A low-code process design system and application based on a workflow engine

Abstract

This invention relates to the field of computer software technology, and in particular to a low-code workflow design system and application based on a workflow engine, comprising a multi-flow collaboration engine layer, an AI-enhanced rule orchestration layer, a distributed architecture layer, a data storage layer, a user interaction layer, and an extension interface layer. This invention achieves deep linkage between logic flow, data flow, and event flow through an event bus, constructing a closed loop of event triggering, logical judgment, and data flow, overcoming the limitations of traditional platforms that only support single process types and lack multi-flow collaboration capabilities. The logic flow engine supports multi-level conditional branches, nested sub-processes, and parallel task splitting; the data flow engine achieves seamless cross-system data mapping; and the event flow engine ensures real-time triggering and asynchronous notification. The three work together to meet the diverse needs of complex enterprise-level processes, such as the linkage between approval flows and data processing flows, and cross-system event-driven process execution.

Description

Technical Field

[0001] This invention relates to the field of computer software technology, and in particular to a low-code workflow design system and application based on a workflow engine. Background Technology

[0002] Low-code development platforms, as core tools for enterprise digital transformation, enable rapid construction of business processes through visual orchestration and minimal code, significantly shortening development cycles. However, existing low-code platforms have significant limitations in complex enterprise-level process scenarios. Traditional platforms often focus on single process types, failing to achieve deep collaboration between logic flows, data flows, and event flows, and thus struggle to support full-process coverage of complex business scenarios. Therefore, a low-code process design system based on a workflow engine is needed. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a low-code workflow design system and application based on a workflow engine.

[0004] The technical solution adopted in this invention is: a low-code workflow design system based on a workflow engine, including a multi-flow collaboration engine layer, an AI-enhanced rule orchestration layer, a distributed architecture layer, a data storage layer, a user interaction layer, and an extension interface layer; The multi-stream collaborative engine layer achieves deep linkage between the logic flow engine module, the data flow engine module, and the event flow engine module through the event bus, forming a closed loop of event triggering, logical judgment, and data flow. The AI-enhanced rule orchestration layer is used to convert natural language into executable rules and perform conflict detection, thereby lowering the threshold for business rule orchestration. The distributed architecture layer is deployed through a multi-engine cluster to improve the system's high-concurrency processing capabilities; The data storage layer is optimized for data persistence to improve data storage performance; The user interaction layer provides drag-and-drop and programmable dual orchestration modes as well as full-process operation monitoring functions; The extended interface layer enables dynamic system expansion and seamless integration with third-party systems through a plug-in architecture and standard connectors.

[0005] As a further description of the above technical solution: The logic flow engine module is designed based on an event-driven model. It uses a lightweight rule engine to parse multi-dimensional conditional expressions, supports multi-level conditional branches, nested sub-process calls, and parallel task splitting, and implements serial and parallel execution logic through a process scheduler.

[0006] As a further description of the above technical solution: The data flow engine module has a built-in visual data mapping tool that supports data access from RESTful APIs, databases, and message queues. Users can configure field mapping relationships by dragging and dropping, and it supports automatic parsing of JSON format to achieve cross-system data mapping and conversion.

[0007] As a further description of the above technical solution: The event stream engine module is designed based on the publish-subscribe pattern. The event source is accessed through Kafka, and the rule engine dynamically matches the event type and distributes it to the target process node, supporting real-time triggering and asynchronous notification functions.

[0008] As a further description of the above technical solution: The AI-enhanced rule orchestration layer includes a natural language to rule conversion module and a rule conflict detection module; The natural language to rule conversion module integrates GPT-4 and Wenxin Yiyan API to parse semantics, extract entities and relationships, and convert natural language descriptions into rule expressions supported by the logic flow engine. The rule conflict detection module constructs a rule dependency network based on a knowledge graph, detects logical contradictions in rules through a logical consistency algorithm, and provides correction suggestions in conjunction with an industry template library.

[0009] As a further description of the above technical solution: The distributed architecture layer includes a multi-engine cluster deployment module, which deploys the process engine as an independent service unit in a cluster according to the business domain, and realizes request load balancing and process instance routing through a routing gateway.

[0010] As a further description of the above technical solution: The data storage layer includes a data persistence optimization module, which adopts an in-memory computing and distributed storage architecture. Real-time data is stored in Redis, historical data is asynchronously archived to the ES database, and process variables are stored in JSON format and support indexing by field.

[0011] As a further description of the above technical solution: The user interaction layer includes a visual orchestration interface module and a runtime monitoring module; The visual orchestration interface module is developed based on Vue components, uses a virtual DOM to achieve incremental updates, provides a drag-and-drop node library and DSL programming method, and supports automatic node snapping and alignment. The operation monitoring module collects node execution logs through Kafka, processes and analyzes the data using Elasticsearch, and displays the node execution status in real time.

[0012] As a further description of the above technical solution: The extended interface layer includes a custom node development module and a third-party system integration module; The custom node development module provides an SDK that allows developers to write components, integrate them into the platform via a Maven repository, and support hot updates. The third-party system integration module implements identity authentication based on OAuth2.0 and SSO, provides pre-built connectors and open APIs, and supports one-click reuse of external system events and custom integration templates via Webhook.

[0013] An application based on the low-code process design system described above is suitable for enterprise-level process automation, cross-system integration, and complex business rule configuration scenarios, enabling dynamic drag-and-drop process design, logical rule-based verification, and cross-scenario reuse.

[0014] The present invention has the following beneficial effects: 1. This invention achieves deep linkage between logic flow, data flow, and event flow through an event bus, constructing a closed loop of event triggering, logical judgment, and data flow, overcoming the limitations of traditional platforms that only support single process types and lack multi-flow collaboration capabilities. The logic flow engine supports multi-level conditional branches, nested sub-processes, and parallel task splitting; the data flow engine achieves seamless cross-system data mapping; and the event flow engine ensures real-time triggering and asynchronous notification. The three work together to meet the diverse needs of complex enterprise-level processes, such as the linkage between approval flows and data processing flows, and cross-system event-driven process execution.

[0015] 2. This invention utilizes a natural language to rule conversion module, allowing non-technical personnel to directly describe business rules in everyday language. The AI ​​model automatically converts these into executable rule expressions, lowering the barrier to rule writing from the code level to the natural language level and significantly improving rule configuration efficiency. Simultaneously, the rule conflict detection module, based on knowledge graphs and logical consistency algorithms, automatically identifies logical contradictions in rules and provides correction suggestions, avoiding manual configuration errors, ensuring the accuracy and reliability of rule execution, and reducing process failures.

[0016] 3. This invention avoids the "god-service" problem of traditional centralized architectures by adopting a multi-engine cluster deployment mode based on business domains, combined with dynamic load balancing of the routing gateway and process instance routing strategies. This disperses the pressure on single nodes and single databases, thereby improving both concurrency and throughput. At the same time, the data persistence optimization adopts an architecture of Redis real-time storage and ES historical archiving. Process variables are split and stored in JSON format and support field-level indexing, which solves the problems of inefficient querying and full table scans caused by serialized storage of process variables in traditional platforms. The response time for querying process historical data is shortened to milliseconds, and the efficiency of log retrieval is improved. Meanwhile, by regularly cleaning up expired data, the long-term stable operation of the system is ensured.

[0017] 4. The drag-and-drop node library of the visual orchestration interface module of this invention is designed for non-technical users. Nodes automatically snap and align, simplifying process design operations and improving efficiency. The DSL programmatic orchestration is designed for technical users, using Groovy to implement code-based orchestration of complex processes, solving the problem of low drag-and-drop efficiency when there are too many nodes. The combination of these two modes meets the needs of different user groups, lowering the barrier to entry for non-technical users while providing technical users with flexible expansion space, adapting to a full range of needs from simple processes to complex business scenarios.

[0018] 5. The plug-in architecture of the extended interface layer in this invention allows developers to write custom nodes via SDK, which can be integrated into the platform through the Maven repository without restarting the backend service, enabling hot updates and significantly improving developer efficiency. Simultaneously, the system provides pre-built connectors and open APIs, implements identity authentication based on OAuth2.0 and SSO, supports seamless integration with third-party systems such as ERP, CRM, and IoT platforms, receives external system events via Webhook, and allows for one-click reuse of custom integration templates, significantly reducing cross-system integration costs and adapting to the system integration needs of enterprises in the process of rapid business iteration and digital transformation. Attached Figure Description

[0019] Figure 1 This is a logical architecture diagram of the system of the present invention. Detailed Implementation

[0020] Reference Figure 1 The present invention provides a low-code workflow design system based on a workflow engine, comprising a multi-flow collaboration engine layer, an AI-enhanced rule orchestration layer, a distributed architecture layer, a data storage layer, a user interaction layer, and an extension interface layer.

[0021] The multi-stream collaborative engine layer achieves deep linkage between the logic flow engine module, the data flow engine module, and the event flow engine module through the event bus, forming a closed loop of event triggering, logical judgment, and data flow. The logic flow engine module is designed based on an event-driven model. It uses a lightweight rule engine (AviatorScript, Groovy) to parse multi-dimensional conditional expressions, supports multi-level conditional branches (such as amount > 5000 && department == R&D department), nested sub-process calls (sub-process input and output parameters are dynamically passed through the parent process), and parallel task splitting (such as parallel approval by multiple departments). It also implements serial and parallel execution logic through a process scheduler. Data Flow Engine Module: Built-in visual data mapping tool, users can configure field mapping relationships by dragging and dropping (such as converting the order amount field of the MES system to the orderMoney parameter of the approval system), supports data access from RESTful API, databases (MySQL / Redis) and message queues (Kafka / RabbitMQ), automatically parses JSON format, and solves the problem of cross-system data format incompatibility; Event Stream Engine Module: Designed based on the publish-subscribe pattern, event sources (such as sensors and API callbacks) are accessed through Kafka, and the rule engine dynamically matches the event type and distributes it to the target process node. It supports real-time triggering (such as triggering a maintenance work order due to abnormal signals from IoT devices) and asynchronous notifications (such as sending emails / notifications after a process node is completed). Collaboration mechanism: The event bus enables the linkage of three flows: the event flow triggers the logic flow, the logic flow drives the data flow, and the data flow feeds back into the event flow.

[0022] The AI-enhanced rule orchestration layer is used to convert natural language into executable rules and perform conflict detection, lowering the threshold for business rule orchestration. This layer aims to reduce the threshold for business rule orchestration and improve the accuracy of rule configuration, including a natural language to rule conversion module and a rule conflict detection module.

[0023] Natural Language to Rule Module: Integrates GPT-4 and Wenxin Yiyan API to parse user-input natural language descriptions (such as triggering three-level approval when the purchase amount is greater than 5,000 yuan and the department is the R&D department), extracts entities such as amount and department, as well as relationships such as triggering conditions and execution order, and maps them into expression syntax supported by the logic flow engine to generate directly executable rules; Rule conflict detection module: Based on the knowledge graph, a rule dependency network is constructed. Logical consistency algorithms such as SAT solver are used to detect logical contradictions in rules (such as requiring both >10000 and <5000). Combined with industry template library, the module recommends the optimal rule combination and provides correction suggestions (such as suggesting merging to >10000).

[0024] The distributed architecture layer is deployed through a multi-engine cluster to improve the system's high-concurrency processing capabilities. The multi-engine cluster deployment module treats the process engine as an independent service unit and splits the clusters according to business domains (such as financial approval clusters and supply chain clusters). It achieves request load balancing and process instance routing through a routing gateway. New process requests are distributed to different clusters according to policies, and operation requests of running processes are routed to their respective clusters to avoid excessive pressure on a single database.

[0025] The data storage layer is optimized for data persistence to improve data storage performance. The data persistence optimization module adopts an in-memory computing and distributed storage architecture. Real-time data (such as process running status) is stored in Redis to ensure read speed, and historical data is asynchronously archived to the ES database. Process variables are stored in JSON format, supporting indexing by field (such as querying related processes by department field), and expired historical data is cleaned up by scheduled tasks. The ES search engine is used to improve query efficiency.

[0026] The user interaction layer provides drag-and-drop and programmable dual orchestration modes and full-process operation monitoring functions; this layer provides an interactive entry point for process design and operation monitoring, including a visual orchestration interface module and an operation monitoring module.

[0027] The visual orchestration interface module provides a drag-and-drop node library (including logic nodes, data nodes, and event nodes) and a real-time rendering engine. Nodes automatically snap and align when dragged, making it suitable for non-technical users. It also supports DSL programming based on the Groovy language, allowing technical personnel to implement code-based orchestration of complex processes and solving the problem of low drag-and-drop efficiency when there are too many nodes. Nodes are developed based on Vue components and use a virtual DOM to achieve incremental updates, ensuring fast interface response. Operation monitoring module: Logs are collected and processed using the high-throughput Kafka distributed messaging system, and further analyzed using the ES data processing system. The module displays node execution logs and process running status (success or failure) in real time, and supports historical process data retrieval and performance bottleneck identification.

[0028] The extended interface layer enables dynamic system expansion and seamless integration with third-party systems through a plug-in architecture and standard connectors. This layer is used to improve system flexibility and scalability, including custom node development modules and third-party system integration modules.

[0029] Custom node development module: Provides an SDK that allows developers to write components, integrate them into the platform through a Maven repository, adopts a plug-in architecture, and allows custom nodes to be registered through a standard interface, supporting hot updates and taking effect without restarting the backend service; Third-party system integration module: Based on OAuth2.0 and SSO, it implements identity authentication, provides pre-built connectors (DingTalk, Lark, Enterprise WeChat) and open APIs, supports receiving external system events (such as IoT device status changes) through Webhook, and supports one-click reuse of custom integration templates to achieve seamless integration with existing enterprise ERP, CRM, IoT platforms and other systems.

[0030] An application based on the low-code process design system described above is suitable for enterprise-level process automation, cross-system integration, and complex business rule configuration scenarios, enabling dynamic drag-and-drop process design, logical rule-based verification, and cross-scenario reuse.

[0031] Specifically: Process design phase: The collaboration between the user interaction layer and the AI-enhanced rule orchestration layer Users design processes through the visual orchestration interface module in the user interaction layer: Non-technical personnel design processes through a drag-and-drop node library (including logical nodes, data nodes, and event nodes), with nodes automatically snapping together when dragged and configured with node attributes through visual operations; Technical personnel can choose the DSL programming method to write process code based on the Groovy language to orchestrate complex processes, and the system uses a virtual DOM to achieve incremental updates and render the process design effect in real time.

[0032] When business rules need to be configured, users input rules described in natural language through the natural language to rule module of the AI-enhanced rule orchestration layer. This module integrates GPT-4 and Wenxin Yiyan API to parse semantics, extract entities (such as amount, department) and relationships (triggering conditions, execution order) in the rules, and map them into expression syntax supported by the logic flow engine to generate executable rule expressions.

[0033] After a rule is generated, the rule conflict detection module is automatically activated. It constructs a rule dependency network based on a knowledge graph and uses logical consistency algorithms such as the SAT solver to detect whether there is a logical contradiction between the rule and the configured rules (such as requiring both amounts >10000 and <5000). If a conflict exists, a correction suggestion is generated (such as suggesting merging it to an amount >10000). After user confirmation, the rule is synchronized to the rule base of the logic flow engine module for logical judgment during process execution.

[0034] Process execution phase: coordination between the multi-stream collaboration engine layer, distributed architecture layer, and data storage layer After the process is deployed, the event flow engine module receives event source triggers: event sources (such as IoT device signals, API callbacks, and third-party system events) are connected to the event flow engine through Kafka. The engine is based on the publish-subscribe pattern and dynamically matches the event type through the rule engine, distributing the event to the target process node and triggering the execution of the logic flow engine.

[0035] After receiving an event trigger, the logic flow engine module, based on the event-driven model, calls a lightweight rule engine (such as AviatorScript or Groovy) to parse the conditional expressions in the rule base and determine the process execution path: if multi-level conditional branches are met, the branch logic is executed; if nested sub-processes are involved, the input / output parameters are dynamically passed through the parent process to call the corresponding sub-process; if parallel tasks exist, the parallel tasks are split and executed through the process scheduler.

[0036] When process execution requires cross-system data interaction, the logic flow engine sends a data processing request to the data flow engine module. The data flow engine uses a built-in visual data mapping tool to obtain data from the target system and perform format conversion according to the preset field mapping relationship (such as converting the order amount in the MES system to the orderMoney in the approval system). The converted data is then fed back to the logic flow engine to support the continued execution of the process. At the same time, the data flow engine synchronizes the processed data to the event flow engine to feed back the event triggering logic, realizing the linkage between the data flow and the event flow.

[0037] The multi-engine cluster deployment module of the distributed architecture layer schedules process execution requests through a routing gateway: new process requests are distributed to the corresponding engine clusters according to business domain policies (such as financial approval processes being distributed to the financial approval cluster, and supply chain processes being distributed to the supply chain cluster); operation requests of already running processes are accurately routed to the cluster by querying the KV database of the routing table, avoiding cross-cluster data interaction and improving execution efficiency.

[0038] Real-time data generated during process execution (such as process running status, current node, and temporary variables) is stored in Redis by the data persistence optimization module of the data storage layer; after the process node is completed, the generated historical data (such as execution results and log information) is archived to the ES database through asynchronous tasks. Process variables are stored in JSON format and support indexing by field (such as querying related processes by department field), which facilitates subsequent querying and analysis.

[0039] Extension and Integration Phase: Coordination between the Extension Interface Layer and the Multi-Stream Collaboration Engine Layer When the system needs to add custom nodes, developers write components through the SDK provided by the custom node development module of the extended interface layer. After the component is developed, it is integrated into the platform through the Maven repository. The system adopts a plug-in architecture. Custom nodes are registered to the node library of the visual orchestration interface module through standard interfaces. Hot updates are supported. Users can use it in process design without restarting the backend service. The execution logic of the node is linked with the multi-flow collaboration engine layer through the event bus.

[0040] When integration with third-party systems (such as ERP, CRM, and DingTalk) is required, the third-party system integration module of the extended interface layer implements identity authentication based on OAuth2.0 and SSO, and establishes connections with external systems through pre-built connectors or open APIs. Events from external systems are pushed to the event flow engine through Webhooks, and data from external systems is mapped and transformed through the data flow engine. Process execution results are fed back to external systems through asynchronous notifications (such as emails and DingTalk messages), realizing cross-system data interaction and process linkage. At the same time, users can save the integration configuration as a custom integration template, which supports one-click reuse.

[0041] Monitoring and optimization phase: coordination between the user interaction layer, multi-stream collaboration engine layer, and data storage layer. During process execution, the operation monitoring module collects the execution logs of each node in real time (such as execution status, time consumption, and error information) through Kafka, transmits the log data to ES for processing and analysis, and displays the node execution status, process progress and other information in real time on the visual operation monitoring panel, so that users can track the process operation in real time.

[0042] When an error occurs during process execution, users can view detailed logs through the monitoring panel to quickly locate the fault node. For historical process data, users can use ES's search capabilities to query relevant process records by field (such as department, time, process type). The system automatically analyzes historical data, identifies performance bottleneck nodes (such as an approval node with an excessively long average time), and recommends optimization solutions based on the industry template library (such as splitting parallel nodes and adjusting rule conditions).

[0043] Once the optimization plan is confirmed, users can modify the process configuration through a visual orchestration interface or DSL programming. The modified process is then redeployed, and the multi-flow collaboration engine layer executes the process according to the new configuration. The data storage layer synchronously updates the relevant rules and process configuration data, thus achieving full lifecycle optimization of the process.

[0044] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A workflow engine based low code process design system, characterized in that, It includes a multi-stream collaboration engine layer, an AI-enhanced rule orchestration layer, a distributed architecture layer, a data storage layer, a user interaction layer, and an extension interface layer; The multi-stream collaborative engine layer achieves deep linkage between the logic flow engine module, the data flow engine module, and the event flow engine module through the event bus, forming a closed loop of event triggering, logical judgment, and data flow. The AI-enhanced rule orchestration layer is used to convert natural language into executable rules and perform conflict detection, thereby lowering the threshold for business rule orchestration. The distributed architecture layer is deployed through a multi-engine cluster to improve the system's high-concurrency processing capabilities; The data storage layer is optimized for data persistence to improve data storage performance; The user interaction layer provides drag-and-drop and programmable dual orchestration modes as well as full-process operation monitoring functions; The extended interface layer enables dynamic system expansion and seamless integration with third-party systems through a plug-in architecture and standard connectors.

2. The low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The logic flow engine module is designed based on an event-driven model. It uses a lightweight rule engine to parse multi-dimensional conditional expressions, supports multi-level conditional branches, nested sub-process calls, and parallel task splitting, and implements serial and parallel execution logic through a process scheduler.

3. The low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The data flow engine module has a built-in visual data mapping tool that supports data access from RESTful APIs, databases, and message queues. Users can configure field mapping relationships by dragging and dropping, and it supports automatic parsing of JSON format to achieve cross-system data mapping and conversion.

4. The low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The event stream engine module is designed based on the publish-subscribe pattern. The event source is accessed through Kafka, and the rule engine dynamically matches the event type and distributes it to the target process node, supporting real-time triggering and asynchronous notification functions.

5. A low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The AI-enhanced rule orchestration layer includes a natural language to rule conversion module and a rule conflict detection module; The natural language to rule conversion module integrates GPT-4 and Wenxin Yiyan API to parse semantics, extract entities and relationships, and convert natural language descriptions into rule expressions supported by the logic flow engine. The rule conflict detection module constructs a rule dependency network based on a knowledge graph, detects logical contradictions in rules through a logical consistency algorithm, and provides correction suggestions in conjunction with an industry template library.

6. The low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The distributed architecture layer includes a multi-engine cluster deployment module, which deploys the process engine as an independent service unit in a cluster according to the business domain, and realizes request load balancing and process instance routing through a routing gateway.

7. The low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The data storage layer includes a data persistence optimization module, which adopts an in-memory computing and distributed storage architecture. Real-time data is stored in Redis, historical data is asynchronously archived to the ES database, and process variables are stored in JSON format and support indexing by field.

8. A low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The user interaction layer includes a visual orchestration interface module and a runtime monitoring module; The visual orchestration interface module is developed based on Vue components, uses a virtual DOM to achieve incremental updates, provides a drag-and-drop node library and DSL programming method, and supports automatic node snapping and alignment. The operation monitoring module collects node execution logs through Kafka, processes and analyzes the data using Elasticsearch, and displays the node execution status in real time.

9. A low-code workflow design system based on a workflow engine according to claim 1, characterized in that: The extended interface layer includes a custom node development module and a third-party system integration module; The custom node development module provides an SDK that allows developers to write components, integrate them into the platform via a Maven repository, and support hot updates. The third-party system integration module implements identity authentication based on OAuth2.0 and SSO, provides pre-built connectors and open APIs, and supports one-click reuse of external system events and custom integration templates via Webhook.

10. An application of the low-code flow design system according to any one of claims 1-9, characterized in that, It is suitable for enterprise-level process automation, cross-system integration and complex business rule configuration scenarios, enabling dynamic drag-and-drop design of processes, logical rule-based verification and cross-scenario reuse.

Images