Multi-agent based data space adaptive data delivery method and device

Abstract

The application relates to the technical field of data space, and discloses a multi-agent-based data space adaptive data delivery method and device, wherein the method comprises the following steps: inputting an unstructured data product API interface document and a data APP interface document into a large language model agent to obtain a structured data product API calling template and a data APP receiving template; inputting the data product API calling template and the data APP receiving template into the large language model agent to obtain a mapping conversion file; inputting a data use contract into the large language model agent to obtain a constraint rule file; inputting the mapping conversion file and the constraint rule file into the large language model agent to obtain a delivery process control graph; and acquiring data from the data product API by using the large language model agent according to the delivery process control graph, and injecting the obtained data into the data APP. The application can improve the automatic adaptation capability of a data delivery process in a data space and promote data circulation and utilization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data space, for example, relates to a multi-agent based data space adaptive data delivery method and device. BACKGROUND

[0002] At present, with the gradual construction of data element market, data as a new type of important production factor is increasingly prominent. In order to realize cross subject, cross organization, cross regional data credible sharing and controllable circulation, data space as a kind of architecture system with the core idea of "data sovereignty, compliance sharing" is widely deployed and applied in various fields.

[0003] In related technologies, data providers provide access capabilities to their data products through connectors. After data users download and deploy data APP, they sign data delivery agreements with data providers, then call the API of data provider connector to obtain data, and push it to the data APP deployed by the local connector to complete delivery.

[0004] In the process of implementing the embodiments of the present disclosure, it is found that at least the following problems exist in the related art:

[0005] Due to the high heterogeneity between the data product API of the data provider and the data APP input interface of the data user in terms of fields and semantics, the data delivery method in the related art relies on a large number of manual configuration and development of adaptation logic, cannot dynamically adapt to interface changes, and cannot realize automatic docking between the data product API and the data APP input interface in the data delivery process.

[0006] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present application, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] In order to have a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. The summary is not a general review, nor does it determine the key / important components or delineate the protection scope of these embodiments, but serves as a prelude to the detailed description below.

[0008] The embodiments of the present disclosure provide a multi-agent based data space adaptive data delivery method and device to improve the automatic adaptation capability between the data product API and the data APP input interface in the data delivery process of the data space, and promote data circulation and utilization.

[0009] In some embodiments, the multi-agent includes a plurality of large language model agents for implementing adaptive data delivery between data product APIs of data providers and data APPs of data users in a data space. The multi-agent-based data space adaptive data delivery method includes: inputting unstructured data product API interface documents and data APP interface documents into a large language model agent to obtain structured data product API call templates and data APP receiving templates; the data product API call templates and the data APP receiving templates are used for semantic mapping analysis. Inputting the data product API call templates and the data APP receiving templates into the large language model agent, and obtaining a mapping conversion file by analyzing the semantic mapping relationship; the mapping conversion file is used for mapping different semantics and structures in the data product API and the data APP. Inputting a data use contract signed between the data user and the data provider into the large language model agent to obtain a constraint rule file; the constraint rule file is used to control data access and use behavior. Inputting the mapping conversion file and the constraint rule file into the large language model agent to obtain a delivery process control graph by arranging the data delivery process in the data space. The large language model agent obtains data from the data product API according to the delivery process control graph, and injects the obtained data into the data APP.

[0010] Optionally, inputting the unstructured data product API interface documents and the data APP interface documents into the large language model agent to obtain the structured data product API call templates and the data APP receiving templates includes: using the large language model agent to respectively perform semantic analysis on the unstructured languages in the data product API interface documents and the data APP interface documents, and automatically identifying key information according to the semantic analysis results; the key information includes one or more of interface call methods, parameter requirements, field structures, and data formats; and generating the structured data product API call templates and the data APP receiving templates according to the key information.

[0011] Optionally, inputting the data product API call templates and the data APP receiving templates into the large language model agent to obtain the mapping conversion file by analyzing the semantic mapping relationship includes: using the large language model agent to analyze fields with different semantics and structures in the data product API call templates and the data APP receiving templates, constructing a mapping relationship between API output fields and APP input fields, and forming a field mapping table; automatically generating conversion logic rules according to data in the field mapping table; the mapping conversion logic includes one or more of type conversion rules, unit conversion rules, and default value filling rules; and generating the mapping conversion file according to the field mapping table and the conversion logic rules.

[0012] Optionally, the data use contract signed by the data user and the data provider is input into the large language model agent to obtain a constraint rule file, including: analyzing the use restriction clauses in the data use contract by using the large language model agent; the use restriction clauses include one or more of access frequency, time window, field range, use restriction, and data retention strategy; and generating the constraint rule file according to the use restriction clauses.

[0013] Optionally, the mapping conversion file and the constraint rule file are input into the large language model agent, and a delivery process control graph is obtained by arranging the data delivery process in the data space, including: analyzing the mapping conversion file and the constraint rule file by using the large language model agent, and arranging the data delivery process according to the analysis result to generate process control logic; and constructing the delivery process control graph according to the process control logic.

[0014] Optionally, the delivery process control graph includes policy nodes and conversion nodes; the large language model agent obtains data from the data product API according to the delivery process control graph, and injects the obtained data into the data APP, including: the large language model agent executes the delivery process control graph, calls the data product API according to the policy nodes in the delivery process control graph to obtain original data, and performs data field conversion on the original data according to the conversion nodes in the delivery process control graph to obtain target data expected to be received by the data APP; and the target data is injected into the data APP receiving interface.

[0015] Optionally, after generating the delivery process control graph, the delivery process control graph is confirmed, specifically including: displaying the delivery process control graph in a visual manner to the data user, and obtaining feedback information of the data user on the delivery process control graph; adjusting the process of generating the delivery process control graph according to the feedback information, and controlling the large language model agent to regenerate the delivery process control graph according to the adjusted process.

[0016] Optionally, the large language model agent is trained and optimized according to previous data delivery tasks, specifically including: collecting historical task information in the data delivery process; the historical task information includes one or more of task execution logs, field adaptation effects, user feedback, and exception information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; and the training and optimization of the large language model agent includes one or more of prompt word template enhancement, fine tuning, parameter adaptation, and transfer learning.

[0017] In some embodiments, the multi-agent-based data space adaptive data delivery device comprises: a document understanding agent for inputting an unstructured data product API interface document and a data APP interface document into a large language model agent to obtain a structured data product API calling template and a data APP receiving template; the data product API calling template and the data APP receiving template are used for semantic mapping analysis. A semantic mapping agent is configured to input the data product API calling template and the data APP receiving template into a large language model agent to obtain a mapping conversion file by analyzing semantic mapping relationships; the mapping conversion file is used for mapping different semantics and structures in the data product API and the data APP. A contract analysis agent is configured to input a data use contract signed by a data user and a data provider into a large language model agent to obtain a constraint rule file; the constraint rule file is used for controlling data access and use behaviors. A delivery process generation agent is configured to input the mapping conversion file and the constraint rule file into a large language model agent to obtain a delivery process control graph by arranging data delivery processes in a data space. An execution control agent is configured to use a large language model agent to obtain data from a data product API according to the delivery process control graph, and inject the obtained data into a data APP.

[0018] Optionally, the data delivery device further comprises: a multi-agent coordination and scheduling module configured to confirm the delivery process control graph after generating the delivery process control graph, specifically comprising: displaying the delivery process control graph to the data user in a visual manner, and obtaining feedback information of the data user on the delivery process control graph; adjusting the process of generating the delivery process control graph according to the feedback information, and controlling the large language model agent to regenerate the delivery process control graph according to the adjusted process; a memory and learning agent configured to train and optimize the large language model agent according to previous data delivery tasks, specifically comprising: collecting historical task information in the data delivery process; the historical task information comprises one or more of task execution logs, field adaptation effects, user feedback, and exception information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; the training and optimization of the large language model agent comprises one or more of prompt word template enhancement, fine-tuning, parameter adaptation, and transfer learning.

[0019] In some embodiments, the multi-agent-based data space adaptive data delivery device comprises a processor and a memory storing program instructions, wherein the processor is configured to execute the multi-agent-based data space adaptive data delivery method as described above when running the program instructions.

[0020] The multi-agent-based data space adaptive data delivery method and device provided by the embodiments of the present disclosure can achieve the following technical effects:

[0021] In the embodiments of the present disclosure, the API interface document and the data APP interface document are parsed by the large language model agent, the calling logic can be automatically extracted from the unstructured interface document, the interface structure is autonomously understood, the data product API calling template and the data APP receiving template are generated, then the semantic mapping relationship is analyzed by the large language model agent, the semantic similarity between the fields of the two templates is identified, and the corresponding conversion rule is generated, the mapping conversion file is obtained, the automatic semantic mapping and parameter conversion between heterogeneous APIs can be realized, the manual adaptation cost is reduced, the data usage contract is parsed by the large language model agent, the constraint rule file is generated, the data delivery process can strictly comply with the contract terms, finally, the delivery process control graph generated by the mapping conversion file and the rule constraint file is used to automatically adjust the delivery process to dynamically adapt to the interface changes, the delivery obstacles caused by the interface heterogeneity between the data provider interface and the data user interface are solved, the automatic adaptation capability between the data product API and the data APP input interface in the data delivery process of the data space is improved, and the data circulation and utilization are promoted.

[0022] The foregoing general description and the following description are only exemplary and explanatory, and are not used to limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0023] One or more embodiments are exemplarily illustrated by corresponding drawings, which do not constitute limitations on the embodiments, elements with the same reference numerals in the drawings are shown as similar elements, the drawings do not constitute proportional limits, and wherein:

[0024] Figure 1 is a schematic diagram of a data space architecture provided by the embodiments of the present disclosure;

[0025] Figure 2 is a schematic diagram of a data space multi-agent delivery engine provided by the embodiments of the present disclosure;

[0026] Figure 3 is a schematic diagram of another data space multi-agent delivery engine provided by the embodiments of the present disclosure;

[0027] Figure 4 is a flowchart of a data space multi-agent delivery engine when delivering data provided by the embodiments of the present disclosure;

[0028] Figure 5 is a schematic diagram of a multi-agent based data space adaptive data delivery method provided by the embodiments of the present disclosure;

[0029] Figure 6is a schematic diagram of a data space self-adaptive data delivery device based on multiple agents provided by the embodiments of the present disclosure. DETAILED DESCRIPTION

[0030] In order to enable a more detailed understanding of the features and technical content of the embodiments of the present disclosure, the implementation of the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings, which are only used for reference and do not limit the embodiments of the present disclosure. In the following technical description, in order to facilitate explanation, a plurality of details are provided to provide a full understanding of the disclosed embodiments. However, one or more embodiments can still be implemented without these details. In other cases, well-known structures and devices can be simplified to facilitate the drawings.

[0031] The terms "first", "second", and the like in the technical solutions described in the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the embodiments of the present disclosure described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion.

[0032] Unless otherwise specified, the term "a plurality of" means two or more.

[0033] In the embodiments of the present disclosure, the character " / " represents a "or" relationship between the objects before and after it. For example, A / B represents: A or B.

[0034] The term "and / or" is a description of the association between objects, which means that there can be three relationships. For example, A and / or B, which means: A or B, or, A and B, three kinds of relationship.

[0035] The term "corresponding" can refer to an association or binding relationship. A and B correspond to each other, which means that there is an association or binding relationship between A and B.

[0036] In combination Figure 1 As shown, the embodiments of the present disclosure provide a data space architecture, including a connector A of a data provider, a connector B of a data user, a data market and an application store.

[0037] The data provider can join the data space through the connector A and register one or more data products (data product 1, data product 2) owned by it in the data directory of the data market. Each data product needs to provide an interface specification document of its data access API in addition to metadata (such as data theme, data type, update time, provider identification, etc.) when publishing. The interface document can be in natural language formats such as OpenAPI, Markdown specification, Word / PDF, describing calling methods, parameter specifications, response formats, etc. Connector A uploads these information to the data market of the data space for other subjects to query and use.

[0038] Developers can develop adaptive data processing applications (APP1, APP2) for specific use cases (such as "product carbon footprint", "city traffic monitoring", "enterprise credit rating", "energy optimization analysis", etc.) through the open capabilities of the data space, and put them on the application store of the data space. Each APP provides a standardized receiving interface (such as REST API, gRPC, etc.) for receiving data and declares the expected data format and semantic structure. Data users browse the application store through connector B, select data APPs (such as "user portrait engine APP") that meet the needs of the scene, and download and deploy them to the local of connector B.

[0039] The data user browses the published data products in the data market, selects a data product (such as "user behavior data"), and initiates a contract signing process with the data provider. Both parties agree on the core elements of data use and form a data use contract, which includes: data access frequency (such as a maximum of 500 calls per day), validity period (such as 30 days from the signing date), field range limit, use limit (such as only for modeling analysis), region limit, response timeliness, access authentication mechanism, etc. The data use contract is registered and supervised by the data space contract service module as the basis for subsequent automatic delivery.

[0040] The connector B of the data user triggers the data space multi-agent delivery engine, i.e. the multi-agent based adaptive data delivery device of the data space, after the contract signing is successful. The multi-agent includes a plurality of large language model agents for realizing adaptive data delivery between the data product API of the data provider in the data space and the data APP of the data user. Figure 2 and Figure 4As shown, the data space multi-agent delivery engine 200 provided by the embodiments of the present disclosure includes a document understanding agent 201, a semantic mapping agent 202, a contract analysis agent 203, a delivery process generation agent 204, and an execution control agent 205. The document understanding agent 201 is configured to input an unstructured data product API interface document and a data APP interface document into a large language model agent, to obtain a structured data product API calling template and a data APP receiving template; the data product API calling template and the data APP receiving template are used for semantic mapping analysis. The semantic mapping agent 202 is configured to input the data product API calling template and the data APP receiving template into a large language model agent, to obtain a mapping conversion file by analyzing semantic mapping relationships; the mapping conversion file is used for mapping of different semantics and structures in the data product API and the data APP. The contract analysis agent 203 is configured to input a data use contract signed by a data user and a data provider into a large language model agent, to obtain a constraint rule file; the constraint rule file is used for controlling data access and use behaviors. The delivery process generation agent 204 is configured to input the mapping conversion file and the constraint rule file into a large language model agent, to obtain a delivery process control graph by arranging a data delivery process in a data space. The execution control agent 205 is configured to use a large language model agent to obtain data from a data product API according to the delivery process control graph, and to inject the obtained data into a data APP.

[0041] Optionally, in combination with Figure 3 and Figure 4 As shown, the data space multi-agent delivery engine 200 further includes a multi-agent coordination and scheduling module 206 and a memory and learning agent 207. The multi-agent coordination and scheduling module 206 is configured to, after generating the delivery process control graph, confirm the delivery process control graph, specifically including: displaying the delivery process control graph to the data user in a visual manner, and obtaining feedback information of the data user on the delivery process control graph; re-adjusting a process of generating the delivery process control graph according to the feedback information, and controlling the large language model agent to regenerate the delivery process control graph according to the adjusted process. The memory and learning agent 207 is configured to train and optimize the large language model agent according to historical data delivery tasks, specifically including: collecting historical task information in a data delivery process; the historical task information includes one or more of a task execution log, a field adaptation effect, user feedback, and exception information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; the training and optimization of the large language model agent includes one or more of prompt word template enhancement, fine-tuning, parameter adaptation, and transfer learning.

[0042] Optionally, to solve the problem that unstructured interface document understanding difficulty leads to the inability to automatically extract delivery semantics and conversion rules, the document understanding agent 201 can automatically extract the calling logic from the unstructured interface document, realize the autonomous understanding of the interface structure and the generation of the calling template. The document understanding agent 201 can identify the key information such as “mandatory parameters”, “return format”, “calling path” contained in the document, and automatically construct a standardized calling template (such as a RESTful style request), which significantly improves the adaptability of the connector to the API.

[0043] Optionally, to solve the problem that semantic inconsistency and structural differences between data provider APIs and data user APPs cause the delivery chain to break, the semantic mapping agent 202 can realize automatic semantic mapping and parameter conversion between heterogeneous APIs, significantly reducing the cost of manual adaptation. Through the semantic mapping agent 202, the semantic equivalence relationship between the data product API fields (such as “uid”, “stime”) and the data application receiving fields (such as “user_id”, “start_time”) is identified, and the mapping rules are automatically established in combination with historical memory, avoiding manual configuration one by one.

[0044] Optionally, to solve the problem that the data delivery process lacks intelligent process orchestration capability and cannot adapt to diverse API behaviors and APP requirements, the delivery process control graph generated by the delivery process generation agent 204 can realize intelligent scheduling and access control based on contract rules, ensuring that the data delivery process complies with policy constraints and adapts to diverse API behaviors and APP requirements. The contract analysis agent 203 can combine contract clauses such as “access times”, “time window”, “data granularity”, automatically reason the feasible scheduling plan, and cooperate with the delivery process generation agent 204 to generate the delivery process control graph, realize the contract-compliant calling, and avoid invalid requests.

[0045] Optionally, to solve the problem that there is a lack of multi-agent collaboration mechanism for data delivery scenarios, and it is impossible to form an autonomous adaptation system with learning ability, the memory and learning agent 207 can dynamically optimize the delivery strategy through the memory and learning mechanism, realize continuous adaptation and migration generalization in cross-product and cross-API scenarios. The system can continuously record the task execution history, including calling success rate, error information, and field mapping effectiveness, and reuse similar historical experience in new tasks to generate more optimal prompts or process paths, support “few-shot” adaptation and “zero-shot” migration, and improve the efficiency and robustness of long-term operation of the system.

[0046] The data space multi-agent delivery engine provided by the embodiments of the present disclosure can realize automatic adaptation between data provider API output and data user APP input, and intelligent arrangement, execution and feedback optimization of the data delivery process. The data space multi-agent delivery engine is composed of multiple agents, which work together through a multi-agent coordination and scheduling module to complete the closed-loop process from interface document analysis, semantic mapping, contract analysis, process generation, execution control to feedback learning. In addition, the delivery process control chart can clearly show the entire process from API request to APP input, including field mapping rules, interface protocol conversion, paging logic processing, call retry mechanism, data format conversion, authentication method insertion, data injection method, etc., and configure contract restrictions such as call frequency and time window. Under the guidance of the delivery process control chart, connector B actively schedules connector A to automatically initiate data requests according to the process. After connector A returns the data, connector B adjusts the data format to a format that the data APP can receive according to the field mapping and conversion rules, and injects the data into the corresponding data application, completing the data delivery process. The system records the delivery log throughout the process, and when an exception occurs, the memory and learning agent assists in diagnosing and automatically adjusting the process to ensure stable, reliable and compliant data delivery.

[0047] By using the above-mentioned data space multi-agent delivery engine, the understanding and reasoning capabilities of large language models and the collaborative autonomous mechanism of multi-agents are introduced to systematically solve the incompatibility problems between data provider API interface output and data user APP input in terms of semantics, structure and process in the data space. In combination with Figure 5 As shown in the drawings, the embodiments of the present disclosure provide a multi-agent-based data space adaptive data delivery method. The execution subject of the data delivery method can be a processor of a data space multi-agent delivery engine. The multi-agent includes a plurality of large language model agents for realizing adaptive data delivery between data product APIs of data providers and data APPs of data users in the data space. The data delivery method includes the following steps.

[0048] S501, the processor inputs the unstructured data product API interface document and the data APP interface document into the large language model agent to obtain a structured data product API calling template and a data APP receiving template.

[0049] The data product API calling template and the data APP receiving template are used for semantic mapping analysis.

[0050] S502, the processor inputs the data product API calling template and the data APP receiving template into the large language model agent to obtain a mapping conversion file by analyzing the semantic mapping relationship.

[0051] The mapping conversion file is used for mapping of different semantics and structures in the data product API and the data APP.

[0052] In S503, the processor inputs a data use contract signed by the data user and the data provider into the large language model agent to obtain a constraint rule file.

[0053] The constraint rule file is used for controlling data access and use behaviors.

[0054] In S504, the processor inputs the mapping conversion file and the constraint rule file into the large language model agent to obtain a delivery process control graph by arranging a data delivery process in the data space.

[0055] In S505, the processor obtains data from the data product API according to the delivery process control graph by using the large language model agent, and injects the obtained data into the data APP.

[0056] In the embodiments of the present disclosure, the API interface document and the data APP interface document are parsed by the large language model agent, the calling logic can be automatically extracted from the unstructured interface document, the interface structure can be autonomously understood, the data product API calling template and the data APP receiving template can be generated, the semantic mapping relationship can be analyzed by the large language model agent, the semantic similarity between the fields of the two templates can be identified, the corresponding conversion rule can be generated, the mapping conversion file can be obtained, the automatic semantic mapping and parameter conversion between heterogeneous APIs can be realized, the manual adaptation cost can be reduced, the data use contract can be parsed by the large language model agent, the constraint rule file can be generated, the data delivery process can be strictly in accordance with the contract terms, and finally the delivery process control graph generated according to the mapping conversion file and the rule constraint file by the large language model agent can automatically adjust the delivery process to dynamically adapt to the interface changes, so that the delivery obstacles caused by the interface heterogeneity between the data provider interface and the data user interface are solved, the automatic adaptation capability between the data product API and the data APP input interface in the data delivery process of the data space is improved, and the data circulation and utilization are promoted.

[0057] Optionally, the unstructured data product API interface document and the data APP interface document are input into the large language model agent to obtain the structured data product API calling template and the data APP receiving template, including: the large language model agent is used to respectively perform semantic analysis on the unstructured languages in the data product API interface document and the data APP interface document, and automatically identify key information according to the semantic analysis results; the key information includes one or more of an interface calling mode, parameter requirements, field structure and data format; and the structured data product API calling template and the data APP receiving template are respectively generated according to the key information.

[0058] In this embodiment, the natural language understanding capability of the data space multi-agent delivery engine document understanding agent can parse the data product API output interface document and the data APP input interface document, identify the interface calling method, parameter requirement, field structure and data format, extract key information and generate a data product API calling template and a data APP receiving template. The data product API calling template and the data APP receiving template are unified structured templates, such as a JSON format (method, url, params, response), including an interface path, a parameter format, a response structure and the like, which are used for subsequent semantic mapping analysis and can process OpenAPI, Markdown, Postman JSON or natural language description and the like. The document understanding agent can quickly and accurately extract key information from unstructured documents through natural language processing technology, generate standardized calling templates and receiving templates, and significantly improve the development efficiency and system adaptability.

[0059] Optionally, the document understanding agent has one or more of the following functions: semantic analysis and key parameter extraction on interface description documents; automatic identification of the request method, URL path, parameter field and type of the interface; automatic completion of missing interface field descriptions (such as completion of URL, parameter description and the like according to the context); and output of a structured interface template (JSON format) for subsequent mapping and calling.

[0060] Optionally, the document understanding agent supports a Few-shot Prompt mechanism, and generates a standardized structure through a preset “interface extraction instruction template” prompt model. The Prompt can be: “Please extract the interface path, method, request parameter name and type, return field name and type from the following API document, and construct a JSON structure.”

[0061] For example, for the "City Travel Trajectory API", the interface document describes the following: the request address is GET / trajectory / list; the parameter description is uid (string), start_time (datetime), end_time (datetime); the response field is lng, lat, t. The document understanding agent identifies the calling method, parameter format, and field semantics, and outputs the following structured data product API calling template: {"method": "GET", "url": " / trajectory / list", "params": ["uid", "start_time", "end_time"], "response": ["lng", "lat", "t"]}. The template describes the calling method (method), URL path (url), request parameter (params), and response field (response) of the API in detail. Among them, "GET" indicates that this is a request to get data from the server; " / trajectory / list" specifies the path of the API, indicating that the API is used to get trajectory data; "uid" indicates the user ID, which is used to identify the user; "start_time" indicates the start time, and "end_time" indicates the end time, which is used to specify the time range of the query; "lng" indicates the longitude coordinate of the trajectory point, "lat" indicates the latitude coordinate of the trajectory point, and "t" indicates the time of the trajectory point. According to the data product API calling template, an API interface for getting user trajectory data can be determined, and the caller can get the trajectory record of a specified user in a specific time range through this interface. When calling, the user ID, start time, and end time need to be provided as parameters.

[0062] Similarly, for the input interface document of a data APP, the document understanding agent can also generate the following data APP receiving template: {"endpoint": " / app / input", "method": "POST", "body": {"user_id": "string", "longitude": "float", "latitude": "float", "timestamp": "datetime"}}. The template describes the input interface of the data APP in detail, including the endpoint for receiving data, the request method, and the fields in the request body and their data types. Among them, " / app / input" specifies the input interface path of the data APP, indicating that data will be sent to this path; "POST" indicates that this is a request to send data to the data APP; "user_id" indicates the user ID, which is used to identify the user, and its data type is string; "longitude" indicates the longitude coordinate of the track point, and its data type is float; "latitude" indicates the latitude coordinate of the track point, and its data type is float; "timestamp" indicates the time of the track point, and its data type is datetime. According to the data APP receiving template, a data APP interface for receiving user trajectory data can be determined, and the caller can send user trajectory data to the data APP through this interface. When calling, the request body needs to be constructed according to the specified fields and data types.

[0063] Optionally, the document understanding and call template generation capabilities introduced by the document understanding agent can build the following automatic parsing link: natural language document, Prompt template, parameter extraction, type recognition, JSON structure, and call semantic template.

[0064] Optionally, the data product API call template and the data APP receiving template are generated according to the following formula:

[0065]

[0066]

[0067] wherein D api and D app are the data product API interface document and the data APP interface document, F LLM is the structure extraction function of the document understanding agent, T api and T app are the data product API call template and the data APP receiving template, and the output is in JSON format.

[0068] Optionally, the data product API call template and the data APP receiving template are input into the large language model agent, and a mapping conversion file is obtained by analyzing the semantic mapping relationship, including: using the large language model agent to analyze the fields with different semantics and structures in the data product API call template and the data APP receiving template, constructing the mapping relationship between the API output fields and the APP input fields, and forming a field mapping table; according to the data in the field mapping table, analyzing the mapping conversion logic, and automatically generating conversion logic rules; the mapping conversion logic includes one or more of type conversion rules, unit conversion rules, and default value filling; according to the field mapping table and the conversion logic rules, a mapping conversion file is generated.

[0069] In this embodiment, through the semantic mapping agent in the data space multi-agent delivery engine, the fields with different semantics and structures in the data product API call template and the data APP receiving template can be analyzed, the semantic mapping relationship between the data product API output and the data APP input can be established, the mapping relationship supports field-level, structure-level, and semantic-level mapping, and a field mapping table is formed. According to the field semantics, types, contexts, and historical experience in the field mapping table, field conversion logic, unit conversion rules, and format adjustment rules can also be automatically generated, conversion logic rules are formed, and finally a mapping conversion file including the field mapping table and the conversion logic rules is output. The mapping conversion file is used for mapping of different semantics and structures in the data product API and the data APP.

[0070] Optionally, the semantic mapping agent has one or more of the following functions: using a large language model to perform semantic matching of cross-interface field names, such as user_id and uid; automatically generating conversion rules, including type conversion, unit conversion, default value filling, etc.; when interface fields are missing or semantic is ambiguous, historical memory is used to complete the completion (such as default value, derived value).

[0071] Optionally, the specific processing details of the semantic mapping agent include: constructing a semantic matching task between field pairs (source_field, target_field), inputting field names, context descriptions, and field types, constructing field semantic embedding vectors based on large language models; performing semantic similarity calculation and structural context matching between field pairs, such as using LLM for embedding comparison or semantic classification matching; using chain prompts to improve the reasoning ability of field disassembly (such as disassembling location into lng+lat) or field fusion (such as merging first_name and last_name into fullname); supporting fuzzy matching (such as matching "uid" with "user_id") and structural splitting (such as splitting "geo_location" into "longitude" and "latitude"); unit conversion support (such as converting the unit of "speed" from km / h to m / s).

[0072] Optionally, the semantic mapping agent supports the Few-shot Prompt mechanism. Wherein, the Prompt can be: "Match the fields in the data source with the APP fields and output the corresponding conversion logic. Please consider the meaning, type and purpose of the fields, and explain the reasoning basis."

[0073] For example, the semantic mapping agent can identify semantic similarity between fields and generate a corresponding rule as follows: {"mappings": {"uid": "user_id", "lng": "longitude", "lat": "latitude", "t": "timestamp"}, "conversions": {"t": {"type": "timestamp", "format": "ISO-8601"}}}. Among them, "mappings" describes the mapping relationship between the output fields of the data provider API and the input fields of the data user APP; "uid": "user_id" indicates that the uid field of the data provider is mapped to the user_id field of the data user; "lng": "longitude" indicates that the lng field of the data provider is mapped to the longitude field of the data user; "lat": "latitude" indicates that the lat field of the data provider is mapped to the latitude field of the data user; "t": "timestamp" indicates that the t field of the data provider is mapped to the timestamp field of the data user; "conversions" describes the data conversion logic that needs to be performed, especially the conversion of field types and formats; "timestamp" specifies that the field t is a timestamp field, and "ISO-8601" specifies that the format of the field t is the date and time string of the ISO 8601 standard.

[0074] In actual data delivery, the field naming, structure organization, and parameter format of the data product API of the data provider often differ significantly from the receiving interface of the data user APP. For example, the data product API fields can be uid, uname, and access_time, while the data APP needs fields of userId, username, and timestamp. Through the semantic mapping agent, semantic-level automatic mapping generation, updating, and repair can be achieved, greatly reducing configuration costs and improving heterogeneous interface adaptation capabilities.

[0075] Optionally, the semantic mapping agent generates mapping rules based on an inter-interface semantic mapping function. The inter-interface semantic mapping function is:

[0076]

[0077] Among them, Match indicates that two fields form a mapping, and f api，i is the output field of the data product API of the data provider, and f app，jis the input field of the data consumer data APP, E is the field semantic embedding generated by the semantic mapping agent, cosine is the cosine similarity between two embedding vectors, measuring the semantic proximity of two fields in the semantic space, ranging from [-1, 1], S i 、S j is the context path of the field in the original structure (such as JSON path, parent node), Sim is the context position similarity of the field in the original data structure, which can be calculated based on path coincidence, field co-occurrence frequency, parent node semantics, and α, β are the weights of adjusting semantic and structural similarity.

[0078] For example, for "the field uname in the API interface document represents the user's name", through the prompt and context, the large model can infer that it should be mapped to the field username of the APP, and accordingly generate the following mapping rule: {"uname":"username"}. It means that the data provider field "uname" is mapped to the data consumer field "username".

[0079] Optionally, the data use contract signed between the data consumer and the data provider is input into the large language model agent to obtain a constraint rule file, including: analyzing the use restriction clauses in the data use contract using the large language model agent; the use restriction clauses include one or more of access frequency, time window, field range, use restriction, and data retention policy; generating a constraint rule file according to the use restriction clauses.

[0080] In this embodiment, through the contract analysis agent in the data space multi-agent delivery engine, the access frequency, time window, field range, use restriction, data retention policy and other use restriction clauses in the data use contract can be identified, and an access plan and quota strategy can be generated. According to the content of the data contract, the interface call frequency, time limit, permission, etc. are scheduled and limited, and a constraint rule file is generated for subsequent delivery scheduling, that is, a structured call limit rule and scheduling strategy set, which is used to control data access and use behavior.

[0081] Optionally, the contract analysis agent has one or more of the following functions: parsing the clauses in the contract document to extract access times, call time limit, use permission, data item limit, etc.; determining the delivery frequency and life cycle in cooperation with the delivery process generation agent; providing strategy checking and limit checking in the call execution stage.

[0082] Optionally, the contract analysis agent supports the Few-shot Prompt mechanism. Among them, the Prompt can be: "Extract the call frequency limit, start and end time, data available field and prohibited use from the following contract, and convert it to structured JSON."

[0083] For example, for the contract "maximum 300 calls per day; limited to weekdays 08:00-18:00; valid from January 1, 2025 to January 31, 2025.", the contract analysis intelligent agent can generate the following constraint rule file: {"max_calls_per_day": 300, "allowed_time_range": ["08:00", "18:00"], "validity": {"start": "2025-01-01", "end": "2025-01-31"}}. Among them, "max_calls_per_day" represents the maximum number of calls allowed per day; "allowed_time_range" represents the time range allowed for calling; "validity" represents the validity period of the contract, "start" represents the start date of the contract, and "end" represents the end date of the contract.

[0084] Optionally, the mapping conversion file and the constraint rule file are input into the large language model intelligent agent, and a delivery process control graph is obtained by arranging the data delivery process in the data space, including: analyzing the mapping conversion file and the constraint rule file using the large language model intelligent agent, and arranging the data delivery process according to the analysis result to generate process control logic; and constructing the delivery process control graph according to the process control logic.

[0085] In this embodiment, the delivery process generation intelligent agent in the data space multi-intelligent agent delivery engine can parse the mapping conversion file and the contract rule file, combine the API call mode and the APP structure requirement, consider the request order, call dependency, paging logic, authentication processing, error retry, and other process control logic, plan and generate an end-to-end data delivery process, and use the delivery process control graph to represent it.

[0086] Optionally, the delivery process control graph includes policy nodes and conversion nodes.

[0087] In this embodiment, the delivery process control graph is a directed graph structure, containing request order, branching logic and control nodes, and exhibiting data flow, conversion nodes, policy nodes, terminal nodes, etc.

[0088] Optionally, the delivery process generation intelligent agent has one or more of the following functions: considering contract strategies, parameter dependencies, and paging mechanisms to construct process control logic; processing complex scenarios such as multi-API combination, incremental update, multi-round request, error retry, etc.; outputting a standardized delivery plan format for use by the execution control intelligent agent.

[0089] Optionally, the delivery process generation agent supports the Few-shot Prompt mechanism. The prompt can be: "Please convert the file and constraint rule based on the following mapping to design a complete process including parameter preparation, API paging request, data conversion, structure reconstruction, and data injection."

[0090] For example, the logical nodes of the generated delivery process control graph are: [initialize call parameters], [get paging data / trajectory / list], [field conversion and structure reconstruction], [POST to / app / input], [log and contract quota update]. The delivery process control graph supports conditional judgment, paging traversal, asynchronous request, retry mechanism, and other control logic, and can be displayed to data users through a graphical interface.

[0091] Optionally, the delivery process control graph DCG can be represented as follows:

[0092]

[0093] Where DCG is the delivery process control graph, N is the set of process nodes such as data request, field mapping, paging processing, data injection, E is the set of execution order edges between nodes, op k is the operation type, param k is the parameter input of the node operation, cond k is the conditional judgment (such as paging continuation condition, contract limit condition).

[0094] In the delivery process control graph DCG, each node represents an API call, judgment, conversion, and other operations. For example: "call / submitJob" represents obtaining jobId; "poll / getResult?jobId=" represents checking the result status; "get data[] field" represents converting the structure and writing to the APP interface.

[0095] Optionally, the delivery process generation agent generates the delivery process control graph based on the mapping conversion file (including the field mapping table and the conversion logic rule) and the constraint rule file according to the following formula:

[0096]

[0097] Where FieldMap is the field mapping table, TransformRules is the conversion logic rule, and ContractConstraintSet is the constraint rule file.

[0098] Optionally, the generation of the delivery process control graph follows the following automatic planning mechanism: parsing key fields in API documents (such as jobId in the response); identifying dependencies (such as the result query depending on the return value of the submitted task); constructing a process control graph with conditional branches and loops.

[0099] Optionally, during the invocation of the delivery process control graph, if a process error occurs (such as missing pagination fields, response structure changes), the execution control agent can repair the process by retrieving historical delivery success examples and performing Few-shot migration, achieving autonomous recovery and evolutionary optimization of the process.

[0100] Optionally, the large language model agent obtains data from the data product API according to the delivery process control graph, and injects the obtained data into the data APP, including: using the large language model agent to execute the delivery process control graph, calling the data product API according to the strategy node in the delivery process control graph, obtaining the original data, and performing data field conversion on the original data according to the conversion node in the delivery process control graph to obtain the target data expected to be received by the data APP; injecting the target data into the data APP receiving interface.

[0101] In this embodiment, through the execution control agent in the data space multi-agent delivery engine, the specific API call, data conversion and result transmission can be performed according to the delivery process control graph, and the task tracking and state feedback capability is also provided, and the delivery result (such as delivery state, delivery data, delivery log) is recorded.

[0102] Optionally, the execution control agent has one or more of the following functions: managing the state and life cycle (start, pause, retry, terminate) of data delivery tasks; supporting authentication mechanisms such as Token injection, signature verification, Session maintenance, etc.; supporting failure recovery mechanisms (such as retry, rollback), and updating experience with the memory and learning agent.

[0103] Optionally, after generating the delivery process control graph, the delivery process control graph is confirmed, specifically including: displaying the delivery process control graph to the data user in a visual manner, and obtaining feedback information of the data user on the delivery process control graph; according to the feedback information, readjust the process of generating the delivery process control graph, and control the large language model agent to regenerate the delivery process control graph according to the adjusted process.

[0104] In this embodiment, after generating the delivery process control graph, the delivery process control graph will be displayed in a visual manner to the data user through the multi-agent coordination scheduling module in the data space multi-agent delivery engine, for "manual review" and "strategy review" of the delivery process. If the data user finds abnormalities (such as field mapping errors, data granularity mismatches, etc.), the system will feed back the problems to the corresponding agent through the multi-agent coordination scheduling module, and trigger local intelligent update to adjust the process of generating the delivery process control graph, and then output the delivery process control graph again for confirmation. This process can optimize the process structure through multiple rounds of question and answer, such as "Please complete the pagination in the current process without considering nextToken".

[0105] Optionally, the multi-agent coordination scheduling module can coordinate the communication and task allocation between agents, output agent scheduling command stream, global task execution plan and task log audit report, and ensure process continuity and state consistency.

[0106] Optionally, the multi-agent coordination scheduling module has one or more of the following functions: controlling the execution flow order (such as first parsing the document, then matching the mapping, then planning the process, and then executing); managing cross-agent communication protocols (based on message queues / event streams); tracking task status, recording logs, and generating auditable reports; supporting human-computer collaborative interface, accepting administrator or developer intervention prompts.

[0107] Optionally, according to the historical data delivery tasks, the large language model agent is trained and optimized, specifically including: collecting historical task information in the data delivery process; the historical task information includes one or more of task execution logs, field adaptation effects, user feedback, and abnormal information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; the training and optimization of the large language model agent includes one or more of prompt word template enhancement, fine-tuning, parameter adaptation, and transfer learning.

[0108] In this embodiment, through the memory and learning agent in the data space multi-agent delivery engine, historical task information including task execution logs, field adaptation effects, user feedback, and abnormal information can be collected, so as to construct a data delivery training set, and train and optimize the large language model agent according to the data delivery training set, for optimizing future tasks. In addition, the memory and learning agent is also used to store, manage and replay historical task experience, that is, interface, parameters, mapping, process and execution log information in the process of executing the tasks, to prompt optimization suggestions, similar task matching results or Prompt enhancement content, to provide prompt enhancement and rapid migration support for other agents.

[0109] Optionally, the memory and learning agent has one or more of the following functions: storing parameter mapping, flowchart, calling state and exception information of completed tasks; forming a task knowledge graph to support similar task matching and Prompt optimization; driving each agent to execute prompt reinforcement or fine-tuning.

[0110] Optionally, the specific processing details of the memory and learning agent include: constructing a field mapping knowledge graph; providing Prompt optimization suggestions for new tasks, using LLM to generate prompt templates to enhance the adaptation of other agents in similar scenarios; based on the knowledge graph, optimizing the large language model instruction template for guiding Prompt selection and instruction combination generation; supporting fine-tuning or parameter adaptation methods for domain adaptation of high-frequency interfaces or specific industry term scenarios; cross-task migration mapping experience (such as user_id, uid, userid merged into synonymous field families).

[0111] For example, in multiple tasks, the matching results of lng and longitude appear, and the memory and learning agent can automatically promote this mapping to a high-confidence rule and preferentially recommend it.

[0112] Optionally, the memory and learning agent records the interface adaptation history, field mapping record, and calling path generation, etc. by continuously recording the contents, and constructs a data delivery knowledge graph according to the following formula:

[0113]

[0114] Where K is the data delivery knowledge graph, API i is the output interface document of the data provider data product API, APP j is the receiving interface document of the data user data APP, FieldMap ij is the field mapping scheme generated by the semantic mapping agent, DCG ij is the delivery process automatically arranged by the delivery flow control chart, Feedback ij is the structured interactive feedback such as actual calling log, error feedback, and user score.

[0115] In this embodiment, the data delivery knowledge graph is managed by the memory and learning agent, and has functions such as memory enhancement, migration generalization, and active error correction. Among them, memory enhancement refers to combining the mapping and process of similar tasks in K as a Few-shot example when adapting to a new interface, to improve understanding and generation accuracy; migration generalization refers to continuously updating parameters through an online learning mechanism, so that it is more efficient when adapting to similar interfaces in the future, supporting migration from adapting to the first platform behavior data API to the second platform behavior data API even if the field name is different; active error correction refers to after multiple delivery failures, the agent can automatically compare the process diagrams and mapping logic of success and failure, infer the error cause, and realize error self-correction.

[0116] For example, a user APP has successfully adapted the "city traffic congestion index API", when it accesses the "highway congestion prediction API", the system identifies that the field semantics of the two are similar (such as city_name vs region_name, congestion_rate vs traffic_score), can directly migrate the existing Map and Flow, and only needs to make a small adjustment in the first call to successfully deliver, significantly reducing the cold start cost. In addition, by using the fine-tuning strategy, the domain data interface document, delivery log and other data of each agent are continuously fine-tuned to optimize their understanding and reasoning ability in the data space delivery task, so that the system has the ability of long-term evolution and continuous enhancement.

[0117] The multi-agent based data space self-adaptive data delivery method and device provided by the embodiments of the present disclosure, each agent in the data space multi-agent delivery engine can complete model reasoning tasks through a unified LLM calling interface (which can be a local model service or a cloud API), and each agent maintains an independent Prompt strategy and prompt template. Cross-validation and model feedback mechanisms are supported during execution, and multiple candidate results are voted to determine the mapping or process path. Each agent module cooperates and cooperates, building a semantic closed loop from API to APP, adaptive delivery process, strategy constraint analysis and task memory feedback mechanism between connectors, realizing efficient, stable, low-code, and migratable data delivery for the data space. The data space multi-agent delivery engine improves the technical level in automatic adaptation of heterogeneous interfaces between data product APIs and data APPs, complex process arrangement, document-driven analysis, knowledge migration generalization, etc. by fusing the semantic understanding ability, natural language processing ability, context reasoning ability of large language models, and task division and cooperation, autonomous learning and evolution mechanism of multi-agent.

[0118] In combination with Figure 6As shown, the embodiment of the present disclosure provides a multi-agent based data space adaptive data delivery device 600, which comprises a processor 601 and a memory 602. Optionally, the device can further comprise a communication interface 603 and a bus 604. Wherein the processor 601, the communication interface 603 and the memory 602 can complete mutual communication through the bus 604. The communication interface 603 can be used for information transmission. The processor 601 can invoke the logic instructions in the memory 602 to execute the multi-agent based data space adaptive data delivery method of the above-mentioned embodiment.

[0119] In addition, the logic instructions in the memory 602 described above can be realized in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium.

[0120] The memory 602 as a computer readable storage medium can be used to store software programs, computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present disclosure. The processor 601 executes the program instructions / modules stored in the memory 602, thereby performing function application and data processing, that is, realizing the multi-agent based data space adaptive data delivery method in the above-mentioned embodiment.

[0121] The memory 602 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; the data storage area can store data created according to the use of the terminal device, etc. In addition, the memory 602 can include a high-speed random access memory, and can also include a non-volatile memory.

[0122] The embodiment of the present disclosure provides an electronic device, which comprises an electronic device body and the multi-agent based data space adaptive data delivery device described above. The multi-agent based data space adaptive data delivery device is installed in the electronic device body. The installation relationship described herein is not limited to placing in the electronic device, but also includes installation connection with other components of the electronic device, including but not limited to physical connection, electrical connection or signal transmission connection, etc. Those skilled in the art can understand that the multi-agent based data space adaptive data delivery device can be adapted to a feasible electronic device body, and thus realize other feasible embodiments.

[0123] The embodiment of the present disclosure provides a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are set to execute the multi-agent based data space adaptive data delivery method described above.

[0124] The technical solutions of the embodiments of the present disclosure can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes one or more instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method disclosed in the embodiments of the present disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0125] The above description and drawings sufficiently illustrate the embodiments of the present disclosure to enable one skilled in the art to practice them. Other embodiments can include structural, logical, electrical, process, and other changes. The embodiments represent only a few of the possible variations. Individual components and functions are optional unless explicitly required, and the order of operations can be changed. Parts and features of some embodiments can be included in or replace parts and features of other embodiments. Also, the words used in this application are only used to describe the embodiments and not to limit the technical solutions described in this application. As used in the technical solutions described in this application, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms as well. Similarly, as used in this application, the term "and / or" refers to any and all possible combinations of one or more associated listed items. In addition, when used in this application, the term "comprise" and its variants "comprises" and / or "comprising" and the like refer to the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups of these. Without more limitations, the element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, or device that includes the stated element. In this document, each embodiment focuses on the differences from other embodiments, and the same or similar parts between various embodiments can be referred to each other. For the method, product, etc. disclosed in the embodiments, if it corresponds to the method part disclosed in the embodiments, the relevant part can be referred to the description of the method part.

[0126] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods for each specific application to realize the described functions, but such implementation should not be considered beyond the scope of the embodiments of the present disclosure. The skilled person can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0127] In the embodiments disclosed herein, the disclosed methods, products (including but not limited to devices, equipment, etc.) can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units can only be a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms. The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to implement the embodiments. In addition, each functional unit in the embodiments of the present disclosure can be integrated in one processing unit, or each unit can be a physically independent unit, or two or more units can be integrated in one unit.

[0128] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

Claims

1. A multi-agent based data space adaptive data delivery method, characterized in that, The multi-agent includes a plurality of large language model agents, and is used for realizing adaptive data delivery between a data product API of a data provider and a data APP of a data user in a data space. Unstructured data product API interface documents and data APP interface documents are input into the large language model agent to obtain structured data product API calling templates and data APP receiving templates; the data product API calling templates and the data APP receiving templates are used for semantic mapping analysis; The data product API calling templates and the data APP receiving templates are input into the large language model agent, and a mapping conversion file is obtained by analyzing a semantic mapping relationship; the mapping conversion file is used for mapping of different semantics and structures in the data product API and the data APP; A data use contract signed by the data user and the data provider is input into the large language model agent to obtain a constraint rule file; the constraint rule file is used for controlling data access and use behaviors; The mapping conversion file and the constraint rule file are input into the large language model agent, and a delivery process control graph is obtained by arranging a data delivery process in the data space; The large language model agent is used to obtain data from the data product API according to the delivery process control graph, and the obtained data is injected into the data APP.

2. The data delivery method of claim 1, wherein, Unstructured data product API interface documents and data APP interface documents are input into the large language model agent to obtain structured data product API calling templates and data APP receiving templates, including: The large language model agent is used to respectively perform semantic analysis on unstructured languages in the data product API interface documents and the data APP interface documents, and automatically identify key information according to a semantic analysis result; the key information includes one or more of an interface calling mode, parameter requirements, field structure and data format; According to the key information, the structured data product API calling templates and the data APP receiving templates are respectively generated.

3. The data delivery method of claim 1, wherein, The data product API calling templates and the data APP receiving templates are input into the large language model agent, and a mapping conversion file is obtained by analyzing a semantic mapping relationship, including: The large language model agent is used to analyze fields with different semantics and structures in the data product API calling templates and the data APP receiving templates, to construct a mapping relationship between API output fields and APP input fields, and to form a field mapping table; According to mapping conversion logic in the field mapping table, conversion logic rules are automatically generated; the mapping conversion logic includes one or more of type conversion rules, unit conversion rules and default value filling rules; According to the field mapping table and the conversion logic rules, a mapping conversion file is generated.

4. The data delivery method of claim 1, wherein, The data use contract signed by the data user and the data provider is input into the large language model agent to obtain a constraint rule file, including: analyzing, by the large language model agent, a usage restriction clause in the data usage contract; the usage restriction clause comprises one or more of access frequency, time window, field range, usage restriction, and data retention policy; generating the constraint rule file according to the usage restriction clause.

5. The data delivery method of claim 1, wherein, inputting the mapping conversion file and the constraint rule file into the large language model agent, and obtaining a delivery process control graph by scheduling a data delivery process in a data space, comprising: analyzing, by the large language model agent, the mapping conversion file and the constraint rule file, and scheduling a data delivery process according to the analysis result to generate process control logic; constructing the delivery process control graph according to the process control logic.

6. The data delivery method of claim 1, wherein, The delivery process control graph comprises policy nodes and conversion nodes; the large language model agent obtains data from a data product API according to the delivery process control graph, and injects the obtained data into a data APP, comprising: The large language model agent executes the delivery process control graph, calls a data product API according to a policy node in the delivery process control graph to obtain raw data, and performs data field conversion on the raw data according to a conversion node in the delivery process control graph to obtain target data expected to be received by a data APP; injecting the target data into a data APP receiving interface.

7. The data delivery method according to any one of claims 1 to 6, characterized by, After generating the delivery process control graph, the delivery process control graph is confirmed, specifically comprising: visually displaying the delivery process control graph to a data user, and obtaining feedback information of the data user on the delivery process control graph; re-adjusting the process of generating the delivery process control graph according to the feedback information, and controlling the large language model agent to regenerate the delivery process control graph according to the adjusted process.

8. The data delivery method according to any one of claims 1 to 6, characterized by, According to the historical data delivery tasks, the large language model agent is trained and optimized, specifically comprising: collecting historical task information in the data delivery process; the historical task information comprises one or more of task execution log, field adaptation effect, user feedback, and exception information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; the training and optimization of the large language model agent comprises one or more of prompt word template enhancement, fine tuning, parameter adaptation, and transfer learning.

9. A multi-agent based data space adaptive data delivery apparatus, characterized by, The multi-agent comprises a plurality of large language model agents, which are used to realize adaptive data delivery between a data product API of a data provider in a data space and a data APP of a data user, and the data delivery device comprises: a document understanding agent, configured to input unstructured data product API interface documents and data APP interface documents into the large language model agent to obtain structured data product API calling templates and data APP receiving templates; the data product API calling templates and the data APP receiving templates are used for semantic mapping analysis; The semantic mapping agent is configured to input the data product API calling template and the data APP receiving template into the large language model agent, obtain a mapping conversion file by analyzing semantic mapping relationships, and map different semantics and structures in the data product API and the data APP by using the mapping conversion file. The contract analysis agent is configured to input a data use contract signed by a data user and a data provider into the large language model agent, obtain a constraint rule file, and use the constraint rule file to control data access and use behaviors. The delivery process generation agent is configured to input the mapping conversion file and the constraint rule file into the large language model agent, obtain a delivery process control graph by arranging a data delivery process in a data space, and use the delivery process control graph to control the data delivery process. The execution control agent is configured to use the large language model agent to obtain data from a data product API according to the delivery process control graph, and inject the obtained data into a data APP.

10. The data delivery device of claim 9, wherein, Further comprising: A multi-agent coordination and scheduling module configured to confirm the delivery process control graph after the delivery process control graph is generated, specifically including: displaying the delivery process control graph in a visual manner to a data user, and obtaining feedback information of the data user on the delivery process control graph; adjusting the process of generating the delivery process control graph again according to the feedback information, and controlling the large language model agent to generate the delivery process control graph again according to the adjusted process. A memory and learning agent configured to train and optimize the large language model agent according to previous data delivery tasks, specifically including: collecting historical task information in a data delivery process; the historical task information includes one or more of task execution logs, field adaptation effects, user feedback, and abnormal information; constructing a data delivery training set according to the historical task information; training and optimizing the large language model agent according to the data delivery training set; the training and optimization of the large language model agent includes one or more of prompt word template enhancement, fine-tuning, parameter adaptation, and transfer learning.

11. A multi-agent based data space adaptive data delivery apparatus comprising a processor and a memory having stored thereon program instructions, the apparatus being characterized by: The processor is configured to execute the program instructions to perform the multi-agent-based data space adaptive data delivery method according to any one of claims 1 to 8 when the program instructions are run.

Images