A method and system for automatic archiving and traceability of AI-generated design resources

Abstract

This invention provides an automatic archiving and traceability method and system for AI-generated design resources. The method includes: 1. Receiving the design product archiving context from the Agent; 2. Identifying newly generated or user-selected design product files in the current session; 3. Categorizing the design products based on the context and file attributes; 4. Generating an archiving directory and file names based on the context; 5. Completing file metadata and generating an upload task list; 6. Calling the resource center interface based on the upload task scale and interface status to upload the design products to the automatically generated archiving directory; 7. Generating descriptive text for each file or group of files based on the session context; 8. Establishing a persistent resource index for upload results and session tasks; 9. Performing design resource review and retrieval based on search criteria in cross-Agent or cross-session scenarios to locate archived design products. This reduces design product loss and improves the reproducibility of design resources.

Description

Technical Field

[0001] This invention relates to the field of AI design conversation technology, and in particular to an automatic archiving and tracing method and system for AI-generated design resources. Background Technology

[0002] When using AI Agents, AI drawing tools, design aids, or automated design systems for design creation, users typically generate a large number of design deliverables through multiple rounds of dialogue or numerous operations, such as design drafts, design drawings, style reference images, partial modification images, resource packs, process versions, and final deliverables. These files may be stored in local temporary directories, download directories, tool output directories, session attachment directories, or project directories.

[0003] Existing file management or resource center systems typically offer basic capabilities such as uploading, downloading, creating directories, renaming, and deleting. However, these capabilities usually require users to manually select files, create directories, and manually enter filenames and comments. In AI-assisted design scenarios, users focus more on design intent, style adjustments, and solution comparisons, and often do not immediately archive standardized outputs after each round of design creation.

[0004] Therefore, existing technologies have the following drawbacks in this scenario: 1. Design artifacts are easily lost. AI-generated files may only be saved in a temporary directory or session directory, making them difficult to find again after the user closes the session or changes devices.

[0005] 2. Multiple design files can easily become mixed up. A design task may generate multiple rounds of process drafts, reference drawings, and the final draft. Without automatic categorization, these files will be mixed up in the same directory.

[0006] 3. The resource directory relies on manual maintenance. Users need to manually create directories by project, date, design theme, and version, which is tedious and prone to inconsistencies.

[0007] 4. The file lacks semantic explanation. When users see the file later, they often cannot know which requirement the file corresponds to, why it was generated, what design problem it solves, or whether it is the final version.

[0008] 5. Similar versions lack version relationships. Images generated in multiple rounds may have similar filenames or content, but they lack version numbers, round numbers, and differences, making it difficult to review the design evolution process later.

[0009] 6. Upload results lack binding with the design session. Even if a file is uploaded to the resource center, there is usually no clear association between the resource ID, remote path, and design task in the session.

[0010] 7. High comprehension costs during team collaboration. Other members may not know the source of the file, its applicable scenarios, the reason for its generation, or usage suggestions after receiving the file from the resource center.

[0011] 8. Difficult to retrieve resources across agents or sessions. After a user generates and uploads design artifacts in an agent window, the original session context may be unavailable when the user reopens a new agent window the next day. If the project, date, theme, version, description, and resource ID were not written as persistent metadata to the resource center or indexing system during upload, even if the new agent can access the resource center, it will be difficult to retrieve the corresponding resources based on natural language descriptions such as keywords like "final draft of the Chinese-style login page made yesterday". Summary of the Invention

[0012] To overcome the above problems, the purpose of this invention is to provide an automatic archiving and traceability method and system for AI-generated design resources, thereby reducing the loss of design products and improving the retrospectiveness of design resources.

[0013] This invention is implemented using the following scheme: an automatic archiving and traceability method for AI-generated design resources, comprising the following steps: Step 1: Receive the design artifact from the Agent session. The Agent calls the Resource Archiving Skill and passes the design artifact archiving context to the Skill. Step 2: Identify newly generated or user-selected design artifact files in this session; Step 3: Based on the context and file attributes, classify the design deliverables into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources; Step 4: Generate the archive directory and file name based on the context; Step 5: Complete the file metadata and generate an upload task list; Step 6: Based on the upload task size and interface status, call the resource center interface to upload the design artifacts to the automatically generated archive directory; Step 7: Generate descriptive text for each file or group of files based on the session context; Step 8: Establish a persistent resource index for upload results and session tasks; Step 9: In cross-Agent or cross-session scenarios, review design resources and locate archived design artifacts based on search criteria.

[0014] Furthermore, step 1 specifically involves: after the user completes the generation of design artifacts in any Agent session, the Agent calls the Resource Archiving Skill and passes the design artifact archiving context to the Skill; the archiving context is obtained by the Agent based on the current session, or supplemented by the user when calling the Skill; the archiving context includes the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path of the file or folder to be archived, session ID, task ID, current time, and user identity.

[0015] Furthermore, step 2 specifically involves: identifying the files generated by this design task based on the file path to be archived, file modification time, results returned by the generation tool, session attachments, user selection records, and file type; design outputs include image files, design draft files, compressed packages, reference materials, configuration files, preview images, or other resources related to the design task; identification is performed based on the file generation or modification time, file extension, design theme in the file name or directory name, user references to keywords in the session, and the output file list returned by the design tool, while excluding temporary cache, log files, empty files, invalid thumbnails, or files unrelated to the design task.

[0016] Furthermore, step 4 specifically involves: generating a resource center directory path based on the project name, date, design theme, deliverable type, and version stage; the directory structure can adopt the format of "Project Name / Date / Design Theme / Reference Material", or "Project Name / Date / Design Theme / Process Draft", or "Project Name / Date / Design Theme / Final Draft"; when the project name is missing, a directory can be generated based on the session title, current working directory, user account, default project, or user-supplemented information; when the design theme is missing, keywords can be extracted from the user requirement text to generate a theme name; when multiple design themes with the same name exist on the same date, task numbers or short session identifiers can be appended to avoid path conflicts; and a filename containing the date, theme, version number, and purpose can be generated.

[0017] Furthermore, step 5 specifically involves: generating upload metadata for each file to be archived, including the local file path, remote parent path, remote file path, file MD5, file extension, file creation time, file modification time, file access time, artifact type, version number, whether it is the final draft, and a summary of notes or descriptions; generating an upload task list based on the metadata; for design artifacts in folder format, recursively enumerating the files within the folder, generating a remote file path and upload task for each file; and marking empty files, files lacking extensions, or unsupported file types as unuploadable or requiring user confirmation.

[0018] Furthermore, step 6 specifically involves: calling the resource center client interface to upload files according to the upload task list; if the resource center supports uploading by path, then uploading files according to the automatically generated remote path is prioritized, and the resource center automatically creates an intermediate directory; if the resource center requires a parent resource ID, then the parent resource ID is searched according to the archive directory or supplemented by the user before uploading; for batch files or local folders, an upload scheduling strategy can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records; the upload scheduling strategy includes direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface has rate limiting or a high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation; during the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag of each file are recorded.

[0019] Furthermore, step 7 specifically involves generating explanatory text for each file or group of files based on the session context, design product type, and version relationship. The explanatory text includes the reason for the file's creation, the corresponding user requirements, the design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final draft, subsequent usage suggestions, and related reference materials or upstream versions.

[0020] Furthermore, step 8 specifically involves: after the upload is completed, establishing a binding relationship between the resource center's returned results and the session task. The index record can be saved as resource center metadata, remarks fields, supporting documentation, database records, or resource indexes accessible to the Agent. The index record includes session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether it has been finally confirmed, and its association with other versions or reference materials. Through this index, the files in the resource center are no longer isolated files, but are associated with the AI ​​design task, design purpose, and version evolution process. Even if the original Agent session ends or the user reopens a new Agent window, the Agent can still retrieve relevant resources through the resource center metadata or persistent resource index.

[0021] Furthermore, the method also includes step 10, exception handling, and user confirmation; specifically: in cases where the project name or design theme cannot be determined, multiple files are final drafts, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, the upload fails or the resource center returns an exception, the user explicitly requests to overwrite, delete, or rename archived resources, or the retrospective search finds multiple similar results, the user can be asked to confirm or supplement information to reduce the risk of misarchiving, misoverwriting, and version confusion.

[0022] An AI-generated design resource automatic archiving and traceability system, the system includes an archiving context receiving module, a design product identification module, a product type determination module, an archiving directory and file name generation module, a file metadata completion module, a resource center upload module, a design specification generation module, a persistent resource index module, and a cross-session review and retrieval module. The archiving context receiving module receives the design artifacts passed in by the Agent session. The Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill. The design product identification module identifies newly generated or user-selected design product files in this session. The product type determination module classifies design products into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources based on context and file attributes. The archive directory and file name generation module generates archive directory and file name names based on the context; The file metadata completion module completes the file metadata and generates an upload task list; The resource center upload module calls the resource center interface according to the upload task size and interface status to upload the design artifacts to the automatically generated archive directory; The design specification generation module generates specification text for each file or a group of files based on the session context. The persistent resource index module establishes a persistent resource index for upload results and session tasks; The cross-session review and retrieval module performs design resource review and locates archived design artifacts based on search criteria in cross-Agent or cross-session scenarios.

[0023] Furthermore, the implementation of the archiving context receiving module is as follows: After the user completes the generation of design artifacts in any Agent session, the Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill; the archiving context is obtained by the Agent based on the current session, or supplemented by the user when calling it; the archiving context includes the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path of the file or folder to be archived, session ID, task ID, current time and user identity.

[0024] Furthermore, the design product identification module is implemented as follows: Based on the file path to be archived, file modification time, results returned by the generation tool, session attachments, user selection records, and file type, the module identifies the files generated in this design task. Design products include image files, design draft files, compressed packages, reference materials, configuration files, preview images, or other resources related to the design task. Identification is performed based on the file generation or modification time, file extension, design theme in the filename or directory name, user references to keywords (such as "this image," "final version," "the folder just generated") in the session, and the output file list returned by the design tool. Temporary cache, log files, empty files, invalid thumbnails, or files unrelated to the design task are excluded.

[0025] Furthermore, the product type determination module specifically classifies design products into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources based on context and file attributes. Version numbers can be generated based on session rounds, user feedback, file generation order, and relationships between similar files. For example, the first image generated under the same design theme is marked as v1, the image generated after user requests adjustments is marked as v2, and after the user confirms "use this version," that version is marked as the final draft.

[0026] Furthermore, the implementation of the archive directory and filename generation module is as follows: It generates the resource center directory path based on the project name, date, design theme, deliverable type, and version stage; the directory structure can adopt the format "Project Name / Date / Design Theme / Reference Material", or "Project Name / Date / Design Theme / Process Draft", or "Project Name / Date / Design Theme / Final Draft"; when the project name is missing, it can generate a directory based on the session title, current working directory, user account, default project, or user-supplemented information; when the design theme is missing, it can extract keywords from the user requirement text to generate a theme name; when multiple design themes with the same name exist on the same date, a task number or short session identifier can be appended to avoid path conflicts; and it can generate filenames containing the date, theme, version number, and purpose. For example, "20260516_National Style Login Page Background_v2_Final Draft.png".

[0027] Furthermore, the implementation of the file metadata completion module is as follows: Upload metadata is generated for each file to be archived. This upload metadata includes the local file path, remote parent path, remote file path, file MD5, file extension, file creation time, file modification time, file access time, product type, version number, whether it is a final draft, and a summary of notes or descriptions. An upload task list is generated based on this metadata. For design products in folder format, the files within the folder are recursively enumerated, and a remote file path and upload task are generated for each file. Empty files, files lacking extensions, or unsupported file types are marked as unuploadable or pending user confirmation.

[0028] Furthermore, the implementation of the resource center upload module is as follows: Files are uploaded by calling the resource center client interface according to the upload task list. If the resource center supports upload by path, files are uploaded first based on the automatically generated remote path, and the resource center automatically creates intermediate directories. If the resource center requires a parent resource ID, the parent resource ID is searched in the archive directory or provided by the user before uploading. For batch files or local folders, an upload scheduling strategy can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records. Upload scheduling strategies include direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface experiences rate limiting or high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation. During the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag for each file are recorded.

[0029] Furthermore, the design specification generation module is implemented as follows: based on the session context, design product type, and version relationship, it generates explanatory text for each file or group of files. The explanatory text includes the reason for file generation, corresponding user requirements, design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final draft, subsequent usage suggestions, and related reference materials or upstream versions. For example, for a final login page background image, the explanatory text could record: "This file was generated based on the user's proposed visual requirements for a traditional Chinese style login page, and is used as a candidate background design for the login page; compared to version v1, brightness has been increased and dark white space has been reduced; the user confirmed this version as the final draft in the 4th session." This explanatory text can serve as part of resource center notes, accompanying explanatory documents, or the upload result index.

[0030] Furthermore, the persistent resource index module is implemented as follows: After the upload is completed, the results returned by the resource center are bound to the session task. The index record can be saved as resource center metadata, remarks fields, supporting documentation, database records, or resource indexes accessible to the Agent. The index record includes session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether it has been finally confirmed, and its association with other versions or reference materials. Through this index, the files in the resource center are no longer isolated files, but are associated with the AI ​​design task, design purpose, and version evolution process. Even if the original Agent session ends or the user reopens a new Agent window, the Agent can still retrieve relevant resources through the resource center metadata or the persistent resource index.

[0031] Furthermore, the cross-session review retrieval module, when a user initiates a review request in the same Agent session or a new Agent session, such as "Find the final draft of the Chinese-style login page generated yesterday," "Where is the bright version from last time," or "Send me the login page design resources uploaded before," queries the resource center metadata or persistent archive index based on time, project, design theme, product type, version description, and user identity in natural language. It can return the resource center path, resource ID, preview image, descriptive text, and a list of related versions. If multiple candidate resources exist, they can be sorted by time, final draft mark, user confirmation status, keyword matching degree, or descriptive text similarity, allowing the user to select.

[0032] Furthermore, the system also includes an exception handling module. The exception handling module is implemented as follows: in cases where the project name or design theme cannot be determined, multiple files are final drafts, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, upload fails or the resource center returns an exception, the user explicitly requests to overwrite, delete or rename archived resources, or a retrospective search finds multiple similar results, the user can be asked to confirm or supplement information to reduce the risk of misarchiving, misoverwriting and version confusion.

[0033] The beneficial effects of this invention are: 1. Reduced loss of design products. After design products are generated, they are automatically identified and uploaded to the resource center, reducing the risk of files being forgotten and left in temporary directories or session attachments.

[0034] 2. Improve archiving consistency. Automatically generate catalogs based on project, date, design theme, deliverable type, and version stage, reducing manual catalog creation and naming inconsistencies.

[0035] 3. Reduce user operating costs. Users no longer need to manually select files one by one, create directories, fill in notes, and record version relationships.

[0036] 4. Improve the reviewability of design resources. Automatically generate explanatory text, recording the reason for the file's creation, the corresponding requirements, design style, usage scenario, and version differences.

[0037] 5. Reduce the risk of version confusion. Mark process drafts, revised drafts, candidate drafts, and final drafts based on session rounds and user feedback, making multi-round design results easier to manage.

[0038] 6. Improve team collaboration efficiency. When team members retrieve files from the resource center, they can simultaneously see explanatory text and version information, making it easier to understand the purpose of the resources.

[0039] 7. Supports cross-Agent and cross-session natural language retrieval. By establishing a persistent resource index, users can locate resources using natural language descriptions such as "yesterday's version," "final draft," or "the background of a project's login page," even if they reopen a new Agent window.

[0040] 8. Improve the data quality of the resource center. When uploading, complete the MD5 hash, file extension, file time, remote path, and description text to make the design files in the resource center more standardized and traceable. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0042] Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0043] The invention will now be further described with reference to the accompanying drawings.

[0044] This invention proposes an automatic archiving and traceability method and system for AI-generated design resources. After a user generates design artifacts using an AI Agent or AI design tool, the Agent passes the archiving context—including the file or folder to be archived, design theme, project name, task description, version stage, and user confirmation status—to the resource archiving Skill. Based on these contexts, the design deliverables generated by this design task are identified, and reference materials, process drafts, revised drafts, candidate drafts, and final drafts are distinguished. Then, a resource center archive directory and file naming rules are automatically generated based on the project, date, design theme, file type, and version stage. Metadata such as MD5, extension, creation time, modification time, and access time are completed for each file, and an upload task list is generated. Based on the resource center interface capabilities, number of files, file size, directory conflict risk, and rate limiting status, a direct upload, batch parallel upload, low-concurrency upload, or serial upload strategy is dynamically selected. After upload, explanatory text is automatically generated based on the task description and deliverable type, recording the reason for file creation, corresponding requirements, design style, usage scenario, differences from the previous version, and whether it is recommended as the final draft. Finally, the resource ID, remote path, file MD5, explanatory text, session ID, task ID, user ID, version relationship, and keyword tags are written into the resource center metadata or persistent archive index, allowing subsequent agent sessions to review the data through natural language or conditional searches.

[0045] This invention is not simply a file upload, nor is it simply a resource center directory management. Instead, it links the "design intent, generation process, file products, resource center objects, and cross-session retrieval index" in the AI ​​design task, so that the design products are transformed from temporary session files into persistent, descriptive, and cross-agent retrievable resources.

[0046] This invention enhances the original scenario by adding mechanisms for design artifact identification, context directory generation, date and version marking, design specification generation, upload scheduling strategy selection, binding of upload results to the persistent index of the resource center, and cross-Agent subsequent review and retrieval. Its innovation lies not in simple interface encapsulation; the addition of cross-session persistent archiving and retrieval mechanisms better demonstrates the technical problems, methods, and verifiable technical effects.

[0047] Please see Figure 1 As shown, an automatic archiving and traceability method for AI-generated design resources is described, which includes the following steps: Step 1: Receive the design artifact from the Agent session. The Agent calls the Resource Archiving Skill and passes the design artifact archiving context to the Skill. Specifically, after a user completes the generation of design artifacts within a given Agent session, the Agent invokes the Resource Archiving Skill, passing the design artifact archiving context to the Skill. The archiving context can be generated by the Agent based on the current session or supplemented by the user during the invocation. The archiving context includes at least the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path to the file or folder to be archived, session ID, task ID, current time, and user identity. For example, after a user completes the design of a "Chinese-style login page background image," the Agent can pass the following to the Resource Archiving Skill: the folder to be uploaded, the design theme as "Chinese-style login page background image," the project name as "a game login page," the task description as "candidate solution for login page background," and the user confirmation status as "final draft."

[0048] Step 2: Identify newly generated or user-selected design artifact files in this session; Specifically, the system identifies files generated during this design task based on file path, file modification time, results returned by the generation tool, session attachments, user selection records, and file type. Design artifacts may include image files, design drafts, compressed files, reference materials, configuration files, preview images, or other resources related to the design task. Identification can be based on file generation or modification time, file extension, design theme in filenames or directory names, user references to "this image," "final version," or "the folder just generated" in the session, and the output file list returned by the design tool. Temporary caches, log files, empty files, invalid thumbnails, or files unrelated to the design task are excluded.

[0049] Step 3: Based on the context and file attributes, classify the design deliverables into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources; Specifically, based on context and file attributes, design deliverables are categorized into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources. Version numbers can be generated based on session rounds, user feedback, file generation order, and relationships between similar files. For example, the first image generated under the same design theme is marked as v1, the image generated after user requests adjustments is marked as v2, and once the user confirms "use this version," that version is marked as the final draft.

[0050] Step 4: Generate the archive directory and file name based on the context; Specifically, the resource center directory path is generated based on the project name, date, design theme, deliverable type, and version stage. The directory structure can take the form of "Project Name / Date / Design Theme / Reference Materials," "Project Name / Date / Design Theme / Process Draft," or "Project Name / Date / Design Theme / Final Draft." When the project name is missing, a directory can be generated based on the session title, current working directory, user account, default project, or user-supplemented information. When the design theme is missing, keywords can be extracted from the user's requirements text to generate a theme name. When multiple design themes with the same name exist on the same date, a task number or short session identifier can be appended to avoid path conflicts. Filenames containing the date, theme, version number, and purpose can be generated, such as "20260516_National Style Login Page Background_v2_Final Draft.png".

[0051] Step 5: Complete the file metadata and generate an upload task list; Specifically, upload metadata is generated for each file to be archived. Upload metadata includes at least the local file path, remote parent path, remote file path, file MD5 hash, file extension, file creation time, file modification time, file access time, artifact type, version number, whether it is the final draft, and a summary of notes or descriptions. An upload task list is generated based on this metadata. For design artifacts in folder format, the files within the folder are recursively enumerated, and a remote file path and upload task are generated for each file. Empty files, files lacking extensions, or unsupported file types are marked as unuploadable or pending user confirmation.

[0052] Step 6: Based on the upload task size and interface status, call the resource center interface to upload the design artifacts to the automatically generated archive directory; Specifically, the process involves calling the resource center client interface to upload files based on the upload task list. If the resource center supports uploads by path, files are uploaded first using the automatically generated remote path, and the resource center automatically creates intermediate directories. If the resource center requires a parent resource ID, the parent resource ID is searched in the archive directory or provided by the user before uploading. For batch files or local folders, upload scheduling strategies can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records. Upload scheduling strategies can include direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface experiences rate limiting or high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation. During the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag for each file are recorded.

[0053] Step 7: Generate descriptive text for each file or group of files based on the session context; Specifically, based on the conversation context, design artifact type, and version relationship, explanatory text is generated for each file or group of files. This explanatory text can include the reason for the file's creation, the corresponding user requirements, the design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final version, suggestions for subsequent use, and related reference materials or upstream versions. For example, for a final login page background image, the explanatory text could record: "This file was generated based on the user's proposed visual requirements for a traditional Chinese style login page and is used as a candidate background design for the login page; compared to version v1, brightness has been improved and dark white space has been reduced; the user confirmed this version as the final version in the 4th round of the conversation." This explanatory text can be used as part of resource center notes, accompanying explanatory documents, or the upload results index.

[0054] Step 8: Establish a persistent resource index for upload results and session tasks; Specifically, after the upload is complete, the results returned by the resource center are bound to the session task. Index records can be stored as resource center metadata, memo fields, accompanying documentation, database records, or resource indexes accessible to the Agent. Index records may include session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether final confirmation is required, and relationships with other versions or reference materials. Through this index, files in the resource center are no longer isolated files, but are associated with AI design tasks, design purposes, and version evolution processes. Even if the original Agent session ends or the user reopens a new Agent window, subsequent Agents can still retrieve relevant resources through resource center metadata or persistent indexes.

[0055] Step 9: In cross-Agent or cross-session scenarios, review design resources and locate archived design artifacts based on search criteria.

[0056] Specifically, when a user initiates a review request within the same Agent session or a new Agent session, such as "Find the final draft of the Chinese-style login page generated yesterday," "Where is the bright version from last time," or "Send me the login page design resources you uploaded before," the system queries the resource center metadata or persistent archive index based on the time, project, design theme, deliverable type, version description, and user identity in natural language processing. It can return the resource center path, resource ID, preview image, descriptive text, and a list of related versions. If multiple candidate resources exist, they can be sorted by time, final draft marker, user confirmation status, keyword matching degree, or descriptive text similarity, allowing the user to select.

[0057] Step 10: Exception handling and user confirmation.

[0058] In cases where the project name or design theme cannot be determined, multiple files may be the final version, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, the upload fails or the resource center returns an error, the user explicitly requests to overwrite, delete or rename archived resources, or a retrospective search finds multiple similar results, the user can be asked to confirm or supplement information to reduce the risk of misarchiving, misoverwriting and version confusion.

[0059] Optional Implementation Example 1: Batch archiving of design folders.

[0060] In one embodiment, a user requests the system to archive a local design output directory. The Agent passes the directory path, project name, design theme, and task description to the Resource Archiving Skill. The system identifies all images and design files in that directory, filtering out temporary and unsupported files. The system generates a remote directory based on the directory name, design theme, and current date, and generates a remote path for each file. Subsequently, the system selects direct upload, batch parallel upload, or low-concurrency upload based on the number of files and interface status, and returns the number of successful uploads, the number of failed uploads, a list of resource IDs, and the reason for each failure.

[0061] Optional Implementation Example 2: Automatic archiving after final draft confirmation.

[0062] In another embodiment, the user expresses keywords such as "use this version," "save this version," or "this is the final draft" in the session. The system marks the confirmed file in the current session as the final draft and archives it in the "Final Draft" directory, while retaining the relevant process drafts in the "Process Draft" directory. The description text records the session round and confirmation statement of the user confirming the final draft.

[0063] Optional Implementation Example 3: Review of Multiple Version Design Resources

[0064] In another embodiment, the user subsequently requests to view the "login page design draft generated yesterday." The system queries the archive index based on the date, design theme, and user identity, returning a list of process drafts, revised drafts, and the final draft, and displaying a description, remote path, and resource ID for each version to help the user review the design evolution process.

[0065] Optional Implementation Example 4: Retrieving Historical Design Artifacts Across Agent Sessions.

[0066] In another embodiment, on the first day, the user generates and archives a set of "traditional Chinese style login page background images" in Agent A. The system uploads the final draft, process draft, and reference materials to the resource center and writes persistent indexes such as project, date, theme, version, description, and resource ID. On the second day, the user reopens Agent B and enters "Help me find the final draft of the traditional Chinese style login page I made yesterday." Agent B invokes the review and retrieval capabilities of the resource archiving skill. The system queries the resource center metadata or persistent indexes based on the current user, the time condition "yesterday," the theme keyword "traditional Chinese style login page," and the product type "final draft," returning the corresponding resource ID, remote path, description text, and a list of related versions.

[0067] like Figure 2 As shown, the present invention provides an automatic archiving and traceability system for AI-generated design resources. The system includes an archiving context receiving module, a design product identification module, a product type determination module, an archiving directory and file name generation module, a file metadata completion module, a resource center upload module, a design specification generation module, a persistent resource index module, and a cross-session review and retrieval module. The archiving context receiving module receives the design artifacts passed in by the Agent session. The Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill. The design product identification module identifies newly generated or user-selected design product files in this session. The product type determination module classifies design products into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources based on context and file attributes. The archive directory and file name generation module generates archive directory and file name names based on the context; The file metadata completion module completes the file metadata and generates an upload task list; The resource center upload module calls the resource center interface according to the upload task size and interface status to upload the design artifacts to the automatically generated archive directory; The design specification generation module generates specification text for each file or a group of files based on the session context. The persistent resource index module establishes a persistent resource index for upload results and session tasks; The cross-session review and retrieval module performs design resource review and locates archived design artifacts based on search criteria in cross-Agent or cross-session scenarios.

[0068] The archiving context receiving module is implemented as follows: After the user completes the generation of design artifacts in any Agent session, the Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill; the archiving context is obtained by the Agent based on the current session, or supplemented by the user when calling it; the archiving context includes the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path of the file or folder to be archived, session ID, task ID, current time and user identity.

[0069] The design product identification module is implemented as follows: It identifies files generated by the current design task based on the file path to be archived, file modification time, results returned by the generation tool, session attachments, user selection records, and file type. Design products include image files, design draft files, compressed packages, reference materials, configuration files, preview images, or other resources related to the design task. The module identifies files based on their generation or modification time, file extension, design theme in filenames or directory names, user references to keywords in the session (such as "this image," "final version," "the folder just generated"), and the output file list returned by the design tool, while excluding temporary cache, log files, empty files, invalid thumbnails, or files unrelated to the design task.

[0070] The product type determination module specifically classifies design products into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources based on context and file attributes. The system can generate version numbers based on session rounds, user feedback, file generation order, and relationships between similar files. For example, the first image generated under the same design theme is marked as v1, the image generated after user requests adjustments is marked as v2, and after the user confirms "use this version," that version is marked as the final draft.

[0071] The implementation of the archive directory and filename generation module is as follows: It generates the resource center directory path based on the project name, date, design theme, deliverable type, and version stage; the directory structure can adopt the format "Project Name / Date / Design Theme / Reference Material", or "Project Name / Date / Design Theme / Process Draft", or "Project Name / Date / Design Theme / Final Draft"; when the project name is missing, it can generate a directory based on the session title, current working directory, user account, default project, or user-supplemented information; when the design theme is missing, it can extract keywords from the user's requirement text to generate a theme name; when multiple design themes with the same name exist on the same date, a task number or short session identifier can be appended to avoid path conflicts; and it can generate filenames containing the date, theme, version number, and purpose. For example, "20260516_National Style Login Page Background_v2_Final Draft.png".

[0072] The file metadata completion module is implemented as follows: Upload metadata is generated for each file to be archived. This upload metadata includes the local file path, remote parent path, remote file path, file MD5, file extension, file creation time, file modification time, file access time, artifact type, version number, whether it is a final draft, and a summary of notes or descriptions. An upload task list is generated based on this metadata. For design artifacts in folder format, the files within the folder are recursively enumerated, and a remote file path and upload task are generated for each file. Empty files, files lacking extensions, or unsupported file types are marked as unuploadable or pending user confirmation.

[0073] The implementation of the resource center upload module is as follows: Files are uploaded by calling the resource center client interface according to the upload task list. If the resource center supports upload by path, files are uploaded first based on the automatically generated remote path, and the resource center automatically creates intermediate directories. If the resource center requires a parent resource ID, the parent resource ID is searched based on the archive directory or supplemented by the user before uploading. For batch files or local folders, an upload scheduling strategy can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records. Upload scheduling strategies include direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface experiences rate limiting or high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation. During the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag for each file are recorded.

[0074] The design description generation module is implemented as follows: Based on the session context, design artifact type, and version relationship, it generates description text for each file or group of files. The description text includes the reason for file generation, corresponding user requirements, design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final draft, subsequent usage suggestions, and related reference materials or upstream versions. For example, for a final login page background image, the description text could record: "This file was generated based on the user's proposed visual requirements for a traditional Chinese style login page, and is used as a candidate background design for the login page; compared to version v1, brightness has been improved and dark white space has been reduced; the user confirmed this version as the final draft in the 4th session." This description text can serve as part of resource center notes, accompanying description documents, or the upload result index.

[0075] The persistent resource index module is implemented as follows: After the upload is completed, the results returned by the resource center are bound to the session task. The index record can be saved as resource center metadata, remarks fields, supporting documentation, database records, or resource indexes accessible to the Agent. The index record includes session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether it has been finally confirmed, and its association with other versions or reference materials. Through this index, the files in the resource center are no longer isolated files, but are associated with the AI ​​design task, design purpose, and version evolution process. Even if the original Agent session ends or the user reopens a new Agent window, the Agent can still retrieve relevant resources through the resource center metadata or the persistent resource index.

[0076] The cross-session review retrieval module, when a user initiates a review request in the same agent session or a new agent session, such as "Find the final draft of the Chinese-style login page generated yesterday," "Where is the bright version from last time," or "Send me the login page design resources I uploaded before," queries the resource center metadata or persistent archive index based on time, project, design theme, product type, version description, and user identity in natural language. The system can return the resource center path, resource ID, preview image, descriptive text, and a list of related versions. If multiple candidate resources exist, the system can sort them by time, final draft mark, user confirmation status, keyword matching degree, or descriptive text similarity, allowing the user to select.

[0077] The system also includes an exception handling module, which is implemented as follows: in cases where the project name or design theme cannot be determined, multiple files are final drafts, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, upload fails or the resource center returns an exception, the user explicitly requests to overwrite, delete or rename archived resources, or a retrospective search finds multiple similar results, the system can ask the user to confirm or supplement information to reduce the risk of misarchiving, misoverwriting and version confusion.

[0078] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A method for automatically archiving and tracing AI-generated design resources, characterized in that: Includes the following steps: Step 1: Receive the design artifact from the Agent session. The Agent calls the Resource Archiving Skill and passes the design artifact archiving context to the Skill. Step 2: Identify newly generated or user-selected design artifact files in this session; Step 3: Based on the context and file attributes, classify the design deliverables into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources; Step 4: Generate the archive directory and file name based on the context; Step 5: Complete the file metadata and generate an upload task list; Step 6: Based on the upload task size and interface status, call the resource center interface to upload the design artifacts to the automatically generated archive directory; Step 7: Generate descriptive text for each file or group of files based on the session context; Step 8: Establish a persistent resource index for upload results and session tasks; Step 9: In cross-Agent or cross-session scenarios, review design resources and locate archived design artifacts based on search criteria.

2. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: Step 1 specifically involves the following steps: After the user completes the generation of the design artifact in any Agent session, the Agent calls the Resource Archiving Skill and passes the design artifact archiving context to the Skill. The archive context is compiled by the Agent based on the current session, or supplemented by the user when it is invoked; the archive context includes the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path of the file or folder to be archived, session ID, task ID, current time and user identity.

3. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: Step 2 specifically involves: identifying the files generated by this design task based on the file path to be archived, file modification time, results returned by the generation tool, session attachments, user selection records, and file type; design deliverables include image files, design draft files, compressed packages, reference materials, configuration files, preview images, or other resources related to the design task. The identification is based on the file's creation or modification time, file extension, design theme in the file name or directory name, user references to keywords in the session, and the list of output files returned by the design tool, while excluding temporary caches, log files, empty files, invalid thumbnails, or files unrelated to the design task.

4. The method for automatic archiving and traceability of AI-generated design resources according to claim 2, characterized in that: Step 4 specifically involves: generating a resource center directory path based on the project name, date, design theme, deliverable type, and version stage; the directory structure can be in the form of "Project Name / Date / Design Theme / Reference Material", "Project Name / Date / Design Theme / Process Draft", or "Project Name / Date / Design Theme / Final Draft"; when the project name is missing, a directory can be generated based on the session title, current working directory, user account, default project, or user-supplemented information; when the design theme is missing, keywords can be extracted from the user requirement text to generate a theme name; when multiple design themes with the same name exist on the same date, task numbers or short session identifiers can be appended to avoid path conflicts; and a filename containing the date, theme, version number, and purpose can be generated.

5. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: Step 5 specifically involves: generating upload metadata for each file to be archived, including local file path, remote parent path, remote file path, file MD5, file extension, file creation time, file modification time, file access time, artifact type, version number, whether it is the final draft, and remarks or explanatory summary; and generating an upload task list based on the metadata. For design artifacts in the form of folders, recursively enumerate the files in the folder and generate a remote file path and upload task for each file; for empty files, files missing extensions, or unsupported file types, mark them as unuploadable or pending user confirmation.

6. The method for automatic archiving and traceability of AI-generated design resources according to claim 5, characterized in that: Step 6 specifically involves: Uploading files by calling the resource center client interface according to the upload task list. If the resource center supports uploading by path, files are uploaded first based on the automatically generated remote path, and the resource center automatically creates an intermediate directory. If the resource center requires a parent resource ID, the parent resource ID is searched in the archive directory or provided by the user before uploading. For batch files or local folders, an upload scheduling strategy can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records. Upload scheduling strategies include direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface experiences rate limiting or high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation. During the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag for each file are recorded.

7. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: Step 7 specifically involves generating descriptive text for each file or group of files based on the session context, design artifact type, and version relationship. The explanatory text includes the reason for the document's creation, the corresponding user needs, the design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final version, suggestions for subsequent use, and related reference materials or upstream versions.

8. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: Step 8 specifically involves: After the upload is completed, the results returned by the resource center are bound to the session task. The index record can be saved as resource center metadata, remarks fields, supporting documentation, database records, or resource indexes accessible to the Agent. The index record includes session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether it has been finally confirmed, and its association with other versions or reference materials. Through this index, the files in the resource center are no longer isolated files, but are associated with the AI ​​design task, design purpose, and version evolution process. Even if the original Agent session ends or the user reopens a new Agent window, the Agent can still retrieve relevant resources through the resource center metadata or persistent resource index.

9. The method for automatic archiving and traceability of AI-generated design resources according to claim 1, characterized in that: The method also includes step 10, exception handling, and user confirmation; specifically, in cases where the project name or design theme cannot be determined, multiple files are final drafts, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, the upload fails or the resource center returns an exception, the user explicitly requests to overwrite, delete, or rename archived resources, or the retrospective search finds multiple similar results, the user can be asked to confirm or supplement information to reduce the risk of misarchiving, misoverwriting, and version confusion.

10. An automatic archiving and traceability system for AI-generated design resources, characterized in that: The system includes an archive context receiving module, a design product identification module, a product type determination module, an archive directory and file name generation module, a file metadata completion module, a resource center upload module, a design specification generation module, a persistent resource index module, and a cross-session review retrieval module. The archiving context receiving module receives the design artifacts passed in by the Agent session. The Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill. The design product identification module identifies newly generated or user-selected design product files in this session. The product type determination module classifies design products into reference materials, process drafts, revised drafts, candidate drafts, final drafts, or auxiliary resources based on context and file attributes. The archive directory and file name generation module generates archive directory and file name names based on the context; The file metadata completion module completes the file metadata and generates an upload task list; The resource center upload module calls the resource center interface according to the upload task size and interface status to upload the design artifacts to the automatically generated archive directory; The design specification generation module generates specification text for each file or a group of files based on the session context. The persistent resource index module establishes a persistent resource index for upload results and session tasks; The cross-session review and retrieval module performs design resource review and locates archived design artifacts based on search criteria in cross-Agent or cross-session scenarios.

11. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The specific implementation of the archiving context receiving module is as follows: after the user completes the generation of design artifacts in any Agent session, the Agent calls the resource archiving Skill and passes the design artifact archiving context to the Skill; The archive context is compiled by the Agent based on the current session, or supplemented by the user when it is invoked; the archive context includes the user's original design requirements, project name, design theme, style keywords, design purpose, modification instructions, generation round, user confirmation feedback, path of the file or folder to be archived, session ID, task ID, current time and user identity.

12. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The design product identification module is implemented as follows: based on the file path to be archived, file modification time, results returned by the generation tool, session attachments, user selection records, and file type, the files generated in this design task are identified; design products include image files, design draft files, compressed packages, reference materials, configuration files, preview images, or other resources related to the design task. The identification is based on the file's creation or modification time, file extension, design theme in the file name or directory name, user references to keywords in the session, and the list of output files returned by the design tool, while excluding temporary caches, log files, empty files, invalid thumbnails, or files unrelated to the design task.

13. The automatic archiving and traceability system for AI-generated design resources according to claim 11, characterized in that: The implementation of the archive directory and filename generation module is as follows: It generates the resource center directory path based on the project name, date, design theme, deliverable type, and version stage; the directory structure can adopt the format "Project Name / Date / Design Theme / Reference Material", or "Project Name / Date / Design Theme / Process Draft", or "Project Name / Date / Design Theme / Final Draft"; when the project name is missing, it can generate a directory based on the session title, current working directory, user account, default project, or user-supplemented information; when the design theme is missing, it can extract keywords from the user requirement text to generate a theme name; when multiple design themes with the same name exist on the same date, a task number or short session identifier can be appended to avoid path conflicts; and it can generate filenames containing the date, theme, version number, and purpose.

14. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The file metadata completion module is implemented as follows: Upload metadata is generated for each file to be archived. This upload metadata includes the local file path, remote parent path, remote file path, file MD5, file extension, file creation time, file modification time, file access time, output type, version number, whether it is a final draft, and a summary of notes or descriptions. An upload task list is then generated based on this metadata. For design artifacts in the form of folders, recursively enumerate the files in the folder and generate a remote file path and upload task for each file; for empty files, files missing extensions, or unsupported file types, mark them as unuploadable or pending user confirmation.

15. The automatic archiving and traceability system for AI-generated design resources according to claim 14, characterized in that: The implementation of the resource center upload module is as follows: Files are uploaded by calling the resource center client interface according to the upload task list. If the resource center supports upload by path, files are uploaded first based on the automatically generated remote path, and the resource center automatically creates intermediate directories. If the resource center requires a parent resource ID, the parent resource ID is searched based on the archive directory or supplemented by the user before uploading. For batch files or local folders, an upload scheduling strategy can be selected based on the number of files, file size, directory conflict risk, resource center interface capabilities, interface rate limiting status, and historical failure records. Upload scheduling strategies include direct upload of a small number of files, batch parallel upload of a large number of files, limiting concurrency when there is a naming conflict risk in the same remote directory, downgrading to low concurrency or serial upload when the resource center interface experiences rate limiting or high failure rate, and adding failed upload files to a retry queue or a list awaiting manual confirmation. During the upload process, the upload status, resource center return code, return information, resource ID, failure reason, and retry flag for each file are recorded.

16. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The design specification generation module is implemented as follows: based on the session context, design artifact type, and version relationship, it generates specification text for each file or group of files. The explanatory text includes the reason for the document's creation, the corresponding user needs, the design theme, style keywords, usage scenarios, differences from the previous version, whether it is recommended as the final version, suggestions for subsequent use, and related reference materials or upstream versions.

17. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The persistent resource index module is implemented as follows: After the upload is completed, the results returned by the resource center are bound to the session task. The index record can be saved as resource center metadata, remarks fields, supporting documentation, database records, or resource indexes accessible to the Agent. The index record includes session ID, task ID, user ID, project name, design theme, local file path, remote resource path, resource ID, file MD5, artifact type, version number, description text, upload time, upload environment, upload status, keyword tags, whether it has been finally confirmed, and its association with other versions or reference materials. Through this index, the files in the resource center are no longer isolated files, but are associated with the AI ​​design task, design purpose, and version evolution process. Even if the original Agent session ends or the user reopens a new Agent window, the Agent can still retrieve relevant resources through the resource center metadata or the persistent resource index.

18. The automatic archiving and traceability system for AI-generated design resources according to claim 10, characterized in that: The system also includes an exception handling module, which is implemented as follows: in cases where the project name or design theme cannot be determined, multiple files are final drafts, a file with the same name already exists in the remote directory, the file type is not supported, the file is empty or lacks an extension, upload fails or the resource center returns an exception, the user explicitly requests to overwrite, delete or rename archived resources, or a retrospective search finds multiple similar results, the system can ask the user to confirm or supplement information to reduce the risk of misarchiving, misoverwriting and version confusion.

Images