A demand tracking system, method, device and medium based on programmable atlas

Abstract

The application discloses a demand tracking system and method based on programmable atlas, equipment and medium. The system comprises: a cross-domain programmable link design module, a graphical interface definition as a tracking template for system operation; a multi-modal atlas construction module, a multi-modal graph neural network performs semantic alignment and fusion on heterogeneous data, and constructs a unified knowledge graph serving the tracking template; an event-driven and data aggregation engine, based on the defined template rule, listens to state change events, and drives knowledge graph update; a full-link tracking view generation module, based on template definition, generates an end-to-end full-link tracking view; an intelligent link completion module, a prediction algorithm completes potential entity association; a risk insight module, predicts delivery risk and triggers alarm. The application solves the problems of data silos, view solidification and insight lag in the research and development tool chain, and realizes configurable and intelligent full-link tracking and active risk management from the business end to the technical end.

Description

Technical Field

[0001] This invention relates to the field of software development technology, and in particular to a requirement tracing system, method, device and medium based on programmable graphs. Background Technology

[0002] In modern DevOps software development processes, multiple stages such as business requirements, code development, test cases, continuous integration, and artifact deployment are typically supported by heterogeneous tools such as Jira, GitLab, Jenkins, and Nexus, resulting in significant "information gaps" and "management silos" throughout the entire development chain. Traditional requirements tracing systems or integration solutions have the following fundamental limitations:

[0003] 1. Shallow Data Association: Traditional solutions often rely on foreign keys, tags, or simple rules for static association, lacking a semantic association model between business requirements and technical artifacts, and thus unable to support complex queries and reasoning. Existing solutions may establish associations by matching requirement numbers in code submission information using regular expressions, but they cannot understand the semantic consistency between requirement descriptions and code changes;

[0004] 2. Fixed Process View: The existing system provides a fixed trace view structure, which cannot support different teams to flexibly define trace links and status rules according to their actual R&D models. Users can only passively accept predefined reports from the system and cannot build value stream views that reflect their own collaboration models as needed;

[0005] 3. Lagging insight capabilities: Existing technologies mainly provide queries and reports after problems occur, lacking the ability to proactively predict risks and pinpoint root causes based on real-time event streams and historical data. Managers often only respond passively after delivery delays or quality incidents occur.

[0006] In summary, existing solutions have limitations in terms of data association depth, process flexibility, and proactive insight. This invention differs from simple pipeline status management or product data association; its core lies in constructing a semantic R&D knowledge graph starting from business needs, and empowering users to flexibly define tracking links through visualization. Based on this graph, intelligent risk prediction and decision support are then achieved. It is applicable to the visualization of R&D value streams and intelligent prediction and decision support for delivery risks in modern, complex collaborative environments. Summary of the Invention

[0007] This invention provides a requirement tracking system, method, device, and medium based on a programmable graph. The system aims to address core issues such as the disconnect between business objectives and R&D execution, rigid management views, and a lack of intelligent predictive capabilities during the R&D process. Ultimately, it achieves end-to-end, real-time, semantic tracking and proactive risk management from the initial business requirement proposal to the final product deployment.

[0008] The technical solution adopted in this invention is as follows:

[0009] In a first aspect, the present invention provides a demand tracing system based on a programmable graph, the system comprising a cross-domain programmable link design module, a multimodal graph construction module, an event-driven and data aggregation engine, a full-link tracing view generation module, an intelligent link completion module, and a risk insight module; wherein;

[0010] The cross-domain programmable link design module provides a graphical interface that supports drag-and-drop operation, allowing users to customize the association rules, state transition conditions, and dependency logic between nodes based on business entity types and R&D entity types, so as to define and store reusable tracing templates as the basis for system operation.

[0011] The multimodal graph construction module is used to extract heterogeneous data from the business management system, code repository, continuous integration platform and artifact repository, and to perform semantic alignment and fusion of the multimodal features of the heterogeneous data through a multimodal graph neural network with a hierarchical attention mechanism including node-level attention units and relation-level attention units, so as to construct a unified knowledge graph that serves the tracking template.

[0012] The event-driven and data aggregation engine is used to monitor the state change events of each source system in real time based on the rules of the tracking template defined by the cross-domain programmable link design module, perform correlation, deduplication and context enhancement processing on multi-source events, and drive the dynamic update of the unified knowledge graph.

[0013] The end-to-end tracing view generation module is used to generate an end-to-end end-to-end tracing visualization view in real time based on the link logic defined by the tracing template and the updated unified knowledge graph.

[0014] The intelligent link completion module is used to automatically complete undefined entity associations based on the structure and historical time sequence pattern of the unified knowledge graph.

[0015] The risk insight module is used to predict delivery risks in the chain through anomaly detection algorithms, generate a visualized location path from risk phenomena to specific technical root causes, and trigger alarms, generate reports, or provide optimization suggestions based on the prediction results.

[0016] In some embodiments, the cross-domain orchestratable link design module includes a visual orchestration interface, an entity library management unit, and a rule definition and storage unit; wherein...

[0017] The visual arrangement interface supports drag-and-drop node selection and connection arrangement, and provides visual browsing and retrieval of the business entity library and the R&D entity library.

[0018] The entity library management unit is used to maintain business entity types and R&D entity types, and supports user-defined entity attributes and relationship templates;

[0019] The rule definition and storage unit is used to provide user-configurable rule options, including: regular expression matching rules for associating work items in the business management system with commit records in the code repository, supporting combined condition judgments of logical "AND" and logical "OR", and logical judgment rules for controlling the flow of R&D status; the rule definition and storage unit is also used to store user-configured rule combinations as reusable tracking templates.

[0020] In some embodiments, the multimodal map construction module includes a heterogeneous data extraction unit, a multimodal feature alignment unit, and a map construction and storage unit; wherein...

[0021] The heterogeneous data extraction unit is used to extract requirement information, code commit logs, pipeline event logs, and artifact metadata from the business management system, code repository, continuous integration platform, and artifact repository in real time or at regular intervals.

[0022] The multimodal feature alignment unit performs vectorized modeling and semantic alignment of text, attribute, and temporal multimodal features through the multimodal graph neural network with the hierarchical attention mechanism.

[0023] The graph construction and storage unit is used to store the aligned multimodal entities and relationships into the graph database to form a unified knowledge graph that supports semantic query and reasoning.

[0024] In some embodiments, the event-driven and data aggregation engine includes an adapter interface group, a stream processing and association unit, and a graph update triggering unit; wherein...

[0025] The adapter interface group is used to connect to various source systems and supports real-time monitoring of status changes through Webhook and API polling. The adapter interface group includes a GitLab Webhook adapter for monitoring code repository merge request events, a Jenkins API adapter for polling the status of the continuous integration platform pipeline, and a Jira Webhook adapter for receiving status updates from the business management system.

[0026] The stream processing and association unit performs real-time association, deduplication, context supplementation and normalization processing on multi-source events based on the rules of the tracking template;

[0027] The graph update triggering unit converts the aggregated event data into graph update operations, driving the real-time evolution of the knowledge graph.

[0028] In some embodiments, the end-to-end tracking view generation module includes a view rendering engine, an interaction control unit, and a state synchronization unit; wherein...

[0029] The view rendering engine generates an end-to-end link visualization view in real time based on the graph query results, and supports multiple view modes such as timeline, topology diagram, and Gantt chart.

[0030] The interactive control unit supports users to perform interactive operations such as zooming, filtering, drill-down, and highlighting on the view;

[0031] The status synchronization unit ensures that the view content and map data are synchronized in real time, supporting collaborative viewing by multiple users.

[0032] In some embodiments, the intelligent link completion module includes a link prediction unit; wherein,

[0033] The link prediction unit, based on graph embedding representation and historical collaboration patterns, uses a link prediction algorithm to automatically recommend potential code reviewers or related requirements for new code submissions.

[0034] In some embodiments, the risk insight module includes a time series graph analysis unit and an anomaly detection and root cause localization unit; wherein...

[0035] The time-series graph analysis unit analyzes the time consumption distribution and failure reasons of each link on similar R&D paths in the historical graph, establishes a prediction model, and quantifies the delivery delay probability and bottleneck links of the current link.

[0036] The anomaly detection and root cause localization unit identifies link risks through anomaly detection algorithms and generates a visualized localization path from risk phenomena to specific technical root causes.

[0037] Secondly, this invention provides a demand tracking method based on a programmable graph, comprising the following steps:

[0038] Step 1: Through a drag-and-drop graphical interface, the user can customize a cross-domain tracking template by dragging and connecting lines. The tracking template includes business nodes, R&D nodes, and the association and flow rules between nodes.

[0039] Step 2: Extract heterogeneous data from the business management system, code repository, continuous integration platform and artifact repository, and use hierarchical attention multimodal graph neural network to perform semantic alignment and fusion to construct a unified knowledge graph that is compatible with the tracking template;

[0040] Step 3: Using the aforementioned tracking template as the rule benchmark, monitor events from each source system in real time through the adapter interface group, and employ stream processing technology to perform correlation aggregation and context enhancement on the events;

[0041] Step 4: Dynamically update the unified knowledge graph based on the aggregated event data, and generate a full-link tracking visualization view in real time based on the tracking template;

[0042] Step 5: Based on the unified knowledge graph, complete the potential entity associations using the link prediction algorithm;

[0043] Step 6: Predict delivery risks through time series graph analysis models, generate root cause localization paths, and trigger alarm notifications, generate analysis reports, or provide process optimization suggestions based on risk prediction results.

[0044] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the steps of the above-described demand tracing method based on a programmable graph. The memory and the processor communicate with each other via an internal connection path.

[0045] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed, implements the steps of the above-described demand tracing method based on a programmable graph.

[0046] The advantages or beneficial effects of the above technical solutions include at least the following:

[0047] This invention achieves the following significant technical effects through the deep integration and synergistic effect of three core technologies: "cross-domain programmable link design", "multimodal graph large model", and "time series graph risk prediction":

[0048] 1. Provides user-programmable process definition and adaptive capabilities: Through graphical, low-code, programmable workflow design, it shifts the power to define tracking views and rules from system developers to frontline business and R&D personnel. User teams can visualize business processes. Figure 1 In this way, firstly, a unique tracking template is defined, and then the system automatically constructs the graph, listens for events, and renders the view based on the template, giving the system unprecedented flexibility and adaptability, and completely breaking through the rigid constraints of traditional fixed reports and views.

[0049] 2. Achieved end-to-end value stream transparency based on deep semantic understanding: Unlike traditional shallow associations based on foreign keys or tags, this invention achieves deep semantic fusion between business language and R&D data through a multimodal graph neural network with a hierarchical attention mechanism, guided by user-programmable semantic association rules. This constructs a queryable, reasonable, and computable "R&D digital twin," fundamentally solving the "black box" problem of value stream information caused by toolchain heterogeneity, and achieving semantic-level end-to-end transparency from business objectives to technology delivery.

[0050] 3. A fundamental paradigm shift from passive operation and maintenance to proactive intelligent prediction has been achieved: Based on the time series graph analysis model, the system can continuously analyze the dynamic time series patterns of R&D activities and their impact propagation effects in the graph topology, thereby proactively and quantitatively predicting the potential impact of anomalies in downstream R&D links on the delivery of upstream business objectives, and providing a visualized and interpretable localization path from risk phenomena to specific technical root causes, thus improving the quality and efficiency of R&D risk management.

[0051] 4. An auditability and decision support foundation built into the system has been constructed: all user-defined rules, the process of intelligent system decision-making, and the graph context on which they depend are recorded completely and in a structured manner, providing fine-grained operational traceability capabilities based on business semantics. At the same time, the unified semantic knowledge graph becomes a reliable data foundation supporting advanced management decisions such as R&D efficiency measurement, intelligent resource scheduling, and return on investment analysis.

[0052] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of this application will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the described drawings are only a part of the embodiments of the present invention, and not all of them. Based on the drawings in the present invention, those skilled in the art can obtain other related drawings without creative effort.

[0054] Figure 1 This is a schematic diagram of the structure of a demand tracing system based on a programmable graph according to the present invention;

[0055] Figure 2 This is a flowchart illustrating a demand tracing method based on a programmable graph according to the present invention. Detailed Implementation

[0056] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0057] As attached Figure 1As shown, this invention provides a demand tracing system based on a programmable graph. The system includes a cross-domain programmable link design module, a multimodal graph construction module, an event-driven and data aggregation engine, a full-link tracing view generation module, an intelligent link completion module, and a risk insight module. The following provides a detailed description of each module:

[0058] 1. A cross-domain orchestratable link design module provides a drag-and-drop graphical interface for users to customize association rules, state transition conditions, and dependency logic between nodes based on business entity types and R&D entity types. This allows for the definition and storage of reusable tracing templates that serve as the basis for system operation. Includes:

[0059] (1) Visual arrangement interface, supporting drag-and-drop node selection and connection arrangement interface, providing visual browsing and retrieval of business entity library and R&D entity library;

[0060] (2) Entity library management unit, used to maintain business entity types and R&D entity types, and supports user-defined entity attributes and relationship templates;

[0061] (3) Rule definition and storage unit, used to provide user-configurable rule options, including: regular expression matching rules for associating work items in the business management system with commit records in the code repository; supporting combined condition judgments of logical "AND" and logical "OR" for controlling the flow of R&D status; the rule definition and storage unit is also used to store user-configured rule combinations as reusable tracking templates.

[0062] 2. A multimodal graph construction module is used to extract heterogeneous data from business management systems, code repositories, continuous integration platforms, and artifact repositories. It then uses a multimodal graph neural network with a hierarchical attention mechanism including node-level and relation-level attention units to semantically align and fuse the multimodal features of the heterogeneous data, constructing a unified knowledge graph serving the tracking template. This includes:

[0063] (1) Heterogeneous data extraction unit, used to extract requirement information, code commit logs, pipeline event logs and artifact metadata from business management system, code repository, continuous integration platform and artifact repository in real time or at regular intervals;

[0064] (2) Multimodal feature alignment unit, through the multimodal graph neural network with the hierarchical attention mechanism, performs vectorized modeling and semantic alignment of multimodal features such as text, attributes, and time series;

[0065] (3) Graph construction and storage unit, used to store aligned multimodal entities and relations into the graph database to form a unified knowledge graph that supports semantic query and reasoning.

[0066] The hierarchical attention mechanism includes:

[0067] (1) Node-level attention unit, used to dynamically assign different attention weights to the text feature vector, attribute feature vector and time sequence feature vector of the same entity node, so as to realize the dynamic fusion of multimodal features;

[0068] (2) Relational attention unit, used to learn the strength weight of different semantic relation types in cross-domain entity association, to achieve non-linear and interpretable semantic association modeling.

[0069] 3. An event-driven and data aggregation engine, used to monitor state change events of each source system in real time based on the rules of the tracking template defined by the cross-domain programmable link design module, perform correlation, deduplication, and context enhancement processing on multi-source events, and drive the dynamic update of the unified knowledge graph. This includes:

[0070] (1) Adapter interface group, used to connect to various source systems, supports real-time monitoring of status changes through Webhook, API polling and other methods. The adapter interface group includes GitLab Webhook adapter for monitoring code repository merge request events, Jenkins API adapter for polling the status of the continuous integration platform pipeline, and Jira Webhook adapter for receiving status updates from the business management system.

[0071] (2) Stream processing and association unit, which performs real-time association, deduplication, context supplementation and normalization processing on multi-source events based on the rules of the tracking template;

[0072] (3) The graph update triggering unit converts the aggregated event data into graph update operations, driving the real-time evolution of the knowledge graph.

[0073] 4. A full-link tracing view generation module, used to generate an end-to-end full-link tracing visualization view in real time based on the link logic defined in the tracing template and the updated unified knowledge graph. This includes:

[0074] (1) View rendering engine, which generates end-to-end link visualization view in real time based on the graph query results, and supports multiple view modes such as timeline, topology map, and Gantt chart;

[0075] (2) Interactive control unit, which supports users to perform interactive operations such as zooming, filtering, drill-down, and highlighting on the view;

[0076] (3) Status synchronization unit to ensure that the view content is synchronized with the map data in real time and to support multi-user collaborative viewing.

[0077] 5. An intelligent link completion module, used to automatically complete undefined entity relationships based on the structure and historical time sequence patterns of the unified knowledge graph. This includes:

[0078] The link prediction unit, based on graph embedding representation and historical collaboration patterns, uses a link prediction algorithm to automatically recommend potential code reviewers or related requirements for new code submissions.

[0079] 6. Risk Insight Module: This module uses anomaly detection algorithms to predict delivery risks in the supply chain, generate a visualized path from risk phenomena to specific technical root causes, and trigger alarms, generate reports, or provide optimization suggestions based on the prediction results. Includes:

[0080] (1) The time-series graph analysis unit analyzes the time consumption distribution and failure reasons of each link on similar R&D paths in the historical graph, establishes a prediction model, and quantifies the delivery delay probability and bottleneck links of the current link.

[0081] (2) Anomaly detection and root cause localization unit: identify link risks through anomaly detection algorithm and generate a visualized localization path from risk phenomena to specific technical root causes.

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. This embodiment takes a complete R&D chain from business requirements to microservice deployment as an example to illustrate the system's working process.

[0083] Example: Full-chain tracking and risk prediction of the demand for "user points redemption for coupons" in e-commerce systems

[0084] 1. Cross-domain programmable link design (user-defined tracing logic)

[0085] The quality assurance team first defined a dedicated tracking template for "financial requirements" using the system's programmable interface:

[0086] (1) Drag and drop the "Business Requirements", "Code Submission", "Pipeline", "Security Scan Report", and "Production Deployment" nodes from the entity library to the canvas;

[0087] (2) Define association rules using lines:

[0088] "Business requirements" and "code submissions" are automatically linked using the "requirement ID mentioned in the submission message";

[0089] When the "pipeline" status is successful and the "security scan report" level is "low risk", the "production deployment" node will automatically unlock to an executable state.

[0090] (3) Save this template as “Financial Demand Release Access Control”.

[0091] 2. Data extraction and map initialization (executed by the system based on the template)

[0092] (1) Data extraction from the business and R&D sides:

[0093] Extract requirement FEA-101 from the Jira business management system: "Implement the function of redeeming coupons with user points" and its subtask TASK-1011: "Develop the points deduction interface". Extracted fields include: Requirement ID, Title, Description, Status, Creation Time, and Person in Charge.

[0094] Extract the code associated with TASK-1011 from the GitLab code repository and commit Commit-abc123. The commit message includes "fix: #TASK-1011 Complete the core logic of deducting points" and obtain the list of changed files.

[0095] Listen to the pipeline named "build-user-service" from the Jenkins continuous integration platform. This pipeline is triggered by the push event of Commit-abc123. Obtain the pipeline ID, status (start, in progress, success / failure), time spent in each stage (compilation, unit test, integration test), and output URL.

[0096] Retrieve the metadata of the Docker image user-service:1.2.0 generated after the pipeline is successfully built from the Nexus artifact repository, including image hash, build time, and dependency list.

[0097] (2) Multimodal graph construction and semantic alignment (template-based):

[0098] The system processes the extracted data based on the entity and relationship types specified in the defined "Financial Demand Posting Access Control" template:

[0099] The multimodal feature alignment unit performs targeted semantic alignment and fusion of the text, attribute, and temporal features of related entities (requirements, tasks, submissions, pipelines, and images) based on the association rules defined in the template (such as matching requirement IDs with submission information).

[0100] A multimodal graph neural network with a hierarchical attention mechanism is used to construct a graph, forming a unified knowledge subgraph that serves the specific tracking template, and then stored in the graph database.

[0101] 3. Event-driven and dynamic updates

[0102] (1) When the status of the Jenkins pipeline build-user-service changes from "in progress" to "failed":

[0103] (2) The event-driven engine captures the event in real time via Webhook;

[0104] (3) The stream processing unit queries the code submission Commit-abc123 and upstream requirement FEA-101 associated with the pipeline based on the "Financial Demand Release Access Control" template;

[0105] (4) The graph update triggering unit writes a (pipeline) - [status updated to] -> (failure) relationship to the graph database and triggers downstream rules.

[0106] 4. Automatic generation of end-to-end view

[0107] The system renders a topology map for requirement FEA-101 in real time, showing the complete link from requirement to failure pipeline and highlighting the faulty node.

[0108] 5. Intelligent link completion

[0109] Based on historical graph patterns, the system discovered that the developer of Commit-abc123 often collaborated with another developer, Developer-B, to modify security-related files, and automatically suggested adding Developer-B as a reviewer for this merge request.

[0110] 6. Risk Insight and Decision-Making

[0111] (1) The time-series graph analysis unit detected that in the past three releases of the user-service involving the "points" function, the integration test failed twice due to database connection timeouts. Based on the similar timeout exception pattern that appeared in the pipeline logs, the system proactively warned that "there is a 70% probability that this release will fail due to database problems", and located the specific test cases and error stacks of the historical failures in the graph, pushing them to the developers as root cause evidence;

[0112] (2) Based on the prediction results, the system will automatically perform one or a combination of the following operations:

[0113] Send risk warning notifications to project managers and development leads;

[0114] Generate a specific risk analysis section in the project weekly report;

[0115] In the visualization view, we recommend optimization measures such as "prioritizing database performance testing" or "splitting database changes with business changes".

[0116] As another embodiment, the present invention provides a demand tracking method based on a programmable graph, as shown in the appendix. Figure 2 As shown, the process includes the following steps:

[0117] S101: Through a drag-and-drop graphical interface, the user can customize a cross-domain tracking template by dragging and connecting lines. The tracking template includes business nodes, R&D nodes, and the association and flow rules between nodes.

[0118] S102: Extract heterogeneous data from the business management system, code repository, continuous integration platform and artifact repository, use hierarchical attention multimodal graph neural network for semantic alignment and fusion, and construct a unified knowledge graph that is compatible with the tracking template;

[0119] S103: Using the aforementioned tracking template as the rule benchmark, the system monitors events from each source system in real time through the adapter interface group, and uses stream processing technology to perform correlation aggregation and context enhancement on the events;

[0120] S104: Dynamically update the unified knowledge graph based on the aggregated event data, and generate a full-link tracking visualization view in real time based on the tracking template;

[0121] S105: Based on the unified knowledge graph, complete the potential entity associations using the link prediction algorithm;

[0122] S106: Predict delivery risks through time-series graph analysis models, generate root cause localization paths, and trigger alarm notifications, generate analysis reports, or provide process optimization suggestions based on risk prediction results.

[0123] In addition, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a program, the processor is used to run the program, and the program executes the steps of the demand tracing method based on programmable graphs provided by the present invention when it runs.

[0124] The device may also preferably include a communication interface for communicating and transmitting data with external devices.

[0125] It should be noted that the memory may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage.

[0126] Furthermore, the present invention also provides another computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the demand tracing method based on programmable graphs provided by the present invention.

[0127] It should be understood that the computer-readable storage medium is any data storage device capable of storing data or programs that can subsequently be read by a computer system. Examples of computer-readable storage media include read-only memory, random access memory, CD-ROM, HDD, DVD, magnetic tape, and optical data storage devices. Computer-readable storage media can also be distributed across network-coupled computer systems, enabling computer-readable code to be stored and executed in a distributed manner.

[0128] Program code contained on a computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), or any suitable combination thereof.

[0129] In some implementations, the computer-readable storage medium may be non-transitory.

[0130] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0131] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A demand tracing system based on programmable graphs, characterized in that, The system includes a cross-domain programmable link design module, a multimodal graph construction module, an event-driven and data aggregation engine, a full-link tracing view generation module, an intelligent link completion module, and a risk insight module; among which... The cross-domain programmable link design module provides a graphical interface that supports drag-and-drop operation, allowing users to customize the association rules, state transition conditions, and dependency logic between nodes based on business entity types and R&D entity types, so as to define and store reusable tracing templates as the basis for system operation. The multimodal graph construction module is used to extract heterogeneous data from the business management system, code repository, continuous integration platform and artifact repository, and to perform semantic alignment and fusion of the multimodal features of the heterogeneous data through a multimodal graph neural network with a hierarchical attention mechanism including node-level attention units and relation-level attention units, so as to construct a unified knowledge graph that serves the tracking template. The event-driven and data aggregation engine is used to monitor the state change events of each source system in real time based on the rules of the tracking template defined by the cross-domain programmable link design module, perform correlation, deduplication and context enhancement processing on multi-source events, and drive the dynamic update of the unified knowledge graph. The end-to-end tracing view generation module is used to generate an end-to-end end-to-end tracing visualization view in real time based on the link logic defined by the tracing template and the updated unified knowledge graph. The intelligent link completion module is used to automatically complete undefined entity associations based on the structure and historical time sequence pattern of the unified knowledge graph. The risk insight module is used to predict delivery risks in the link through anomaly detection algorithms, generate a visualized location path from risk phenomena to specific technical root causes, and trigger alarms, generate reports or provide optimization suggestions based on the prediction results. The cross-domain orchestratable link design module includes a visual orchestration interface, an entity library management unit, and a rule definition and storage unit; wherein... The visual arrangement interface supports drag-and-drop node selection and connection arrangement, and provides visual browsing and retrieval of the business entity library and the R&D entity library. The entity library management unit is used to maintain business entity types and R&D entity types, and supports user-defined entity attributes and relationship templates; The rule definition and storage unit is used to provide user-configurable rule options, including: regular expression matching rules for associating work items in the business management system with commit records in the code repository; supporting combined condition judgments of logical "AND" and logical "OR"; and logical judgment rules for controlling the flow of R&D status. The rule definition and storage unit is also used to store user-configured rule combinations as reusable tracking templates.

2. The demand tracking system based on programmable graphs according to claim 1, characterized in that, The multimodal graph construction module includes a heterogeneous data extraction unit, a multimodal feature alignment unit, and a graph construction and storage unit; wherein... The heterogeneous data extraction unit is used to extract requirement information, code commit logs, pipeline event logs, and artifact metadata from the business management system, code repository, continuous integration platform, and artifact repository in real time or at regular intervals. The multimodal feature alignment unit performs vectorized modeling and semantic alignment of text, attribute, and temporal multimodal features through the multimodal graph neural network with the hierarchical attention mechanism. The graph construction and storage unit is used to store the aligned multimodal entities and relationships into the graph database to form a unified knowledge graph that supports semantic query and reasoning.

3. The demand tracking system based on programmable graphs according to claim 1, characterized in that, The event-driven and data aggregation engine includes an adapter interface group, a stream processing and association unit, and a graph update triggering unit; wherein... The adapter interface group is used to connect to various source systems and supports real-time monitoring of status changes through Webhook and API polling. The adapter interface group includes a GitLab Webhook adapter for monitoring code repository merge request events, a Jenkins API adapter for polling the status of the continuous integration platform pipeline, and a Jira Webhook adapter for receiving status updates from the business management system. The stream processing and association unit performs real-time association, deduplication, context supplementation and normalization processing on multi-source events based on the rules of the tracking template; The graph update triggering unit converts the aggregated event data into graph update operations, driving the real-time evolution of the knowledge graph.

4. The demand tracking system based on programmable graphs according to claim 1, characterized in that, The end-to-end tracking view generation module includes a view rendering engine, an interaction control unit, and a state synchronization unit; wherein... The view rendering engine generates an end-to-end link visualization view in real time based on the graph query results, and supports multiple view modes such as timeline, topology diagram, and Gantt chart. The interactive control unit supports users to perform interactive operations such as zooming, filtering, drill-down, and highlighting on the view; The status synchronization unit ensures that the view content and map data are synchronized in real time, supporting collaborative viewing by multiple users.

5. The demand tracking system based on programmable graphs according to claim 1, characterized in that, The intelligent link completion module includes a link prediction unit; wherein... The link prediction unit, based on graph embedding representation and historical collaboration patterns, uses a link prediction algorithm to automatically recommend potential code reviewers or related requirements for new code submissions.

6. The demand tracking system based on programmable graphs according to claim 1, characterized in that, The risk insight module includes a time-series graph analysis unit and an anomaly detection and root cause localization unit; wherein... The time-series graph analysis unit analyzes the time consumption distribution and failure reasons of each link on similar R&D paths in the historical graph, establishes a prediction model, and quantifies the delivery delay probability and bottleneck links of the current link. The anomaly detection and root cause localization unit identifies link risks through anomaly detection algorithms and generates a visualized localization path from risk phenomena to specific technical root causes.

7. A demand tracing method based on a programmable graph, applied to the demand tracing system based on a programmable graph as described in any one of claims 1-6, characterized in that, include: Step 1: Through a drag-and-drop graphical interface, the user can customize a cross-domain tracking template by dragging and connecting lines. The tracking template includes business nodes, R&D nodes, and the association and flow rules between nodes. Step 2: Extract heterogeneous data from the business management system, code repository, continuous integration platform and artifact repository, and use hierarchical attention multimodal graph neural network to perform semantic alignment and fusion to construct a unified knowledge graph that is compatible with the tracking template; Step 3: Using the aforementioned tracking template as the rule benchmark, monitor events from each source system in real time through the adapter interface group, and employ stream processing technology to perform correlation aggregation and context enhancement on the events; Step 4: Dynamically update the unified knowledge graph based on the aggregated event data, and generate a full-link tracking visualization view in real time based on the tracking template; Step 5: Based on the unified knowledge graph, complete the potential entity associations using the link prediction algorithm; Step 6: Predict delivery risks through time series graph analysis models, generate root cause localization paths, and trigger alarm notifications, generate analysis reports, or provide process optimization suggestions based on risk prediction results.

8. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the steps of the demand tracking method based on a programmable graph as described in claim 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the steps of the demand tracing method based on a programmable graph as described in claim 7.

Images