchestration plan to the visualization rule-protocol conversion engine module.

[0011] By adopting the above technical solution, natural language instructions are automatically parsed into structured, assignable task plans, and a communication link with the rule engine is established, laying the foundation for the subsequent generation of final execution instructions by combining interactive rules.

[0012] Furthermore, the visual rule-protocol conversion engine module includes: a rule configuration interface unit, used to provide graphical interface elements and receive user configuration operations for speaking order mode, triggering conditions, or priority strategies; and a rule conversion and execution unit, connected to the rule configuration interface unit and the communication interface unit, used to map the interactive rule logic defined by the configuration operation into underlying communication control instructions that control the information transmission order or triggering logic between intelligent agents, and generate the underlying protocol in combination with the received task orchestration plan.

[0013] By adopting the above technical solution, the high-level collaborative strategies set by users through the graphical interface (such as who speaks first, and under what conditions to trigger) are automatically translated into specific control instructions that drive the interaction of intelligent agents within the system. This is a key technical bridge to achieve code-free orchestration.

[0014] Furthermore, the system also includes a low-level communication and scheduling module connected to the rule conversion and execution unit and the agent management unit; the low-level communication and scheduling module is used to drive each agent in the group collaboration container to communicate and execute tasks through an asynchronous communication mechanism according to the low-level protocol, and to feed back the task execution status to the visualization interaction unit.

[0015] By adopting the above technical solutions, a complete closed loop is formed from user instruction parsing and rule conversion to the final driving of intelligent agent execution and status feedback, ensuring the reliable operation and process visualization of the entire orchestration process.

[0016] Furthermore, the speaking order modes provided by the rule configuration interface unit include: equal speaking mode, polling mode, or response mode triggered by a specific output event.

[0017] By adopting the above technical solutions, users are provided with intuitive and optional typical collaborative strategies, covering common multi-agent interaction scenarios and reducing the complexity of rule configuration.

[0018] Furthermore, the predefined agent capability library on which the task orchestration engine unit performs task decomposition and agent matching is maintained by the agent management unit, which records the pre-encapsulated capability tags of each agent's functional modules.

[0019] By adopting the above technical solution, a data linkage between the agent management module and the task orchestration module was established, enabling task orchestration to accurately match based on the real-time capability description of the agent, thereby improving the accuracy and automation level of task allocation.

[0020] Furthermore, the system also includes a consensus extractor unit connected to the underlying communication and scheduling module; the consensus extractor unit is used to track and analyze the process information generated by the interaction of multiple agents in the group collaboration container according to the underlying protocol in real time, and output summary data generated based on the process information, the summary data containing interaction conclusion information.

[0021] By adopting the above technical solutions, the dynamic collaboration process of multiple agents can be monitored and analyzed, key conclusions and consensus can be automatically extracted, and the manageability and usability of the output results can be enhanced.

[0022] Furthermore, the consensus extractor unit is also used to attach time metric labels to the summary data or the task orchestration plan.

[0023] By adopting the above technical solutions, a timeliness management dimension is introduced into the collaboration process and task planning, which helps to identify and process outdated information and meets the "preservation" requirements in the context of rapid knowledge updates.

[0024] Furthermore, the group collaboration container building module can dynamically build or modify group collaboration containers in response to user voice commands or selection of predefined container templates.

[0025] By adopting the above technical solutions, the no-code operation method has been expanded, providing more convenient ways to create groups, such as voice interaction and template reuse, thereby improving the flexibility and efficiency of user interaction.

[0026] In summary, this application has at least the following beneficial effects:

[0027] A no-code orchestration system based on intelligent agents is provided, enabling non-technical personnel to easily build and command multi-agent collaborative systems through visualization and natural language.

[0028] By automatically converting graphical interaction rules into underlying communication protocols, the decoupling of high-level business logic and underlying technical implementation is achieved;

[0029] By introducing consensus extraction and time-stamping functions, the management of dynamic collaboration processes and the ability to preserve knowledge have been enhanced.

[0030] It should be understood that the description in the Summary Section is not intended to limit the key or essential features of the embodiments of this application, nor is it intended to restrict the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0031] The above and other features, advantages, and aspects of the embodiments of this application will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0032] Figure 1 A schematic diagram of an agent-based no-code orchestration system according to an embodiment of this application is shown. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0035] This application provides a no-code orchestration system based on intelligent agents. Through visual group building, dialogue-driven tasks, and graphical rule-making, it enables non-technical personnel to intuitively and efficiently organize and schedule multiple autonomous intelligent agents to work collaboratively. This significantly reduces the threshold for building and using multi-agent systems and improves the orchestration efficiency of complex intelligent tasks.

[0036] Figure 1 A schematic diagram of an agent-based no-code orchestration system according to an embodiment of this application is shown.

[0037] Reference Figure 1 The agent-based no-code orchestration system provided in this application can be implemented through various software or hardware / software combinations. The following description aims to clearly illustrate the implementation principles and various feasible methods of its technical solution, and is not intended to limit the scope of protection.

[0038] This system aims to provide a no-code environment for intelligent agent collaboration. Within this environment, the system's architecture allows users to organize multiple intelligent agents, assign tasks, and set collaboration rules through non-programming interaction. The system typically manifests as a software application or service platform executed by one or more computing devices, its core consisting of several functionally interconnected components.

[0039] The system includes a component for providing a user interface and building collaborative organizations. This component responds to user actions—such as visualization, voice, or selecting predefined templates—to dynamically create or adjust a logical workspace. This workspace accommodates and defines a group of agents that will work collaboratively. User interaction with this component is intuitive. For example, in a graphical user interface, users can drag and drop agent icons representing different functions (such as text analysis, data querying, image generation, etc.) into or out of a specific visual area, which represents the workspace. Besides dragging and dropping, users can also trigger workspace creation via voice commands, such as "create a customer service group," or by directly clicking a pre-configured template called "project review." During this process, users can also configure attributes for agents added to the workspace through a configuration panel on the interface, such as associating with an external knowledge base file or writing system prompts to guide their behavior. This component internally includes an interface rendering component and an agent management component. The interface rendering component is responsible for rendering the graphical interface and capturing the user's actions mentioned above. The agent management component is responsible for registering the agent functional units connected to the system, maintaining their configuration attributes (such as updating the knowledge base path recently associated with the user), and tracking their operational status. The agent management component maintains a capability list, which records the functional tags claimed or provided by each registered agent, such as "translation," "summarization," and "computation." This list provides other parts of the system with the basis for finding and scheduling agents by capability.

[0040] The system also includes a component for understanding user intent and planning tasks. This component receives natural language instructions input by the user within the aforementioned workspace and automatically transforms them into a set of specific action plans that can be executed by agents within the space. At its core is a task parsing and planning component that uses natural language processing technology to analyze user instructions. For example, when a user inputs "Analyze last week's sales data and create a chart report," the component identifies two core sub-intents: "data analysis" and "chart generation." It then queries a capability list maintained by the agent management component, searching for agents whose function tags match. For example, it might find an agent A tagged "data analysis" and an agent B tagged "chart generation." Next, it orchestrates an execution sequence based on the task logic (usually requiring data analysis before chart generation), forming a structured task orchestration plan that specifies the execution steps, participating agents, and data flow. Finally, a communication interface component outputs this plan and passes it to subsequent stages of the system. In this process, the confidence level of intent recognition is an adjustable parameter, and its threshold can be set between, for example, 0.65, 0.8 or 0.9 depending on the application scenario. A typical preferred value is 0.75 to balance accuracy and flexibility.

[0041] The system further includes a section for defining and transforming collaboration rules. This section provides users with a graphical way to set rules for how agents interact and translates these high-level rules into internal system execution instructions. It includes a rule configuration interface component, presented to the user in the form of graphical controls (such as drop-down menus, radio buttons, or visual logic block diagrams). Users can use these controls to select or define interaction modes for the workspace, such as setting an "equal discussion" mode where all agents can speak freely, a "polling" mode where actions are executed in a fixed order, or defining an event response rule that "if agent A's output contains the keyword 'urgent,' immediately notify the administrator." Once the user has completed the configuration, a rule transformation and execution component begins operation. This component receives user settings from the rule configuration interface and task orchestration plans from the task parsing and planning section. Its responsibility is to translate the user-defined, business-oriented interaction rules into control commands required for the system's underlying scheduling. For example, it translates the user-selected "polling mode" into a set of instructions that specify the strict order and triggering conditions for message passing. Then, it integrates the translated control commands with the task orchestration plan to generate a final, directly executable underlying collaboration protocol.

[0042] To execute the aforementioned protocol, the system typically includes a component responsible for runtime scheduling and communication. This component coordinates and drives the various agents within the workspace to work according to the generated underlying cooperation protocol. It manages information exchange between agents through asynchronous communication mechanisms, such as a message queue-based architecture. Specifically, this component sends task start instructions and input data to designated agents according to the protocol. It manages a communication channel to which agents output their task results. Based on rules defined in the protocol (such as polling order), this component controls which agent can consume information from the channel or send new information at what time. Simultaneously, it monitors the task execution status of each agent and provides real-time feedback of this status information (such as "in progress" or "completed") to the user interface, allowing the user to understand the progress. The asynchronous communication mechanism used here is key to achieving decoupling and flexible scheduling; its specific form can be a message queue, a publish-subscribe system, or other event-driven architecture.

[0043] A more sophisticated system implementation could include a component for refining collaborative outcomes and managing their validity. This component monitors in real-time the flow of information generated during interactions between all agents within the workspace, including dialogues, intermediate results, and final outputs. By analyzing and processing these information flows—for example, using text summarization, key information extraction, or consensus detection algorithms—it can automatically generate a summary report reflecting the core conclusions of the collaborative work. Furthermore, to address the issue of information becoming outdated, this component or other relevant system components can attach a timestamp or expiration date tag to the generated summary report, or even to the task scheduling plan itself. This expiration date tag, or lifespan, can be set to different values ​​depending on the type of information. For example, for stock price information, the expiration date might be as short as a few seconds or minutes (e.g., 30 seconds, 5 minutes, 1 hour, preferably 5 minutes); for market analysis reports, the expiration date might be several days (e.g., 1 day, 7 days, 30 days, preferably 7 days); and for company policy documents, the expiration date might be as long as several months (e.g., 30 days, 90 days, 180 days, preferably 90 days). When this information is used subsequently, the system can use the label to determine whether it is still valid.

[0044] In summary, the system described in this embodiment, through the coordinated operation of the aforementioned components, achieves the entire process from user no-code command input to automated collaborative execution by multiple agents. Specifically, the user interaction and organization building component provides a low-barrier entry point; the intent understanding and task planning component transforms vague requirements into clear steps; the rule definition and transformation component empowers users to control the collaborative process; the scheduling and communication component ensures reliable execution of the plan; and the results extraction component enhances the value and usability of the collaborative output. These components are interconnected through data interfaces and communication links, forming an organic whole that collectively achieves the technical objective of lowering the barriers to building and using multi-agent systems. Those skilled in the art will understand that the specific implementation technologies of the aforementioned components, such as natural language processing models for intent recognition, state machines or compilers for rule transformation, and message middleware selection for asynchronous communication, all have various mature existing technologies available for selection and combination, which does not deviate from the scope of the core concept of this invention.

[0045] Building upon the closed-loop process from no-code user operation to multi-agent collaborative execution achieved in the aforementioned embodiments, this system can further integrate a deep "collaborative cognition and evolution kernel." This kernel is not an independent add-on module, but rather an algorithmic framework that runs through and enhances the core logic of the system. Its fundamental goal is to enable the system to possess advanced cognitive capabilities for internal (collaborative dynamics) modeling and understanding, and external (new tasks and new environments) creative adaptation and evolution.

[0046] The core of this kernel lies in constructing a model capable of providing a continuous and unified mathematical representation of the cooperative state of a group of intelligent agents, called the "Unified Representation and Algorithm Space for Cooperative States." This space aims to precisely describe the core elements and their dynamic relationships in the cooperative process in a quantitative manner. At any given time... The state is represented by a tuple vector It is indicated that it is defined as:

[0047]

[0048] in, Let be the belief state tensor. Its elements are... Indicates at time , No. The agent in task number The data represents the strength and direction of opinions on each dimension or sub-topic. For example, a value close to 1 indicates strong agreement or a positive judgment, close to -1 indicates strong opposition or a negative judgment, and 0 indicates neutrality or no opinion formed. This data comes from the consensus extractor unit's real-time sentiment and stance analysis of the agent's output. This is the interaction relation hypermatrix. Its elements... Quantified in terms of issue dimension Above, intelligent agent The viewpoint is influenced by the intelligent agent The degree of influence or trust weight. This relationship data is derived by analyzing message response patterns, referencing relationships, and historical collaboration performance statistics in the underlying communication and scheduling modules. The knowledge context embedding vector is a fixed-dimensional semantic vector mapped from external knowledge base entries, data sources, and the agent's internal memory activated or referenced in the current collaboration, using a pre-trained language model (such as a Transformer encoder). This vector originates from knowledge base association records maintained by the agent management unit and data access logs during task execution. Encoding high-level intents and rule policies, it is a vector that combines discrete and continuous elements, encoding the original intent label parsed by the task orchestration engine unit for the current task, as well as the key features of the underlying protocol that is in effect and generated by the visualization rule-protocol conversion engine module.

[0049] Based on this unified representation, the system runs a "state dynamics learning engine". The goal of this engine is to learn a parameterized state transition function. This is used to simulate and predict the evolution of cooperative states. Its mathematical form is:

[0050]

[0051] in, Indicates that the system is in The actions taken at any given moment specifically refer to interventions implemented by the visualization rule-protocol conversion engine module or by the user through the visualization interaction unit, such as introducing a new data source. (Corresponding update) Alternatively, switch the interaction rule to "polling mode" (corresponding to the update). (Regular code section). It represents environmental context parameters, including fixed information such as task type, time constraints, and the set of available agents. It is a set of parameters for a neural network, which typically employs a hybrid architecture of graph neural networks and temporal models (such as LSTM or Transformer) to simultaneously capture the structured relationships and temporal dependencies between agents. This engine is trained on massive amounts of historical collaborative data, with the goal of improving the prediction state. The goal is to minimize the difference between the observed subsequent states and the actual values. After training, the engine's function is to predict and infer: given the current cooperative state... and a hypothetical intervention action The engine is able to calculate the expected next state. This allows us to derive macro-level indicators of collaborative effectiveness. Examples include consensus convergence speed (the rate at which opinion variance decreases) or task completion prediction. This enables the system to perform quantitative "scenario simulations" of different management or intervention strategies.

[0052] To go beyond correlation prediction and gain a deeper understanding of the causal mechanisms of collaborative dynamics, the system further deploys a "causal counterfactual explorer." This explorer uses observed cooperative state sequences... Based on this, a causal discovery algorithm based on score matching or gradient search is used to identify the potential causal structure between state variables. Its output is a directed acyclic graph. The node is Key variables in the process (such as a specific belief dimension of an agent) or a certain relation weight The edge represents the direction of causal influence, and the edge weight is... This represents the strength of the causal effect. Based on the learned causal graph. The system can perform counterfactual queries. For a previously occurred event with a result quality of [missing information], [missing information]. For collaborative instances, the system can calculate counterfactual quantities:

[0053]

[0054] in, This indicates the change in the variable at a certain historical moment. (For example, whether or not) Inject key knowledge at all times Implement intervention (set its value) After that, by the causal model The derived expected results. This calculation involves operations on the conditional probability distribution in the causal graph. This capability enables the system to answer diagnostic questions such as, "If a certain report were provided midway through the discussion, by what percentage would the final consensus quality improve?", providing fundamental insights for retrospective analysis and strategy optimization.

[0055] In terms of creative strategy generation, the system embeds a "meta-collaborative strategy generation field." This module transforms the problem of finding the optimal collaborative strategy into an optimization search on a manifold of a unified representation space. Its core is a meta-learning framework that does not directly output specific rules but instead learns a policy generation function. The input to this function is the current novel cooperative state. and mission objectives The output is a new collaborative strategy tailored to this scenario. The function first has a policy encoding layer, which encodes all basic policies in the system's rule base (such as "equal speaking", "round-robin", and "chairmanship") into a series of basis vectors. Then, an attention network computes the importance weights for each basis based on the current input:

[0056]

[0057] in, It is a learnable similarity function. Ultimately, the generation strategy is a weighted combination of these basis vectors, which is then refined and specified by a decoder network.

[0058]

[0059] This is how it was generated. It could be a completely new interaction protocol that combines the advantages of multiple basic strategies and adapts to specific situations, such as "initially using equal speaking to brainstorm, automatically switching to class representative polling when clustering of viewpoints is detected, and introducing external knowledge for final decision-making in the final stage".

[0060] To facilitate the transfer of cross-domain intelligence, a "semantic topology reconstructor" was designed. This component posits that successful collaboration patterns across different business domains manifest as different data manifolds within a "cooperative unified representation space." The reconstructor aims to transform efficient cooperative trajectories within a domain (source domain, such as "product design review") into a learnable nonlinear mapping. This can be transformed onto the manifold of another domain (the target domain, such as "investment risk assessment") to generate virtual trajectories that may be equally efficient in the target domain. This is achieved through an adversarial variational autoencoder. Encoder Source domain state Encode into a domain-independent latent space vector decoder Then try to from Reconstruct the state of the target domain At the same time, a discriminator It is trained to distinguish between the true target domain state and the reconstructed state. The overall optimization objective function of the model is:

[0061]

[0062] in These are weighted parameters that balance reconstruction errors with adversarial losses. After successful training, the system can transform dynamic patterns in the source domain that inspire creativity or improve efficiency into novel collaborative process suggestions applicable to the target domain, achieving fundamental innovation transfer.

[0063] Ultimately, all the knowledge generated by the above learning and generation processes—including the state transition model—is... Cause-and-effect diagram Strategy generation function and cross-domain mapping functions All of these are systematically integrated into a dynamic "cognitive metagraph." This graph is organized in the form of a knowledge graph, with nodes representing algorithmic models, strategy patterns, causal mechanisms, and performance indicators, and edges representing the derivative, complementary, contradictory, or combinable relationships between them. Through continuous graph embedding updates and reasoning, the system can perform high-level knowledge retrieval and fusion, such as automatically answering questions like: "When facing highly uncertain and time-sensitive decision-making tasks, based on historical causal analysis, which strategy generation patterns and which knowledge injection methods have most effectively improved decision robustness?" This constantly evolving cognitive metagraph constitutes the intelligent core of the system's self-understanding, self-evaluation, and strategic self-improvement.

[0064] By integrating this series of deeply collaborative algorithmic kernels, this system, building upon its original automated orchestration capabilities, has acquired fundamental abilities for quantitative modeling of collaborative processes, causal inference, creative strategy generation, and cross-domain intelligent transfer. These capabilities enable the system not only to execute user-defined collaborative tasks but also to proactively understand the key factors of collaborative effectiveness, design better collaborative solutions, and continuously evolve its collaborative intelligence from experience, ultimately forming an intelligent collaborative organism capable of adapting to complex and ever-changing needs and possessing endogenous evolutionary capabilities.

[0065] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. An agent-based no-code orchestration system, characterized in that, include: Group collaboration container building module, dialogue-driven intelligent agent collaboration network module, and visualization rule-protocol conversion engine module; The group collaboration container building module is used to dynamically build or modify a group collaboration container for accommodating multiple agents in response to user no-code operations. The no-code operations include dragging and dropping pre-packaged agent function modules into or out of the group collaboration container through a visual interface. The dialogue-driven agent collaboration network module is connected to the group collaboration container construction module and is used to receive natural language instructions input to the group collaboration container. By performing semantic understanding and task decomposition on the natural language instructions, a task orchestration plan is generated for collaborative execution by multiple agents in the group collaboration container. The visualization rule-protocol conversion engine module is connected to the group collaboration container construction module and the dialogue-driven agent collaboration network module. It is used to provide a visualization interface to receive the interaction rules configured for the group collaboration container and convert the interaction rules into a low-level protocol that controls the communication and scheduling order between agents in the group collaboration container, so as to drive the agent interaction based on the task orchestration plan and the low-level protocol.

2. The system of claim 1, wherein, The group collaboration container construction module includes: A visual interaction unit is used to provide a graphical interface to receive drag-and-drop operations of intelligent agent functional modules and attribute configuration operations of intelligent agents within the group collaboration container. The attribute configuration includes associating with a knowledge base or setting system prompt words. The intelligent agent management unit is connected to the visual interaction unit and is used to encapsulate and register the intelligent agent functional modules, maintain their attributes, and monitor their operating status.

3. The system of claim 2, wherein, The dialogue-driven agent collaboration network module includes: The task orchestration engine unit is used to perform intent recognition on the received natural language instructions, and to perform task decomposition and agent matching based on the recognized intent and the predefined agent capability library to form a task orchestration plan containing task allocation logic. The communication interface unit is used to send the task orchestration plan to the visualization rule-protocol conversion engine module.

4. The system of claim 3, wherein, The visualization rule-protocol conversion engine module includes: The rule configuration interface unit is used to provide graphical interface elements and receive user configuration operations for speaking order mode, trigger conditions or priority strategies. The rule conversion and execution unit, connected to the rule configuration interface unit and the communication interface unit, is used to map the interactive rule logic defined by the configuration operation into underlying communication control instructions that control the information transmission order or triggering logic between intelligent agents, and generate the underlying protocol in combination with the received task orchestration plan.

5. The system of claim 4, wherein, The system also includes an underlying communication and scheduling module, which is connected to the rule conversion and execution unit and the agent management unit; The underlying communication and scheduling module is used to drive each intelligent agent in the group collaboration container to communicate and execute tasks through an asynchronous communication mechanism according to the underlying protocol, and to feed back the task execution status to the visualization interaction unit.

6. The system of claim 4, wherein, The speaking order modes provided by the rule configuration interface unit include: equal speaking mode, polling mode, or response mode triggered by a specific output event.

7. The system of claim 3, wherein, The predefined agent capability library on which the task orchestration engine unit performs task decomposition and agent matching is maintained by the agent management unit, and records the capability tags pre-packaged for each agent's functional modules.

8. The system of claim 5, wherein, The system also includes a consensus extractor unit, which is connected to the underlying communication and scheduling module; The consensus extractor unit is used to track and analyze the process information generated by the interaction of multiple agents in the group collaboration container according to the underlying protocol in real time, and output summary data generated based on the process information, wherein the summary data includes interaction conclusion information.

9. The system of claim 8, wherein, The consensus extractor unit is also used to attach time metric labels to the summary data or the task orchestration plan.

10. The system according to claim 1, characterized in that, The group collaboration container building module can dynamically build or modify group collaboration containers in a way that responds to user voice commands or selection of predefined container templates.

Images