Zero-code agent construction platform

Abstract

The invention belongs to the technical field of computers, and particularly relates to a zero-code agent construction platform. The system comprises a module database, a semantic analysis and intention decomposition unit, a task flow tree generation unit and an agent construction unit, the semantic analysis and intention decomposition unit is used for performing image language analysis on a natural language instruction input by a user, decomposing the natural language instruction into a plurality of intention units with a sequence, and constructing an intention vector; the task flow tree generation unit is used for generating a plurality of corresponding task flow trees according to the intention vector; and the agent construction unit is used for analyzing all the task flow trees, finding out an optimal task flow tree, calling all target function modules in the task flow tree and completing the construction of the agent. According to the method, the precision of semantic analysis, the flexibility of task logic generation and the automation degree of agent construction are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and specifically relates to a zero-code intelligent agent construction platform. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, intelligent task execution and human-computer interaction technology are increasingly being used in all walks of life. However, the construction of most intelligent systems currently still requires professional programming skills, which creates a high threshold for users with non-technical backgrounds. In order to lower this technical threshold, zero-code development platforms have gradually gained attention and promotion. Zero-code platforms aim to help users develop and deploy complex tasks without writing code through intuitive user interfaces, modular task construction, and visual logic design. However, existing technologies still have deficiencies in certain key links of zero-code intelligent agent construction, and cannot fully meet users' intelligent needs for diversified and dynamic tasks.

[0003] Most zero-code development platforms only support structured text input or pass task requirements through interface operations, and cannot effectively handle unstructured or complex input. For example, for handwritten images containing natural language instructions, existing technologies often lack effective parsing capabilities and can only recognize characters through OCR, and then rely on static rules to map task logic. This approach not only has limited parsing accuracy, but also cannot adapt to multimodal input scenarios (such as those containing both images and text). Current mainstream platforms rely on simple keyword matching or predefined rules to achieve semantic parsing, rather than dynamic analysis based on deep learning or high-dimensional geometric features. For example, when users enter multi-level complex task instructions (such as "analyze sales data to generate a bar chart, and send the results to management by email"), traditional platforms often find it difficult to accurately decompose them into independent intent units. This causes users to need to perform additional disassembly of the task, losing the simplicity of zero-code development. In existing technologies, task logic is usually implemented through fixed process definitions or templated module calls, which is difficult to dynamically adjust or optimize. For example, the optimal path selection of the task process is done manually on most platforms, lacking intelligent solutions based on computational optimization. When users are faced with high-concurrency tasks or resource-constrained scenarios, existing platforms are unable to provide sufficient task flexibility. Although some platforms provide storage and reuse capabilities for functional modules, existing technologies still have shortcomings in terms of dynamic updates of modules, redundant call management, and adjustment of logical relationships between modules. For example, the addition or deletion of modules usually requires system downtime for maintenance, or relies on complex manual operations, which makes it difficult to meet real-time adjustments to task requirements. In addition, there is a lack of effective redundant management mechanisms for multiple versions of the same functional module, which can easily lead to system instability. Summary of the invention

[0004] The main purpose of the present invention is to provide a zero-code intelligent agent construction platform, which significantly improves the accuracy of semantic parsing, the flexibility of task logic generation, and the degree of automation of intelligent agent construction. Its beneficial effects include supporting efficient parsing of multimodal input, dynamically optimizing task process trees, enhancing module management and expansion capabilities, improving the fault tolerance and automation level of intelligent agent construction, greatly reducing the user development threshold, and being widely applicable to complex and diverse intelligent task scenarios.

[0005] In order to solve the above technical problems, the present invention provides a zero-code intelligent agent construction platform, which includes: a module database, a semantic parsing and intention decomposition unit, a task process tree generation unit and an intelligent agent construction unit; the module database stores multiple functional modules that perform a single function at fixed positions, and each functional module is connected to the adjacent stored functional modules; there are at least N functional modules with the same function, and N is a set value; the semantic parsing and intention decomposition unit is used to perform image language analysis on the natural language instructions input by the user, decompose the natural language instructions into multiple intention units with a sequence, take each intention unit as an element in a vector, and construct an intention vector according to the sequence of the intention units; the task process tree generation unit is used to generate corresponding multiple task process trees according to the intention vector; the intelligent agent construction unit is used to analyze all task process trees, find the optimal task process tree, call all target function modules in the task process tree according to the sequence of the target function modules in the optimal task process tree, and complete the construction of the intelligent agent.

[0006] Furthermore, the updating process of the module database specifically includes: if any functional module in the module database is to be deleted, the functional module is directly put into permanent hibernation; if a functional module is to be added to the module database, the functional module is directly added and stored in any free position in the module database, and connected with other adjacent functional modules.

[0007] Furthermore, the process of generating any task flow tree by the task flow tree generation unit specifically includes: finding all the functional modules corresponding to each intention unit in the intention vector in the module database as the target functional modules of each intention unit; each functional module includes a keyword set for matching the intention unit; in the keyword set, each keyword is represented by a feature code, and each intention unit is also a code; for each intention unit, all the corresponding target functional modules are grouped into a set as the target functional module set of the intention unit; arbitrarily select a target functional module from each target functional module set and connect it with the target functional modules in other target functional module sets in the order of the intention units to generate a corresponding target task tree; during the connection process, if the functional module adjacent to the target functional module is not the target functional module, the functional module is put into dormancy and grayed out.

[0008] Furthermore, the agent construction unit analyzes all task flow trees, and the process of finding the optimal task flow tree includes: taking the task flow tree with the least dormant grayed-out functional modules as the optimal task flow tree.

[0009] Furthermore, the semantic parsing and intention decomposition unit performs image language parsing on the natural language instructions input by the user, and the method of decomposing the natural language instructions into multiple intention units with a sequence specifically includes: obtaining an image containing natural language instructions, analyzing the fractal geometric feature vector of the image, and constructing a multi-fractal spectrum; based on the multi-fractal spectrum, constructing a dynamic topological structure vector of the strokes; constructing a high-dimensional feature space vector based on the multi-fractal spectrum and the dynamic topological structure vector; mapping the high-dimensional feature space vector into the discriminant manifold space to obtain a discriminant manifold vector; and performing a prior probability-based transformation on each element in the discriminant manifold vector to obtain a corresponding intention unit.

[0010] Furthermore, the following formula is used to obtain an image containing natural language instructions, analyze the fractal geometric feature vector of the image, and construct a multifractal spectrum: ; in, is the integral area, which represents the area of ​​the image; is the fractal sequence parameter, is the set value; is the time scale parameter, is the set value; is the space curvature parameter, is the set value; For the The size of the observation window of each scale; For the input image Gray value of the position; is the coordinate of the X axis; is the coordinate of the Y axis; is an integer subscript index; is the scale number; It is a multi-typing spectrum.

[0011] Furthermore, the dynamic topological structure vector of the strokes is constructed based on the multifractal spectrum through the following formula: : ; in, To make The largest value; is the highest curvature order; is an integer subscript index; for The gradient of is the imaginary number symbol; For the The curvature characteristic scale of the order is a set value ranging from 0.5 to 0.9; is the regional integral differential; is the characteristic angle, ranging from 10 to 30 degrees.

[0012] Furthermore, a high-dimensional feature space vector is constructed based on the multifractal spectrum and the dynamic topological structure vector through the following formula: ; in, is a high-dimensional feature space vector; is the coordinate of the first coordinate axis in the high-dimensional feature space; is the coordinate of the second coordinate axis in the high-dimensional feature space; is the resolution parameter of the high-dimensional feature space; for Bessel function of order.

[0013] Furthermore, the high-dimensional feature space vector is mapped to the discriminant manifold space through the following formula to obtain the discriminant manifold vector: ; in, is the harmonic function in the discriminant manifold space; is the F-norm operation; is the absolute value operator; is the manifold coordinate system Quantity; is the manifold coordinate system The characteristic scale of the component is the set value.

[0014] Furthermore, each element in the discriminant manifold vector is transformed based on the prior probability through the following formula to obtain the corresponding intention unit: ; in, For the Intention unit; The mean and variance are The prior probability of the Gaussian distribution of ; for The gradient of is the first vector in the discriminant manifold elements.

[0015] The zero-code intelligent agent construction platform of the present invention has the following beneficial effects: The present invention breaks through the traditional zero-code platform's reliance on structured input and realizes the direct parsing of natural language images for the first time. Through the image parsing method based on fractal geometry, the present invention can extract geometric and semantic features from handwritten instructions, fuzzy images and even complex unstructured texts, providing powerful technical support for diversified input scenarios. Using multiple fractal spectra and dynamic topological structure vectors, the present invention can accurately capture the local geometric characteristics and global distribution laws in the image, and realize efficient parsing of complex natural language instruction images. This capability greatly expands the scope of application of the zero-code platform, giving it significant advantages in multimodal input scenarios and meeting the needs of users in diversified business scenarios.

[0016] The present invention greatly improves the accuracy and robustness of natural language instruction parsing by constructing a high-dimensional feature space vector and mapping it to a discriminative manifold space. Unlike traditional keyword matching and static rule methods, the present invention utilizes dynamic topological structures and high-dimensional feature analysis techniques to perform in-depth analysis of complex, multi-level natural language instructions. For example, for complex task instructions containing multi-step logic, the present invention can accurately decompose them into intent units with a sequence, and optimize the parsing results through a priori probability models. This semantic parsing method based on high-dimensional features and geometric optimization not only improves the accuracy of the parsing results, but also significantly enhances the system's adaptability to fuzzy expressions, handwritten input, and irregular instructions.

[0017] Traditional zero-code platforms usually rely on static rules to generate task logic, and it is difficult to dynamically adjust the process path. The present invention realizes dynamic optimization of task logic through a task process tree generation unit. By analyzing the intention vector and the module database, the present invention can generate multiple candidate task process trees, and select the optimal task process tree based on the principle of the least dormant grayed-out modules. This dynamic optimization method can significantly improve the execution efficiency of the process tree when resources are limited or the task complexity is high. In addition, the present invention introduces a dynamic adjustment mechanism in the task logic generation process, which can update the process path in real time according to changes in the task environment (such as module call failures or resource allocation problems). BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0019] Figure 1 A schematic diagram of the system structure of a power grid primary frequency regulation intelligent control system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The method of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments of the present invention.

[0021] Example 1, reference Figure 1 : A zero-code intelligent agent construction platform, which includes: a module database, a semantic parsing and intention decomposition unit, a task flow tree generation unit and an intelligent agent construction unit; the module database stores multiple functional modules that perform a single function at fixed positions, and each functional module is connected to the adjacent stored functional modules; there are at least N functional modules with the same function, and N is a set value; the semantic parsing and intention decomposition unit is used to perform image language analysis on the natural language instructions input by the user, decompose the natural language instructions into multiple intention units with a sequence, take each intention unit as an element in a vector, and construct an intention vector according to the sequence of the intention units; the task flow tree generation unit is used to generate multiple corresponding task flow trees according to the intention vector; the intelligent agent construction unit is used to analyze all task flow trees, find the optimal task flow tree, call all target function modules in the task flow tree according to the sequence of the target function modules in the optimal task flow tree, and complete the construction of the intelligent agent.

[0022] Specifically, the principle of the module database is based on the modularization, redundant storage and structured connection of functional modules to achieve efficient module calling and task execution. Each functional module is an independent unit that implements a single function, such as data input, information processing, result output, etc. The module database uses a fixed-position storage method to arrange and connect these functional modules according to logical associations, so that the module calling path is deterministic and predictable. The advantage of this storage method is that when the intelligent agent construction unit accesses the module database according to the task process tree, it can quickly find the target functional module without additional index calculations or complex path searches, thereby greatly reducing the delay in task execution.

[0023] In addition, in order to improve the stability of the system, each functional module in the module database has multiple redundant versions. The multi-version design mainly considers the module call failure or resource competition problems that may exist in actual application scenarios. Through the introduction of redundant modules, the system can dynamically switch to the alternative module when the main module cannot be called normally, so as to achieve seamless recovery of the task. Taking the construction of an intelligent agent as an example, if a task process tree needs to call the "semantic analysis module", but the main module fails due to resource occupation or calculation error, the system can immediately select other redundant modules to replace it to ensure the continuity of the task. This design greatly improves the fault tolerance and robustness of the platform, so that the entire system can still maintain stable operation when facing high-concurrency tasks or sudden failures. There are structured connections between the functional modules in the module database. This connection is not random, but is designed based on the relevance of the module functions and the task logic. The direct connection between adjacent modules not only provides an efficient calling path for the task process tree, but also provides the possibility for dynamic adjustment of the task process. For example, when the task process tree analysis finds that the execution time of a module call is long or the computing load is too high, the system can re-plan the process through the module connection path and select another more efficient path to complete the task. This flexibility comes from the tight coupling between modules in the module database and the efficient data interaction mechanism. Unlike traditional modular design, the module database also plays the role of implicit knowledge in the zero-code intelligent agent construction platform, that is, the module database is not only a collection of functions, but also an embedding point for task process rules. For example, each functional module contains its input and output specifications and interface definitions with other modules when it is stored. This standardized design makes the combination between modules more concise and clear, thereby realizing the dynamic generation of intelligent agents. Users only need to provide high-level natural language instructions, and the module database can quickly determine the functional modules required for the task through the guidance of semantic parsing and intention decomposition units, and complete the call in a logical order, avoiding the complex code writing and module configuration process in traditional development. Another key function of the module database is to support dynamic expansion and hot update. As business needs change, new functional modules can be directly added to the database without interrupting the operation of the existing system. This dynamic expansion capability enables the zero-code intelligent agent construction platform to quickly adapt to new scenarios and new needs, and maintain its competitiveness in technological evolution. At the same time, the hot update function of the module database ensures that the maintenance and optimization of the module can be completed during operation. For example, when the algorithm of a certain functional module is improved, it only needs to be updated to the module database, and the system can automatically identify and prioritize the latest version, further improving the intelligence level of the platform.

[0024] When using the platform, users only need to input a natural language, such as "analyze this set of data and generate a bar chart". This unit will first perform grammatical and semantic analysis on the instruction, understand its true meaning in combination with the context, and gradually decompose it into a series of intention units with a clear logical order. The core of semantic analysis is to understand human language in a modeling way, including lexical analysis, syntax tree construction, semantic rule mapping, and context dependency analysis. Through this process, the system can identify the main actions, target objects, and constraints contained in the user's instructions, such as "analysis" as an action, data as a target object, and "generate a bar chart" as a subsequent specific goal. After completing the semantic analysis, the intention decomposition unit will further decompose the parsed instructions into a group of intention units according to their logical hierarchy. Each intention unit represents an independent execution step in the instruction, with a clear execution target and sequence. For example, "analyze this set of data and generate a bar chart" can be decomposed into two intention units: "data analysis" and "generate a bar chart". There is a sequential relationship between these two units, that is, data analysis needs to be completed before the bar chart is generated. To achieve this goal, the intent decomposition unit uses pre-trained models (such as BERT or GPT) combined with task context to convert instructions into a high-dimensional vector representation. This vector representation is called an intent vector and is an important input for the system to generate subsequent task processes. Through the construction of intent vectors, the semantic parsing and intent decomposition unit maps complex natural language questions into structured data that can be processed by machines, thereby completing the conversion of user language into system task logic.

[0025] It is worth noting that when processing user instructions, the semantic parsing and intention decomposition unit not only focuses on the surface meaning of a single instruction, but also needs to reason and complete ambiguous or omitted expressions in combination with the context. For example, the user may enter "analyze data first, then draw a picture", and the specific data set and chart type are omitted in the instruction. In this case, the system needs to identify the implicit information of the user's instruction through contextual reasoning, such as determining the specific data set involved in the current task, or judging the type of chart drawn by the user's historical operation habits. After completing this reasoning, the system automatically completes the missing part so that the generated intent vector can fully describe the task logic. The principle of the semantic parsing and intention decomposition unit is also reflected in its multi-level decomposition ability for complex instructions. In actual applications, the natural language input by the user may not be a single task, but a complex logic with multiple levels and branches. For example, "analyze data set A, then compare the results with data set B, and generate a report" can be decomposed into multiple intent units with nested relationships, including "analyze data set A", "analyze data set B", "compare the two sets of results" and "generate a report". This decomposition requires the system to have the ability to understand logical dependencies and the ability to judge task priorities. The semantic parsing and intention decomposition unit realizes the decomposition and encoding of complex instructions through a deep learning model combined with task tree generation rules, ensuring that each intention unit is independent and operable. Furthermore, the implementation principle of the semantic parsing and intention decomposition unit also integrates the mechanism of real-time learning and dynamic adjustment. When the user enters a new instruction, the system will update the language model in real time to adapt to the new semantic pattern. For example, the user may habitually use certain specific terms or phrases, and these expressions may be different from the training corpus of the conventional language model. Through the real-time learning mechanism, the system can dynamically adjust the weight of the model to better adapt to the user's language style and improve the accuracy of semantic parsing. At the same time, when the parsing fails or the result does not meet the user's expectations, the intention decomposition unit can further optimize the task decomposition process through interaction with the user, such as by prompting the user to provide more details, or automatically recommending possible task decomposition methods.

[0026] The task flow tree generation unit is a key link in the zero-code agent construction platform. It converts the intent vector generated by the semantic parsing and intent decomposition unit into the logical structure of the agent task execution, providing a clear operation path for the construction of the agent. Its core principle is to use a graphical data structure, the task flow tree, to represent and organize the hierarchical relationship, dependency relationship and execution order between tasks, so as to achieve efficient task planning and execution. The workflow of the task flow tree generation unit begins with the parsing of intent vectors, which contain the task elements and their logical order in the user input instructions. The task flow tree generation unit maps each intent unit to a node in the tree structure by parsing the intent vector, and the connection between these nodes represents the dependency relationship of the intent unit. Through this mapping process, the system can formalize the user's natural language instructions into an executable task model.

[0027] The generation process of the task flow tree not only focuses on the surface logic of the intent unit, but also requires an in-depth understanding of the implicit dependencies between tasks. For example, when a user enters a complex instruction, such as "analyze sales data, predict future trends, and generate reports", the task flow tree generation unit will identify the hierarchy and order of these tasks: data analysis is a prerequisite for prediction, and reports can only be generated after the prediction is completed. This hierarchical relationship is the basis for the generation of the task flow tree, enabling the system to correctly organize the order of tasks. The task flow tree generation unit ensures the logical consistency of the generated flow tree through dependency graph modeling technology and task constraint analysis. Each task node not only contains the description of the intent unit, but also attaches the module information, input and output constraints, and possible branch paths required for its execution to support flexible task scheduling. In the process of generating the task flow tree, the unit will also comprehensively consider the parallelism and dependencies of the tasks to optimize the execution efficiency. For example, for independent tasks, if there is no dependency between them, the task flow tree generation unit can arrange them in parallel to save time costs. This parallelization capability is particularly suitable for scenarios where multiple modules collaborate, such as data processing and chart generation can be performed simultaneously without affecting each other. However, when there are strong dependencies between tasks, the task flow tree generation unit will strictly arrange the execution order according to the dependencies to avoid errors or failures caused by task conflicts.

[0028] The innovation of the task flow tree generation unit also lies in its optimization capability. In many cases, the user's natural language instructions may result in the generation of multiple task flow trees. The task flow tree generation unit will analyze and optimize these candidate trees to select the optimal execution path. This optimization process is based on multi-dimensional evaluation criteria, such as execution time, resource consumption, and reliability of module calls. Taking the "generate report" task as an example, the system may generate multiple implementation paths based on the available module versions, such as calling different chart generation modules or selecting different data processing methods. The task flow tree generation unit will evaluate the pros and cons of each path through cost function calculation and heuristic search algorithm, and select the task flow tree with the lowest cost and the highest execution efficiency as the final generation result. In addition, the task flow tree generation unit has dynamic adjustment capabilities. In actual operation, the execution environment of the task may change, such as module call failure or resource unavailability. At this time, the task flow tree generation unit can adjust the task flow tree in real time. By regenerating local subtrees or adjusting task paths, the system can ensure the stability and continuity of the entire process. This dynamic adjustment capability relies on the modular design of the task process tree generation unit, where each node and edge can be updated independently without affecting the structure of the entire tree. This flexibility enables the zero-code agent construction platform to cope with complex task environments and provide reliable task execution guarantees. Another important function of the task process tree generation unit is to support scalability and reusability. In a multi-task scenario, instructions from different users may involve similar tasks, such as data preprocessing or specific analysis logic. The task process tree generation unit can quickly generate standardized process trees for these common tasks through the mechanism of task templates, and personalize them according to the specific needs of users. This templated process tree generation method not only significantly improves the speed of task generation, but also reduces the workload of repeated modeling, reflecting the efficiency of the zero-code development platform.

[0029] In the process of agent construction, the unit first parses the node information of the task flow tree and identifies the specific functional modules that need to be called for each task node. Since the module database adopts a redundant storage design for functional modules, the agent construction unit does not simply call the main module directly when calling the module, but gives priority to the module version with the highest execution efficiency or the least resource consumption based on the module status and availability. For some high-concurrency scenarios, the agent construction unit can also assign multiple task nodes in the flow tree to different modules for parallel calls based on the parallelization characteristics of the modules, thereby significantly improving the efficiency of task execution. For example, in the task flow tree, if data preprocessing and preliminary analysis belong to two independent tasks, the agent construction unit can schedule the corresponding functional modules to execute these two tasks at the same time, and finally summarize the results to the subsequent nodes, thereby reducing the overall execution time of the task. Another key function of the agent construction unit is the dynamic management of task logic. In actual execution, the logical path described by the task flow tree is not static. Module calls may be interrupted due to insufficient resources, execution failure, or inconsistent input and output. At this time, the agent construction unit needs to monitor the status of module execution in real time and replan the task path according to the dynamic adjustment rules of the task flow tree. For example, when a function module corresponding to a node fails, the agent construction unit will quickly call the redundant module in the module database to complete the replacement operation, and update the status of the task process tree at the same time, so that the entire task process can continue to execute. This dynamic management capability relies on the real-time monitoring of the process tree by the agent construction unit and the efficient access to the module database, which reflects the high fault tolerance and robustness of the platform.

[0030] In order to further optimize the agent construction process, the unit also dynamically adjusts the module call strategy through real-time analysis of task execution. For example, for complex multi-stage tasks, the agent construction unit will optimize the module call order and allocation strategy based on the structure of the task flow tree and the execution performance of the task nodes. In some scenarios, the user's task requirements may require data flow across modules, such as from the data acquisition module to the analysis module, and then to the report generation module. The agent construction unit achieves efficient data sharing and transmission between modules by building a unified task context, avoiding execution delays caused by inconsistent data or repeated loading. Context management is a major technical highlight of the agent construction unit. By standardizing the input and output of each task node, the system can ensure data compatibility and operation continuity between module calls. The principle of the agent construction unit is also reflected in the automatic generation capability of the agent. Traditional agent construction usually requires users to define logical processes, call API interfaces, and write configuration files. In the zero-code agent construction platform, the agent construction unit completely hides these cumbersome operations from the user's view. Through the complete analysis of the task process tree and the seamless scheduling of modules, the system can complete the construction of the intelligent agent with almost no human intervention. Taking the user input "analyze financial data and generate trend charts" as an example, the intelligent agent construction unit will call the data loading module, data cleaning module, analysis module and chart generation module from the database in sequence according to the process tree, and finally output the results required by the user. In this process, the user does not need to understand the underlying implementation details, and the construction and operation of the entire intelligent agent are fully automated.

[0031] Embodiment 2: The updating process of the module database specifically includes: if any functional module in the module database is to be deleted, the functional module is directly put into permanent dormancy; if a functional module is to be added to the module database, the functional module is directly added and stored in any free position in the module database, and connected with other adjacent functional modules.

[0032] Specifically, when deleting a functional module, the system adopts a permanent dormancy strategy instead of direct deletion. This design idea fully considers the traceability of the module database and the fault tolerance of the system. Permanent dormancy means that the module marked for deletion will not be physically cleared, but will be logically disabled. The advantage of this approach is that if the subsequent task process tree still accidentally attempts to call the module, the system can detect and take remedial measures in time, such as automatic replacement through redundant modules to avoid task interruption caused by module deletion. At the same time, permanent dormancy also provides convenience for possible module recovery in the future. Users or system administrators can reactivate the module when needed without redevelopment or redeployment. In addition, the permanent dormancy mechanism has a low cost in technical implementation. It only needs to mark the module status as "dormant" or "disabled", and does not involve complex storage operations, thereby maintaining the stability and efficiency of the module database. When adding new functional modules, the design of the module database is more flexible. The new module is directly stored in any free position in the module database, avoiding the common index reconstruction or storage structure adjustment problems in traditional databases. This storage method reflects the decentralized design of the module database. The logical connection between functional modules is realized through the direct association of adjacent modules without relying on fixed positions or global indexes. This mechanism is very efficient in terms of technical implementation: when a new module is added, the system automatically detects the free position in the database, stores the module in that position, and connects it to the adjacent modules according to logical requirements. The connection method can be determined by the preset interface specifications and the dependencies between modules, so that the new module can be seamlessly integrated into the existing module network.

[0033] This dynamic insertion mechanism also provides technical support for the scalability of the module database. In practical applications, the zero-code agent construction platform may need to frequently update the module database to cope with the emergence of new scenarios or new requirements. For example, users may want to add a new data analysis module to support the latest algorithms or processing capabilities. At this time, the new module can be quickly inserted into the module database and is immediately available without shutting down or reconfiguring the system. Through the connection with adjacent modules, the new module can be quickly integrated into the generation and execution process of the task flow tree to ensure that the task scheduling capability and execution efficiency of the entire system are not affected. From the perspective of technical implementation, the storage and connection of new modules need to solve two core problems: one is the matching problem of the input and output interfaces of the module, and the other is the problem of resolving the dependency relationship between the new module and the existing module. The module database ensures that the interface of the new module is compatible with the adjacent modules through standardized interface definition and module metadata management. Each module records its functional description, input and output specifications, and possible dependencies when it is stored. When a new module is inserted, the system automatically resolves its dependency relationship with adjacent modules and generates connection rules so that data flow and function calls between modules can proceed smoothly.

[0034] Embodiment 3: The process of generating any task flow tree by the task flow tree generation unit specifically includes: finding all the functional modules corresponding to each intention unit in the intention vector in the module database as the target functional module of each intention unit; each functional module includes a keyword set for matching the intention unit; in the keyword set, each keyword is represented by a feature code, and each intention unit is also a code; for each intention unit, all the corresponding target functional modules are grouped into a set as the target functional module set of the intention unit; arbitrarily select a target functional module from each target functional module set and connect it with the target functional modules in other target functional module sets in the order of the intention units to generate a corresponding target task tree; during the connection process, if the functional module adjacent to the target functional module is not the target functional module, the functional module is put into dormancy and grayed out.

[0035] Specifically, in the process of generating the task flow tree, the first step is to parse each intent unit in the intent vector and find all the target function modules corresponding to them in the module database. Each function module contains a keyword set, which is an abstract description of the module function. The system identifies all modules that can implement the function of the intent unit by matching the intent unit code with the feature code in the module keyword set. This matching method is essentially a semantic mapping operation. The keyword set expresses the function of the module through feature coding, while the intent unit code reflects the requirements in the user's instruction. Through this matching, the system can quickly filter out the function modules that meet the requirements from the module database, thereby generating a target function module set for the intent unit. After the target function module set is determined, the system organizes the target function module set of each intent unit in sequence and gradually generates a task flow tree. The generation process of the task flow tree needs to consider the logical connection between modules and the order of intent units. For each intent unit, a function module is selected from the target function module set and connected to the target function modules of other intent units in the order of intent units. This design ensures that the task flow tree can faithfully reflect the logical structure of user instructions. At the same time, the flexibility in selecting the target functional module (i.e., “arbitrary selection”) provides the possibility of generating diverse task process trees. This diversity provides a basis for subsequent process tree optimization and selection of the optimal task path.

[0036] In the process of connecting the task flow tree, the system also needs to dynamically adjust the state of the function module to ensure the clarity of the process and the efficiency of execution. Specifically, if the function module adjacent to the target function module is not the target function module required by the current task, these irrelevant function modules will be marked as "dormant graying". "Drowsy graying" is a logical disabling operation, which means that these modules will not be activated or called in the current task flow, but they are still retained in the module database for use when subsequent tasks are required. This design not only ensures the simplicity of the task flow tree, but also avoids the problem of changing the module database structure or breaking the task dependency chain due to physical deletion of the module. This embodiment also has important technical features and advantages. First, by matching the module keyword set with the feature encoding of the intention unit, the system realizes the rapid mapping from the user's natural language instructions to the function module. This mapping mechanism can tolerate a certain degree of ambiguity or diversity in the user input, so that the platform has higher accuracy and robustness in understanding the user's intention. Secondly, the construction process of the target function module set supports redundant calls of modules. By selecting one from multiple alternative function modules for connection, the system realizes the flexibility of module selection and the diversification of task flows. Especially in the face of resource-constrained or high-concurrency scenarios, this flexibility provides more possibilities for system performance optimization. In addition, the dynamic adjustment capability of the task flow tree generation unit in module connection is also a highlight. When the non-target function module is dormant and grayed out, the structure of the task flow tree is simplified, avoiding resource waste or execution delay caused by unnecessary module calls. This dynamic adjustment capability not only improves the execution efficiency of the task flow tree, but also enhances the fault tolerance of the system. For example, if a certain target function module fails during execution, the system can select another module from the corresponding target function module set as a replacement to ensure the continuity of the task flow. From the perspective of the overall system architecture, the task flow tree generation process in Example 3 lays a solid foundation for the automation and intelligence of the zero-code intelligent agent construction platform. It converts the user's high-level needs into low-level execution logic through the seamless connection of the module database and the intent vector, and optimizes the executability and efficiency of the task flow through mechanisms such as dormant graying. This design not only simplifies the construction process of complex tasks, but also improves the system's adaptability to diverse application scenarios, providing users with highly flexible and reliable intelligent agent construction support.

[0037] Embodiment 4: The agent construction unit analyzes all task flow trees, and the process of finding the optimal task flow tree includes: taking the task flow tree with the least dormant grayed-out functional modules as the optimal task flow tree.

[0038] Specifically, in the process of generating the task flow tree, different task flow trees may contain multiple paths or implementation methods, which correspond to different function module call sequences and resource usage. Since the module database contains a large number of function modules, and these modules may have redundant connections or be marked as dormant and grayed out, the number of dormant and grayed out modules in the task flow tree becomes an important indicator to measure its optimization degree. Dormant and grayed out modules represent modules that are logically disabled but physically exist in the current task flow. Although they will not directly participate in task execution, their existence may cause unnecessary complexity or resource redundancy in the module database. Therefore, minimizing the number of dormant and grayed out modules not only simplifies the structure of the task flow tree, but also improves the accuracy of resource allocation and the efficiency of overall task execution. When analyzing the task flow tree, the agent construction unit will perform a comprehensive scan of each tree, count the status of all nodes (i.e., function modules) in it, and record the number of dormant and grayed out modules. This statistical process is not a simple counting operation, but a combination of task logic parsing and module status dynamic judgment. For example, a module may be dormant and grayed out due to the current task logic, but in specific scenarios (such as fault recovery or dynamic adjustment), it may re-participate in execution. Therefore, the intelligent agent construction unit also needs to consider the potential usage of the module during the statistical process to ensure the comprehensiveness and accuracy of the statistical results.

[0039] Among all task flow trees, the agent construction unit selects the optimal task flow tree with the goal of minimizing the number of dormant grayed-out modules. This selection strategy has important practical significance. First, minimizing the number of dormant grayed-out modules can directly reduce the redundant consumption of the system in task execution. For example, in a task flow tree, if a branch path contains a large number of dormant grayed-out modules, then this path may contain unnecessary complex logic or resource occupation, and these logic and resources cannot bring any gain to the current task. By selecting the task flow tree with the least dormant grayed-out modules, the system can avoid calling invalid paths and use more computing and storage resources for actual task execution, thereby improving overall performance. Secondly, this strategy can enhance the certainty and reliability of task execution. When the number of dormant grayed-out modules decreases, the structure of the task flow tree becomes clearer and the calling path of the module becomes more stable, thereby reducing the potential error risk caused by path complexity. For example, in a high-concurrency scenario, fewer dormant grayed-out modules means that the calling of functional modules in the task flow is more concentrated, the possibility of resource competition is reduced, and the system can complete the task with higher efficiency. It is worth noting that the optimization process of the agent construction unit is not only limited to the selection of a single task flow tree, but also has dynamic adaptability. In actual operation, the task environment may change, such as module status updates, dynamic resource allocation, or user demand adjustments. At this time, the agent construction unit can re-analyze all task flow trees and adjust the selection of the optimal task flow tree in real time based on the optimization principle of the number of dormant grayed-out modules. This dynamic adjustment capability enables the zero-code agent construction platform to remain efficient and stable in a changing operating environment.

[0040] Embodiment 5: The semantic parsing and intention decomposition unit performs image language parsing on the natural language instructions input by the user, and the method of decomposing the natural language instructions into multiple intention units with a sequence specifically includes: obtaining an image containing natural language instructions, analyzing the fractal geometric feature vector of the image, and constructing a multi-fractal spectrum; based on the multi-fractal spectrum, constructing a dynamic topological structure vector of the strokes; constructing a high-dimensional feature space vector based on the multi-fractal spectrum and the dynamic topological structure vector; mapping the high-dimensional feature space vector into the discriminant manifold space to obtain the discriminant manifold vector; performing a prior probability-based transformation on each element in the discriminant manifold vector to obtain the corresponding intention unit.

[0041] Specifically, first, the method starts with obtaining images containing natural language instructions. These images may come from scans of user handwritten instructions, picture uploads, or even real-time handwritten input. The system analyzes the fractal geometry features of the image and extracts fractal geometry feature vectors that reflect the shape, structure, and distribution of the text. Fractal geometry features can capture the subtle structure and complex changes of the text in the image, and this information far exceeds the ability of traditional optical character recognition (OCR) technology to analyze the shape of simple characters. For example, fractal geometry feature vectors can capture the changes in stroke thickness, writing speed, and continuity in user handwriting input, thereby providing more refined data support for subsequent task logic modeling. After extracting fractal geometry features, the system further constructs a multifractal spectrum. The multifractal spectrum is a mathematical tool that reflects the uneven distribution in an image. It describes the global and local complexity of the natural language instruction image through changes in the distribution dimension. Through the multifractal spectrum, the system can extract hidden patterns of semantic information from the image, such as the arrangement of text strokes and the trend of morphological changes. This process provides global context information for image language parsing, enabling the system to understand complex handwritten text input or image instructions with certain ambiguity. Next, the system constructs the dynamic topological structure vector of the strokes based on the multifractal spectrum. The dynamic topological structure vector of the strokes captures the geometric connectivity and spatial distribution relationship in the image. It further refines the semantic level of image language parsing by describing the relative position and interaction between the strokes of the text. For example, the stroke sequence of a Chinese character can be accurately modeled by the dynamic topological structure vector, allowing the system to not only parse the shape of the text, but also understand the logical order of the text writing. This ability is particularly important in handwriting input or non-standard font parsing, because users may have large differences in their writing methods, and traditional OCR methods often perform poorly in these scenarios. On the basis of the above, the system combines the multifractal spectrum and the dynamic topological structure vector to further construct a high-dimensional feature space vector. The high-dimensional feature space vector is a comprehensive expression of image language information. It integrates the global and local features of the image and describes the structure of natural language instructions in a highly abstract way. This vector space provides a rich information basis for the subsequent semantic parsing and intention decomposition of the system. By mapping high-dimensional feature space vectors to discriminative manifold space, the system can further extract significant semantic patterns in natural language instructions. Discriminative manifold space is a low-dimensional embedding space that retains task-related discriminative information in high-dimensional feature space while removing redundant features and noise. For example, a complex sentence can be decomposed into multiple key phrases and action words through manifold mapping, thereby greatly improving the efficiency and accuracy of semantic parsing. Finally, the system performs a prior probability-based transformation on each element in the discriminative manifold vector to generate the corresponding intent unit.This process makes comprehensive use of the semantic information and task background knowledge in the manifold vector, and gradually parses complex natural language instructions into structured intent units with a logical order. For example, for a handwritten input instruction such as "analyze this set of data and then generate a bar graph", the system will identify the two main intent units of "analyze data" and "generate a bar graph" from the fractal features and topological structure of the image, and organize them in sequence. These intent units are then input into the task process tree generation unit to construct a task logic that meets user needs.

[0042] Embodiment 6: Obtain an image containing natural language instructions, analyze the fractal geometric feature vector of the image, and construct a multifractal spectrum through the following formula: ; in, is the integral area, which represents the area of ​​the image; is the fractal sequence parameter, is the set value; is the time scale parameter, is the set value; is the space curvature parameter, is the set value; For the The size of the observation window of each scale; For the input image Gray value of the position; is the coordinate of the X axis; is the coordinate of the Y axis; is an integer subscript index; is the scale number; It is a multi-typing spectrum.

[0043] Specifically, in this formula, each part of the input image is represented as a grayscale distribution function , which is in pixel position is the independent variable. By analyzing the fractal characteristics of the image, the formula aims to reveal the law of grayscale value changing with space. This analysis is done by integrating the area The definition of , realizes the global observation of the entire image, while also retaining the detailed features of the local area. In fractal analysis, multi-scale observation windows The introduction of is one of the core of the formula. The size of these windows controls the accuracy of the system analysis. Smaller windows capture detailed changes, while larger windows reveal the overall structure of the image. Through multi-scale step-by-step analysis, the formula can establish a connection between the global and the local, and accurately quantify the complexity of the image. The numerator of the formula is , by calculating the gray value difference The power is integrated to extract the grayscale change characteristics of the image at different scales. is a fractal order parameter used to adjust the sensitivity to different grayscale ranges. When it is smaller, the formula pays more attention to small grayscale fluctuations, which helps capture subtle stroke differences in handwritten text; When the value is larger, the formula emphasizes more on significant grayscale changes, such as character boundaries or prominent stroke features. The integral result accumulates the grayscale change characteristics in the entire image area, thus providing a preliminary quantitative description of the fractal geometric characteristics.

[0044] At the same time, the denominator of the formula A normalization and scaling mechanism is introduced. is the time scale parameter, is the spatial curvature parameter, which together adjusts the sensitivity of the fractal spectrum to different scale and dimensional features. Emphasizes the dynamic characteristics of grayscale changes in the image, which is similar to describing the rate of change of the image; while the spatial curvature parameter It captures the nonlinear characteristics of the grayscale value with spatial distribution, reflecting the overall geometric complexity of the image. This normalized design ensures the robustness of the formula, enabling it to maintain consistent feature extraction effects on images of different sizes and resolutions. The final output of the formula is a multifractal spectrum. , which is achieved by and The analysis of the values ​​generates a series of feature points, which form a curve that can reflect the complexity and unevenness of the image. The fractal spectrum not only expresses the static characteristics of the image, but also reveals the dynamic behavior of grayscale changes in different areas of the image. For example, in an image containing natural language instructions, the fractal spectrum can clearly distinguish the dense and sparse parts of the characters, thereby providing basic support for the subsequent construction of dynamic topological vectors. The multi-scale characteristics of the fractal spectrum also enable it to show higher accuracy and adaptability when processing blurred or handwritten text. The application of this formula in the zero-code intelligent agent construction platform has greatly expanded the platform's adaptability to diversified inputs. Traditional text parsing technologies often rely on structured inputs, such as standardized digital text or high-quality printed text. However, in actual application scenarios, the natural language instructions entered by users may exist in the form of pictures, including handwritten input, screenshots, or scans. These input forms often have fuzzy and irregular features, which are difficult to be accurately parsed by traditional technologies. The formula based on fractal geometry makes up for the shortcomings of traditional methods by extracting features from multi-scale changes in images. It can not only accurately capture the complex geometric information in the image, but also reveal hidden semantic patterns through fractal spectra. For example, when a user inputs the instruction "analyze sales data and generate a bar chart" in handwriting, this formula can decompose the input image into a gray value matrix, construct a fractal spectrum through multi-scale integration and normalization operations, and thus extract geometric features related to "analysis", "sales data" and "bar chart". Subsequently, these features can be mapped to the discriminative manifold space and further decomposed into structured intent units. This parsing process not only overcomes the ambiguity of handwritten input, but also understands the context of the input by capturing writing features (such as stroke order or intensity), thereby providing an accurate basis for task process tree generation.

[0045] Embodiment 7: Constructing the dynamic topological structure vector of the stroke based on the multifractal spectrum by the following formula : ; in, To make The largest value; is the highest curvature order; is an integer subscript index; for The gradient of is the imaginary number symbol; For the The curvature characteristic scale of the order is a set value ranging from 0.5 to 0.9; is the regional integral differential; is the characteristic angle, ranging from 10 to 30 degrees.

[0046] Specifically, the formula Indicates that in the integral area This region usually covers the entire image or the part of the image related to the target text. By accumulating the characteristics of pixels in this region, the formula can capture the global and local geometric characteristics of the strokes and provide a comprehensive mathematical description for the construction of dynamic topological structure vectors. The first important component of the formula is the multifractal spectrum ,in Yes Reaching the maximum This means that the formula pays special attention to the areas of the image that are most fractal, that is, those with the highest geometric complexity. This choice can highlight the areas with the most dramatic changes in the strokes, such as inflection points, intersections, or sharp changes at edges. , which is the gradient of the grayscale value of the image, and the formula further captures the direction and magnitude of the change in grayscale distribution. Gradient is an important feature in image analysis, which directly reflects the direction and intensity of the change in pixel value, so that the edges and details of the strokes can be accurately located.

[0047] The formula contains plural forms ,in is a characteristic angle, ranging from 10 to 30 degrees. By introducing this complex form, the formula can not only capture the amplitude changes of the strokes, but also reflect their directional characteristics. This processing method is very suitable for modeling complex handwritten characters, because the direction and angle of handwritten strokes often contain rich semantic information. For example, in some characters, strokes in a specific direction may have important distinguishing characteristics, and the complex form representation can effectively encode these features. The denominator of the formula This section emphasizes the normalization of the second-order derivative of the fractal spectrum. The rate of change under the value can capture the subtle differences in the fractal characteristics of the strokes. By introducing the denominator into the square and normalization, the formula smoothes the interference of high-frequency noise on feature extraction, making the dynamic topological structure vector more robust and stable. Another key part of the formula is , which introduces the curvature characteristics of the stroke into the modeling process. Indicates the stroke in The curvature of the order, is the corresponding characteristic scale, ranging from 0.5 to 0.9. The design of this part enables the formula to capture the multi-order curvature characteristics of the strokes, especially at the inflection points or curved areas where the high-order curvature characteristics are significant, and the formula can effectively describe their geometric characteristics. The curvature characteristic is particularly important in stroke analysis because it can distinguish different types of strokes such as straight lines, arcs, and sharp corners, thereby providing support for the accurate characterization of the text form. The dynamic topological structure vector finally output by the formula It is a comprehensive expression of all the above characteristics. It integrates the fractal characteristics (through ), gradient direction (through ) and curvature information (via ), forming a high-dimensional vector. This vector can fully describe the geometric structure, dynamic changes and spatial distribution of the strokes, and provides high-precision feature support for semantic parsing and intent decomposition in the zero-code intelligent agent building platform. At the application level, this formula provides a powerful tool for parsing complex natural language instruction images. For example, for a handwritten input instruction "generate a trend chart of sales data", the formula can extract the stroke structure of each word from the image and use the dynamic topological structure vector Accurately describe the geometric shape and writing characteristics of the strokes. These vectors can then be matched with the functional module keywords in the module database to complete the semantic mapping of instructions to functional modules.

[0048] Embodiment 8: A high-dimensional feature space vector is constructed according to the multifractal spectrum and the dynamic topological structure vector by the following formula: ; in, is a high-dimensional feature space vector; is the coordinate of the first coordinate axis in the high-dimensional feature space; is the coordinate of the second coordinate axis in the high-dimensional feature space; is the resolution parameter of the high-dimensional feature space; for Bessel function of order.

[0049] Specifically, the input of the formula includes the multifractal spectrum and dynamic topology vector , which capture the global geometric characteristics and local stroke structure of the image respectively. The fractal spectrum reveals the self-similarity and complexity of the character structure by modeling the changes in the grayscale values ​​of the image at multiple scales; the topological structure vector further describes the dynamic connection relationship of the character strokes. By combining these two parts of information, the formula can generate a high-dimensional vector with rich semantic features from both global and local dimensions. The first term in the formula It is the core feature coupling mechanism. Dynamic topology vector Provides a basic description of the stroke shape, and the mixed second-order partial derivative of the fractal spectrum It reveals that the fractal spectrum varies with the parameter and The complexity of simultaneous changes. This part of the features is condensed into a scalar value to express the interdependence of the characters in terms of geometric properties and local connectivity. This coupling method can capture the complex interactions between strokes in natural language instruction images and the dynamic changes of the overall structure composed of different strokes. is the distribution control factor of high-dimensional space, where is a resolution parameter used to control the diffusion degree of features in high-dimensional space. The design of this part is based on the idea of ​​Gaussian distribution, and its purpose is to ensure that the density and distribution of features have a certain degree of continuity and consistency in high-dimensional feature space. Through this item, the formula can effectively avoid the parsing problems caused by features that are too sparse or too concentrated, so that the generated high-dimensional feature vector is both expressive and smooth and stable in high-dimensional space. The third item yes The introduction of the 2-order Bessel function further enhances the ability of high-dimensional feature vectors to express complex stroke structures. The Bessel function is a special function that is widely used in describing wave and vibration phenomena. In this formula, it captures the periodic changes in the stroke shape and connection relationship through radial symmetry. and The Bessel function can describe the geometric center distribution of different characters and strokes in the natural language instruction image. This feature extraction method is particularly suitable for processing those structures with cyclic or symmetric properties, such as closed ring structures in some letters or symbols.

[0050] The formula is obtained by double integration The above feature components are and Accumulation is performed in two dimensions. Integration area Represents the range of the high-dimensional feature space, which may include the feature distribution of the entire character or be limited to certain local areas. The result of double integration Is a high-dimensional feature vector used to represent the semantic characteristics of natural language instruction images in high-dimensional space. Through this high-dimensional vector, the system can map the features to the discriminative manifold space in subsequent processing to further extract intent units. Example 8 deeply integrates the global geometric information of the fractal spectrum with the local dynamic information of the topological structure, and realizes a seamless conversion from natural language images to high-dimensional feature spaces. This conversion can capture the diverse stroke styles and geometric relationships in handwritten instruction images, thereby enhancing the input adaptation capabilities of the platform. Secondly, through the combination of Gaussian distribution and Bessel function, the formula ensures that the distribution of high-dimensional features is both continuous and geometrically consistent, laying a solid foundation for subsequent feature discrimination and classification. In application scenarios, the function of this formula is particularly prominent. Suppose a user enters a command by handwriting, such as "analyze sales data and generate a line chart", the formula will first extract the stroke structure of each character in the image, and through the dynamic topological vector Combined with fractal spectrum, high-dimensional feature space vector is generated This high-dimensional feature space vector can not only accurately describe the geometric structure of the text, but also reveal the logical relationship between characters, providing high-quality input data for the intent decomposition unit.

[0051] Embodiment 9: The high-dimensional feature space vector is mapped to the discriminant manifold space by the following formula to obtain the discriminant manifold vector: ; in, To discriminate the harmonic function in the manifold space; is the F-norm operation; is the absolute value operator; is the manifold coordinate system Quantity; is the manifold coordinate system The characteristic scale of the component is the set value.

[0052] Specifically, the first part of the formula is an input vector generated from a high-dimensional feature space. It has been modeled through multi-fractal spectra and dynamic topological structures, and contains global geometric features and local connectivity in natural language images. On this basis, the formula further performs geometric optimization on high-dimensional features to generate feature vectors with stronger discriminability. is a key geometric control term. is the dynamic topology vector The Frobenius norm of the vector The square root of the sum of the squares of all elements in is a measure of A scalar value for the size of the The derivative calculation of captures the rate of change of the topological structure vector in the direction. Then, by taking the logarithmic function and absolute value operation, the formula amplifies the features with significant changes while suppressing smaller or insignificant features. The design of this part can highlight the key stroke direction information in natural language images, such as the edges, inflection points and intersections of characters, thereby providing strong geometric support for the discrimination of intent. The second part of the formula is an orthogonal adjustment term in a manifold space, where and They are the first components and their characteristic scales. This part is obtained by the hyperbolic cosine function The introduction of , makes nonlinear adjustments to the feature distribution in the manifold space. The hyperbolic cosine function has the characteristic of smooth growth, which can effectively suppress the components of features that are too large or too small, making the vector The distribution of is more uniform, avoiding the information imbalance caused by some feature components being too strong. The setting value of adjusts the characteristic scale so that components of different dimensions can be reasonably expressed in the same space. The last part of the formula is a harmonic function in manifold space. Harmonic function is a special function that satisfies Laplace equation. It is geometrically smooth and symmetric, and can optimize the distribution of features in manifold space and further improve vector In practical applications, the harmonic function It can also play a regularization role to prevent the vector from overfitting the noise in the high-dimensional features. The overall result of the formula is to generate the discriminant manifold vector , which is a feature expression in the discriminative manifold space, has higher discriminability and compactness. The geometric optimization formula projects the high-dimensional features into a low-dimensional but discriminative manifold space, thereby retaining the core semantic information of natural language images and removing redundant features and noise. In this way, the manifold vector It can be directly input into the subsequent intent decomposition unit or task logic generation unit to complete the key transition from image input to task parsing. Through the calculation of the Frobenius norm and derivative value of the dynamic topological structure vector, the formula can extract the most significant geometric changes from high-dimensional features, which is particularly suitable for capturing directional features in natural language instructions. This optimization method overcomes the curse of dimensionality problem that is prone to occur in traditional high-dimensional feature vectors, so that the generated manifold vector is not only compact but also has strong distinguishing ability. Secondly, the formula reasonably distributes the feature components in the manifold space through the combination of hyperbolic cosine function and smooth harmonic function. This design avoids the information loss caused by uneven feature distribution, while enhancing the vector The ability to distinguish between different intent units. For example, in a complex natural language image containing multiple instructions, the strokes of different characters may overlap or be blurred. This optimization mechanism can accurately separate the intent features with independent semantics from them, providing support for multi-task decomposition. In actual applications, suppose the user inputs a handwritten image, such as "analyze data and generate trend charts", and the formula is obtained through a high-dimensional feature vector Extract global and local features, and then generate vectors through discriminative manifold mapping This vector can not only distinguish the two main tasks of "analyzing data" and "generating trend graphs", but also capture the semantic features of each task, providing high-quality input for subsequent module matching and task logic generation.

[0053] Embodiment 10: Each element in the discriminant manifold vector is transformed based on the prior probability by the following formula to obtain the corresponding intention unit: ; in, For the Intention unit; The mean and variance are The prior probability of the Gaussian distribution of ; for The gradient of is the first vector in the discriminant manifold elements.

[0054] Specifically, the first part of the formula is the gradient-driven feature resolution term In this section, Represents the discriminant manifold vector The gradient of an element directly reflects The rate and direction of change in manifold space. This gradient captures the natural language image Related geometric features, such as mutation points or direction changes in character structures. This local geometric feature is particularly important for semantic discrimination of intent units because it can accurately locate certain image areas with special meanings. At the same time, the formula is The determinant operation further extracts the high-dimensional feature space vector The calculation of the determinant of this second-order derivative reflects the nonlinear changes in the feature space and is used to supplement the details that may be missed in the gradient information. For example, in some complex natural language instructions, the subtle curvature changes of character strokes may be important for semantic discrimination. By combining the gradient and curvature characteristics, the formula can comprehensively capture the geometric information of natural language images and generate a more complete semantic expression. The second part of the formula is a contrast suppression term used to weaken the contribution of features irrelevant to the intent. and plural forms It captures the local changes and directional information of grayscale values ​​in natural language images. This part of the features may contain interference factors, such as local irregularities or noise of handwritten characters. By subtracting these interference factors, the formula can focus on the key features related to the intent during the discrimination process.

[0055] The most important thing is the prior probability term in the formula , which is a value with mean and variance The Gaussian distribution of is used to introduce statistical information to enhance the robustness of intent discrimination. Prior probability The introduction of has important statistical significance: it combines prior knowledge and observed data, so that the discrimination results of intent units can adapt to data characteristics and avoid overfitting to a certain extent. For example, for intent units with higher frequencies, the distribution tendency of prior probability will give them higher weights, thereby ensuring that these common intent units have higher confidence in the discrimination process. The formula as a whole is maximized through a By integrating the above parts, the transformation from discriminant manifold vector to intention unit is realized. The goal of the maximization process is to find a , so that the value of the formula reaches the maximum, that is, select an intention unit that can most accurately represent the current discriminant feature. This process not only integrates geometric information (gradient and curvature), statistical information (prior probability) and dynamic characteristics (feature change rate), but also effectively eliminates interference through feature comparison, ensuring the accuracy of intention unit discrimination. The formula of Example 10 is of great practical significance in the zero-code intelligent agent construction platform. For example, when a user inputs a handwritten instruction image "Generate a line graph to analyze sales data", the formula can be used through the manifold vector Element-by-element analysis is performed to extract geometric features related to "generating a line graph" and "analyzing sales data", and these features are semantically distinguished using prior probability, ultimately generating two independent intent units. and . These intent units can then be input into the task flow tree generation unit to build the task logic. The technical advantage of the formula lies in its ability to jointly model multiple features. Through the integration of gradients, curvatures, and prior probabilities, the formula is able to achieve high-precision intent unit discrimination in complex scenarios, especially when dealing with blurred, handwritten, or irregular characters. In addition, the formula reduces the impact of interference information through contrast suppression terms and improves the robustness of the overall discrimination. This feature enables the platform to efficiently adapt to diverse natural language image inputs and meet the complex needs of users.

[0056] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A zero-code intelligent agent construction platform, characterized in that: It includes: A module database, a semantic parsing and intention decomposition unit, a task flow tree generation unit and an agent construction unit; the module database stores multiple function modules that perform a single function at fixed positions, and each function module is connected to an adjacent function module; there are at least N function modules with the same function, where N is a set value; the semantic parsing and intention decomposition unit is used to perform image language parsing on the natural language instructions input by the user, decompose the natural language instructions into multiple intention units with a sequence, use each intention unit as an element in a vector, and construct an intention vector according to the sequence of the intention units; The task process tree generating unit is used to generate a plurality of corresponding task process trees according to the intention vector; The intelligent agent construction unit is used to analyze all task process trees, find the optimal task process tree, and call all target function modules in the task process tree according to the order of the target function modules in the optimal task process tree to complete the construction of the intelligent agent.

2. The zero-code intelligent agent construction platform as described in claim 1, characterized in that: The updating process of the module database specifically includes: if any functional module in the module database is to be deleted, the functional module is directly put into permanent dormancy; if a functional module is to be added to the module database, the functional module is directly added and stored in any free position in the module database, and connected with other adjacent functional modules.

3. The zero-code intelligent agent construction platform as described in claim 2, characterized in that: The process of generating any task flow tree by the task flow tree generation unit specifically includes: finding all the functional modules corresponding to each intention unit in the intention vector in the module database as the target functional module of each intention unit; each functional module includes a keyword set for matching the intention unit; in the keyword set, each keyword is represented by a feature code, and each intention unit is also a code; for each intention unit, all the corresponding target functional modules are grouped into a set as the target functional module set of the intention unit; a target functional module is arbitrarily selected from each target functional module set and connected with the target functional modules in other target functional module sets in the order of the intention units to generate a corresponding target task tree; during the connection process, if the functional module adjacent to the target functional module is not the target functional module, the functional module is put into dormancy and grayed out.

4. The zero-code intelligent agent construction platform as described in claim 3, characterized in that: The intelligent agent construction unit analyzes all task flow trees, and the process of finding the optimal task flow tree includes: taking the task flow tree with the least dormant grayed-out functional modules as the optimal task flow tree.

5. The zero-code intelligent agent construction platform as described in claim 4, characterized in that: The semantic parsing and intention decomposition unit performs image language parsing on the natural language instructions input by the user, and the method of decomposing the natural language instructions into multiple intention units with a sequence specifically includes: obtaining an image containing the natural language instructions, analyzing the fractal geometric feature vector of the image, and constructing a multi-fractal spectrum; based on the multi-fractal spectrum, constructing a dynamic topological structure vector of the strokes; constructing a high-dimensional feature space vector based on the multi-fractal spectrum and the dynamic topological structure vector; mapping the high-dimensional feature space vector into the discriminant manifold space to obtain the discriminant manifold vector; and performing a prior probability-based transformation on each element in the discriminant manifold vector to obtain the corresponding intention unit.

6. The zero-code intelligent agent construction platform as claimed in claim 5, characterized in that: The following formula is used to obtain an image containing natural language instructions, analyze the fractal geometric feature vector of the image, and construct a multifractal spectrum: ； in, is the integral area, which represents the area of ​​the image; is the fractal sequence parameter, is the set value; is the time scale parameter, is the set value; is the space curvature parameter, is the set value; For the The size of the observation window of each scale; For the input image Gray value of the position; is the coordinate of the X axis; is the coordinate of the Y axis; is an integer subscript index; is the scale number; It is a multi-typing spectrum.

7. The zero-code intelligent agent construction platform as claimed in claim 6, characterized in that: The dynamic topological structure vector of the strokes is constructed based on the multifractal spectrum through the following formula: : ； in, To make The largest value; is the highest curvature order; is an integer subscript index; for The gradient of is the imaginary number symbol; For the The curvature characteristic scale of the order is a set value ranging from 0.5 to 0.9; is the regional integral differential; is the characteristic angle, ranging from 10 to 30 degrees.

8. The zero-code intelligent agent construction platform as claimed in claim 7, characterized in that: The high-dimensional feature space vector is constructed based on the multifractal spectrum and dynamic topological structure vector through the following formula: ； in, is a high-dimensional feature space vector; is the coordinate of the first coordinate axis in the high-dimensional feature space; is the coordinate of the second coordinate axis in the high-dimensional feature space; is the resolution parameter of the high-dimensional feature space; for Bessel function of order.

9. The zero-code intelligent agent construction platform as claimed in claim 8, characterized in that: The high-dimensional feature space vector is mapped to the discriminant manifold space through the following formula to obtain the discriminant manifold vector: ； in, To discriminate the harmonic function in the manifold space; is the F-norm operation; is the absolute value operator; is the manifold coordinate system Quantity; is the manifold coordinate system The characteristic scale of the component is the set value.

10. The zero-code intelligent agent construction platform as claimed in claim 8, characterized in that: Through the following formula, each element in the discriminant manifold vector is transformed based on the prior probability to obtain the corresponding intention unit: ； in, For the Intention unit; The mean and variance are The prior probability of the Gaussian distribution of ; for The gradient of is the first vector in the discriminant manifold elements.

Images