Configuration information generation method, program operation method and device based on large model

Abstract

The present disclosure provides a configuration information generation method based on a large model, a program execution method, and an apparatus, relating to the field of artificial intelligence technology, particularly to large language models, computer technology, and other technical fields. A specific implementation scheme of the configuration information generation method is as follows: using the large model, generating first configuration information in a target format based on an initial text, wherein the initial text is used to describe the configuration intent for the target program; using the large model, generating a target text based on the first configuration information; wherein the target text is used to describe the configuration content of the first configuration information; and in response to determining that the configuration content matches the configuration intent, determining that the first configuration information is the target configuration information for running the target program.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to technical fields such as large language models and computer technology, and involves a configuration information generation method based on a large model, a program running method and device, an electronic device, a storage medium, and a program product. Background Art

[0002] Large language models have rich natural language understanding capabilities and are widely used in natural language processing and text processing, such as chat interaction, machine translation, sentiment analysis, intelligent customer service, medical diagnosis, speech recognition, computer vision, etc.

[0003] As the application areas of large language models continue to expand, they play an important role in various industries, such as improving work efficiency, improving user experience, and promoting technological innovation. Summary of the Invention

[0004] The present disclosure provides a configuration information generation method based on a large model, a program running method and device, an electronic device, a storage medium, and a program product.

[0005] According to one aspect of the present disclosure, a method for generating configuration information based on a large model is provided, comprising: utilizing the large model to generate first configuration information in a target format based on an initial text, wherein the initial text is used to describe the configuration intent for a target program; utilizing the large model to generate a target text based on the first configuration information; wherein the target text is used to describe the configuration content of the first configuration information; and in response to determining that the configuration content matches the configuration intent, determining that the first configuration information is the target configuration information for running the target program.

[0006] According to another aspect of the present disclosure, a program running method is provided, comprising: in response to receiving a running instruction for a target program, obtaining target configuration information; wherein the target configuration information is generated based on the above-mentioned large model-based configuration information generation method; and loading the target configuration information into the target program, and restarting the target program.

[0007] According to another aspect of the present disclosure, a configuration information generation device based on a large model is provided, comprising: a first generation module, a second generation module, and a first determination module. The first generation module is configured to generate first configuration information in a target format based on an initial text using the large model, wherein the initial text is used to describe the configuration intent for the target program. The second generation module is configured to generate a target text based on the first configuration information using the large model; wherein the target text is used to describe the configuration content of the first configuration information. The determination module is configured to determine that the first configuration information is the target configuration information for the target program operation in response to determining that the configuration content matches the configuration intent.

[0008] According to another aspect of the present disclosure, a program execution device is provided, comprising: an acquisition module and an execution module. The acquisition module is configured to, in response to receiving an execution instruction for a target program, acquire target configuration information; wherein the target configuration information is generated based on a configuration information generation method of a large model. The execution module is configured to load the target configuration information into the target program and restart the target program.

[0009] According to another aspect of the present disclosure, an artificial intelligence agent is provided, comprising: an input module for receiving input information; a processing module for determining a target task based on the input information received by the input module, determining a large model based on the target task, and obtaining output information by calling the large model to execute the method described above; and an output module for outputting the output information obtained by the processing module.

[0010] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described above.

[0011] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method described above.

[0012] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the method described above when executed by a processor.

[0013] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0015] Figure 1 Schematically illustrates an exemplary system architecture to which a configuration information generation method based on a large model, a program running method, and an apparatus according to an embodiment of the present disclosure can be applied;

[0016] Figure 2 The flowchart of the method for generating configuration information based on a large model according to an embodiment of the present disclosure is schematically shown;

[0017] Figure 3AA schematic diagram schematically illustrates a method for generating configuration information based on a large model according to an embodiment of the present disclosure;

[0018] Figure 3B A schematic diagram schematically illustrates a method for generating configuration information based on a large model according to another embodiment of the present disclosure;

[0019] Figure 4 Schematically shows a schematic diagram of generating first configuration information based on initial text using a large model according to an embodiment of the present disclosure;

[0020] Figure 5 A schematic diagram of generating a target text based on first configuration information using a large model according to an embodiment of the present disclosure is schematically shown;

[0021] Figure 6A A schematic diagram of a closed-loop update prompt of a configuration information generation method based on a large model according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 6B Schematically illustrates a closed-loop update Prompt of a configuration information generation method based on a large model according to another embodiment of the present disclosure;

[0023] Figure 6C Schematically illustrates a closed-loop update Prompt of a configuration information generation method based on a large model according to yet another embodiment of the present disclosure;

[0024] Figure 7 A flowchart schematically illustrates a method for executing a program according to an embodiment of the present disclosure;

[0025] Figure 8 Schematically shows a block diagram of a configuration information generating device based on a large model according to an embodiment of the present disclosure;

[0026] Figure 9 A block diagram schematically illustrates a program execution device according to an embodiment of the present disclosure;

[0027] Figure 10 A block diagram schematically illustrates a structure of an artificial intelligence agent according to an embodiment of the present disclosure; and

[0028] Figure 11 A block diagram of an electronic device suitable for implementing a configuration information generating method or a program running method based on a large model according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0029] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0030] In related examples, parameter configuration for target programs is typically based on manual experience and rule engines. However, for scenarios such as resource intervention and user group targeting in recommendation systems, advertising delivery, and duration analysis, these scenarios involve large amounts of data and complex and ever-changing logical relationships between data. Configuration templates created based on manual experience are difficult to adapt to rapidly changing scenario requirements, resulting in poor flexibility and inefficient configuration, which can affect normal program operation.

[0031] Figure 1 An exemplary system architecture is schematically shown to which a configuration information generation method based on a large model, a program running method, and an apparatus can be applied according to an embodiment of the present disclosure.

[0032] It should be noted that Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present disclosure may be applied, to help those skilled in the art understand the technical content of the present disclosure, but do not imply that the embodiments of the present disclosure may not be applied to other devices, systems, environments, or scenarios. For example, in another embodiment, an exemplary system architecture to which the method and apparatus for generating configuration information based on a large model may be applied may include a terminal device, but the terminal device may implement the method and apparatus for generating configuration information based on a large model provided by the embodiments of the present disclosure without interacting with a server.

[0033] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a terminal device 101 , a large model 102 , a database 103 , a network 104 and a server 105 .

[0034] The terminal device 101 may be any electronic device having a display screen and supporting web browsing, including but not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, and the like.

[0035] It should be noted that the configuration information generation method and program execution method based on the large model provided in the embodiments of the present disclosure can generally be executed by the terminal device 101. Accordingly, the configuration information generation device and program execution device based on the large model provided in the embodiments of the present disclosure can also be set in the terminal device 101.

[0036] For example, the terminal device 101 can obtain model parameters from the server 105 via the network 104 and load the large model 102 into the terminal device 101. The user can input the description text by voice input or keyboard input, and the terminal device 101 uses the large model 102 to generate configuration information based on the description text. Then, the large model 102 is used to generate the target text based on the configuration information. The target text can be displayed to the user through the terminal device 101 so that the user can compare the configuration content described in the target text with their own configuration intentions. When the user determines that the configuration content meets their user intentions, they can perform a confirmation operation to store the configuration information generated by the large model 102 in the database 103. When the terminal device 101 needs to run the target program, it obtains the configuration information from the database 103, loads it into the target program, and restarts the target program to complete the operation.

[0037] Alternatively, the configuration information generation method based on the big model provided in the embodiment of the present disclosure may also be generally executed by the server 105. Accordingly, the configuration information generation apparatus based on the big model provided in the embodiment of the present disclosure may generally be set in the server 105. The configuration information generation method based on the big model provided in the embodiment of the present disclosure may also be executed by a server or server cluster that is different from the server and can communicate with the terminal device 101 and / or the server 105. Accordingly, the configuration information generation apparatus based on the big model provided in the embodiment of the present disclosure may also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal device 101 and / or the server 105.

[0038] For example, the terminal device 101 may send the description text input by the user to the server 105 via the network 104. The server 105 generates target configuration information by executing the method of the embodiment of the present disclosure, and returns the target configuration information to the terminal device 101 via the network 104. The target configuration information is then stored in the database 103. When the terminal device 101 needs to run the target program, it obtains the configuration information from the database 103, loads it into the target program, and restarts the target program to complete the operation.

[0039] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0040] In the technical solution disclosed herein, the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information involved comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0041] In the technical solution disclosed herein, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0042] Figure 2 The flowchart of the method for generating configuration information based on a large model according to an embodiment of the present disclosure is schematically shown.

[0043] like Figure 2 As shown, the method includes operations S210 to S230.

[0044] In operation S210 , first configuration information in a target format is generated based on an initial text using a large model.

[0045] In operation S220 , a target text is generated based on the first configuration information using the large model.

[0046] In operation S230 , in response to determining that the configuration content matches the configuration intention, the first configuration information is determined to be target configuration information for running the target program.

[0047] According to an embodiment of the present disclosure, the initial text may be used to describe the configuration intention for the target program. The specific content of the initial text may depend on the specific configuration scenario, and the embodiment of the present disclosure does not specifically limit this.

[0048] For example, for resource intervention scenarios, the initial text might include the target program's scope of effectiveness, the user group scope, the resource scope, and the intervention strategy. For example, the target program might be a video player, the scope of effectiveness might be channel 1, the user group might be males aged 18-25, the resource scope might be video resources, and the intervention strategy might be filtering for low-traffic numbers like xxx.

[0049] According to an embodiment of the present disclosure, the target format may be determined according to the parameter configuration rules of the target program, for example, it may be in json format.

[0050] According to an embodiment of the present disclosure, the target text is used to describe the configuration content of the first configuration information. The specific content of the target text may also be determined according to a specific configuration scenario, and the embodiment of the present disclosure does not specifically limit this.

[0051] For example, for resource intervention application scenarios, the target text may also include the scope of effectiveness of the target program, the scope of user groups, the scope of resources, and intervention strategies.

[0052] According to an embodiment of the present disclosure, the large model may be any large language model with natural language processing capabilities, such as ERNIE Bot.

[0053] For example, Prompt A can be constructed based on the initial text to generate configuration information. Then, Prompt A is input into the large model to output the first configuration information. Next, Prompt B can be constructed based on the first configuration information to generate the target text, and the target text can be output.

[0054] In some embodiments, prompts A and B can be generated based on a static prompt template. The static prompt template can be generated by analyzing historical configuration cases in a specific scenario and extracting characteristic data from the historical configuration cases. Based on the characteristic data, the template is then constructed according to the format requirements of the target program's configuration parameters.

[0055] For example, a static prompt template may include a task description, configuration rules, and reference examples. The configuration rules in prompts A and B can be the same or similar. The reference example in prompt A is an example of generating configuration information based on a text description, while the reference example in prompt B is an example of generating a text description based on configuration information.

[0056] In some embodiments, the reference examples in Prompt A and Prompt B can also be dynamically updated. For example, when the configuration content described in the target text meets the configuration intent, the target configuration information and the initial text can be used as correct examples to update the Prompt template. When the configuration content described in the target text does not meet the configuration intent, the first configuration information and the initial text can be used as incorrect examples to update the Prompt template, further improving the accuracy of configuration information generated by the large model.

[0057] The degree of match between the configuration content described in the target text and the configuration intent described in the initial text can be determined based on manual verification or based on the similarity of intent keywords.

[0058] For example, the target text can be displayed on a visual interface so that users can determine whether the configuration content described by the target text meets their configuration intentions. When the current configuration content is determined to meet their configuration intentions, the user can click the confirmation button in the visual interface to confirm the current configuration information as the target configuration information.

[0059] In some embodiments, a multi-level manual review mechanism can be configured, for example: Person A with one year of experience in the field, Person B with three years, and Person C with more than five years of experience. Weights for different levels of manual review can be pre-configured, and the scores for each level of manual review can be weighted to produce a final score. The weighting can be varied based on the experience of different personnel.

[0060] In some embodiments, intent keywords may be extracted from the target text and the initial text respectively for similarity matching to determine whether the configuration content described in the target text conforms to the configuration intent described in the initial text.

[0061] In some embodiments, a large model may be used to perform intent matching between the target text and the initial text to determine whether the configuration content described in the target text conforms to the configuration intent described in the initial text.

[0062] According to the embodiments of the present disclosure, the natural language understanding capability of the large model can be utilized to generate the first configuration information while fully understanding the configuration intent described in the initial text. Then, the content described in the first configuration information can be generated while fully understanding the first configuration information. Based on multiple rounds of interaction of the large model, the ability of the large model to understand the configuration intent can be continuously improved until configuration information that meets the configuration intent is generated, thereby reducing the dependence of the configuration parameters on human experience and improving the accuracy of the configuration parameters and the configuration efficiency.

[0063] The logical relationships between configuration parameters vary in different application scenarios. For example, when targeting user groups, the criteria can be based on a combination of factors such as age, gender, and historical preferences. For example, targeting males aged 18-25, or individuals under 40 with at least 10 years of experience in a specific technical field.

[0064] Therefore, based on the logical hierarchical relationship between different configuration parameters, the corresponding Prompt template can be matched to the initial text, so that the large model can better understand the coordination intention.

[0065] According to an embodiment of the present disclosure, using a large model, generating first configuration information based on a first text may include the following operations: parsing the initial text to obtain a first logical hierarchical relationship between fields used to represent the configuration intent; and using the large model to generate the first configuration information based on the first logical hierarchical relationship and the initial text.

[0066] For example, the initial text can be a vague description of the configuration intent, such as "I want to conduct an experiment and need to target xx users for video resource yy." It can also be a clear description of the configuration intent, such as "Effective scope: Channel 1; Group targeting: Males aged 18-25 who do not work in a certain field; Resource targeting: Video resources; Experimental plan: Hit low-traffic number xxx and filter it." The first logical level of this configuration intent includes four parallel nodes: "Effective scope," "Group targeting," "Resource targeting," and "Experimental plan." The second logical level corresponding to the "Group targeting" node includes: Males aged 18-25 who do not work in a certain field. The fields "18-25 years old" and "Male" in this node have an "and" relationship. The relationship between "Male" and "Working in a certain field" is "not."

[0067] By analyzing the logical hierarchical relationship of the configuration intent, a Prompt template with the same or similar logical hierarchical relationship as the above can be determined from the candidate Prompt template library. Prompt templates can also be dynamically generated based on the logical hierarchical relationship and in accordance with the configuration rules of the target format. For example, the logical hierarchical relationship can be split into multiple single-layer logical relationships, and then the corresponding configuration rules and examples of each single-layer logical relationship can be extracted from the candidate Prompt template library. Then, the corresponding configuration rules and examples of each single-layer logical relationship are combined according to the original logical hierarchical relationship to dynamically generate a Prompt template.

[0068] By parsing the logical hierarchical relationship of configuration intent and determining the prompt template based on the logical hierarchical relationship, the big model's ability to understand configuration intent is further improved, and the big model is able to generate configuration information that meets the user's configuration intent, reducing the number of repeated interactions with the big model, and further improving the accuracy and efficiency of configuration information generation.

[0069] Reference below Figure 3A to Figure 6C , combined with specific embodiments Figure 2 The method shown is further explained.

[0070] In some embodiments, a prompt template with the same or similar logical hierarchical relationship as the above can be determined from a candidate prompt template library. For example, a first prompt text matching the first logical hierarchical relationship is obtained; the initial text and the first prompt text are input into the macro model, and the first configuration information is output.

[0071] Figure 3A A schematic diagram of a configuration information generation method based on a large model according to an embodiment of the present disclosure is schematically shown.

[0072] like Figure 3AAs shown, in embodiment 300A, first, the initial text 311 is parsed to obtain a logical hierarchical relationship 312. Based on the logical hierarchical relationship 312, Prompt A313 is matched. The [Task] in Prompt A313 can be "Convert the input text into a string that conforms to the json logic specification." The configuration rules can include [Constraints] and [Rule Descriptions]. [Constraints] can be "The json object must include X, Y keywords and corresponding data." [Rule Descriptions] can include keyword descriptions, field configuration descriptions, logical rule descriptions, etc. [Example] can be an example of using description information xxx to generate the corresponding configuration yyy.

[0073] Then, the initial text 311 is filled into the text input position in Prompt A 313 , and then input into the large model 310 to generate a json string 321 .

[0074] Next, Prompt B322 can be matched based on JSON string 321. The [Task] in Prompt B322 can be "Translate the input JSON string into a natural language description." The [Constraint] can be "The natural language description must include the following core elements C and D." The [Rule Description] can be the same as the rule description in Prompt A. The [Example] can be an example of generating the corresponding description text nnn from the JSON string mmm.

[0075] Then, the json string 321 and Prompt B 322 are input into the large model 310 to generate the target text 323.

[0076] Finally, operation S330 is performed to match the configuration content described in the target text 323 with the configuration intent described in the initial text 311. If the configuration content matches the configuration intent, operation S331 is performed to determine that the JSON string is the target configuration. If the configuration content does not match the configuration intent, the process returns to reusing the large model to generate a new JSON string based on the initial text.

[0077] Since the configuration rules and examples in the Prompt template have the same or similar logical hierarchical relationship with the configuration intent, it is helpful for the big model to understand the configuration intent and generate configuration information that meets the configuration intent according to the configuration rules based on the examples in the template, further improving the efficiency of configuration information generation, reducing the number of repeated interactions between users and the big model, reducing operation complexity, and improving user experience.

[0078] In some embodiments, different people may use different language descriptions for the same configuration intent, and such differences may affect the large model's accurate understanding of the configuration intent.

[0079] Therefore, before parsing the initial text or processing the initial text using a large model, intent keywords can be extracted from the initial text; and based on the intent keywords, the initial text can be enhanced using a large model to generate an enhanced intent text.

[0080] Figure 3B The following schematically illustrates a method for generating configuration information based on a large model according to another embodiment of the present disclosure.

[0081] like Figure 3B As shown, in embodiment 300B, the difference from embodiment 300A is that before parsing the initial text, intent keywords 341 are first extracted from the initial text 311. Then, Prompt C 342 is matched based on the intent keywords 341. Prompt C 342 may include reference examples of intent reinforcement tasks and intent reinforcement. Next, the intent keywords 341 and Prompt C 342 are input into the large model 310, and the intent reinforcement text 343 is output. The same operation as that for the initial text 311 in embodiment 300A is performed based on the intent reinforcement text 343. No further details are given here.

[0082] By strengthening the intent of the initial text, the impact of different people's language expression ability on the large model's ability to understand the configuration intent is reduced, further improving the accuracy of the large model in intent recognition.

[0083] As market demands change rapidly and application scenarios become increasingly complex, pre-set prompt templates based on historical configuration examples are difficult to adapt to flexible and changing configuration intentions.

[0084] Therefore, in some embodiments, the initial text can be split into multiple sub-texts according to the first logical hierarchical relationship; the second prompt texts matching each single-layer logical relationship are obtained; the initial text and the second prompt texts are input into the large model, and the configuration sub-information is output; and the configuration sub-information is combined according to the first logical hierarchical relationship to generate the first configuration information.

[0085] Figure 4 The diagram schematically shows a method of generating first configuration information based on initial text using a large model according to an embodiment of the present disclosure.

[0086] like Figure 4 As shown, the logical hierarchical relationship 312 is obtained by parsing the initial text 311. Then, the initial text is split into sub-texts T1411~sub-texts T n41n. Each subtext is used to describe a single-layer logical relationship of configuration intent. For example, the subtext may only include any one of the logical relationships "and," "or," or "not." The logical hierarchical relationship 312 may be a complex logical link or logical topology structure composed of single-layer logical relationships.

[0087] Then, based on the single-layer logical relationship corresponding to each sub-text, the corresponding Prompt1~Prompt n Then, each sub-text and its corresponding prompt are input into the large model 310 to generate character strings J1421~J n 42n.

[0088] Finally, the character strings are recombined according to the logical hierarchical relationship 312 to generate a json character string 321.

[0089] By splitting the initial text into multiple sub-texts with a single-level logical relationship that describes the configuration intent, there's no need to pre-configure prompt templates with complex logical relationships. Furthermore, using the large model to generate strings based on the sub-texts of the single-level logical relationship reduces the difficulty for the large model to understand complex logic. Combining them according to the initial logical hierarchy allows for more flexible adaptation to a variety of complex and changing application scenarios, improving configuration flexibility.

[0090] Using the big model to generate the target text based on the first configuration information can include the following operations: parsing the first configuration information to obtain the second logical hierarchical relationship between the fields in the first configuration information; and using the big model to generate the target text based on the second logical hierarchical relationship and the first configuration information.

[0091] For configuration information in JSON format, you can determine the logical hierarchical relationship by identifying logical identifiers, such as "equal", "and", "or", etc.

[0092] By analyzing the logical hierarchical relationship of the first configuration information, a prompt template with the same or similar logical hierarchical relationship can be determined from the candidate prompt template library. Prompt templates can also be dynamically generated based on the logical hierarchical relationship and in accordance with the configuration rules of the target format. The method for generating a dynamic prompt template can be the same as the method for generating a prompt template for generating configuration information described above, and will not be repeated here.

[0093] By parsing the logical hierarchical relationship of the first configuration information and determining the Prompt template based on the logical hierarchical relationship, the big model's ability to understand the configuration content is further improved, and the big model is able to generate text that can accurately describe the configuration content, so that users can determine whether the first configuration information meets the configuration intent. This reduces the complexity of verification, reduces the dependence of configuration parameters on human experience, and further improves the user experience.

[0094] Using the big model to generate the target text based on the second logical hierarchical relationship and the first configuration information can include the following operations: obtaining a third prompt text that matches the second logical hierarchical relationship; inputting the first configuration information and the third prompt text into the big model, and outputting the target text.

[0095] According to an embodiment of the present disclosure, the third prompt text includes a configuration rule for indicating a logical hierarchical relationship between fields in the configuration information of the target format and a third reference example.

[0096] Figure 5 A schematic diagram of generating a target text based on first configuration information using a large model according to an embodiment of the present disclosure is schematically shown.

[0097] like Figure 5 As shown, in this embodiment 500, the logical hierarchical relationship 323 is obtained by parsing the json string 321. Based on the logical hierarchical relationship 323, the Prompt T e1 521. The Prompt T e1 The [Task] in 521 can be "Translate the input JSON string into a natural language description." Configuration rules can include [Constraints] and [Rule Descriptions]. A [Constraint] can be "The natural language description must include the following core elements: 'C' and 'D'." A [Rule Description] can include keyword descriptions, field configuration instructions, logical rule descriptions, etc. An [Example] can be an example of using the JSON string sss to generate the corresponding description text ttt.

[0098] In some embodiments, the matched Prompt T e1 521 and the json string 321 are directly input into the large model 310 to output the target text.

[0099] In some embodiments, operations for interacting with the user can be added by displaying a Prompt T e1 521, to confirm Prompt T e1 Whether the description text in the reference example in 521 is similar to or matches the user's configuration intention, thereby further improving the large model's ability to understand the content of the json string 321.

[0100] When it is determined that the reference example matches the user's configuration intention, the Prompt T e1 521 and the json string 321 are directly input into the large model 310 to output the target text Ta 323a.

[0101] When it is determined that the reference example does not match the user's configuration intention, for example, when the user clicks the update button in the visual interface or the user voice "mismatch" is detected, the Prompt T in the current visual interface e1 521 can enter the edit state so that the user can change the Prompt T e1 Reference example in 521, generate Prompt T e2 522.

[0102] Then, change Prompt T e2 522 and the json string 321 are directly input into the large model 310 to output the target text Tb323b.

[0103] When the large model fully understands the configuration information content, it generates text to describe the configuration content, which improves the configuration efficiency in complex scenarios.

[0104] Figure 6A The following schematically illustrates a closed-loop update prompt of the configuration information generation method based on a large model according to an embodiment of the present disclosure.

[0105] like Figure 6A As shown in embodiment 600A, user A, who has been working for one year, can first enter the intent description into platform 610 by voice or manual input. Platform 610 can be the running terminal of the target program. By calling the large model 310, a JSON string is generated based on the intent description and Prompt A.

[0106] Then, by calling the large model 310, a content description is generated based on the json string and Prompt B.

[0107] Next, the manual verification phase begins. The content description and intent description can be presented to User A and User B, who has been in the industry for three years. User A scores the configuration based on the degree of match between the content and intent, resulting in a score S1 601. User B scores the configuration based on the degree of match between the content and intent, resulting in a score S2 602. Based on a predetermined weighting ratio, the scores S1 601 and S2 602 can be weighted to obtain a total score 603.

[0108] In some embodiments, multiple rounds of interaction can be performed with the large model for the same intent description, resulting in multiple scores for the same intent description and the JSON string generated each time. The higher-scoring intent description and its corresponding JSON string are used as examples to update Prompt A. Similarly, the higher-scoring intent description and its corresponding JSON string can also be used as examples to update Prompt B.

[0109] In some embodiments, multiple rounds of interaction can be performed with the large model for different intent descriptions, thereby obtaining multiple scores for different intent descriptions and the JSON strings generated each time. Similarly, the intent descriptions with higher scores and the corresponding JSON strings can be used as examples to update Prompt A. Similarly, the intent descriptions with higher scores and the corresponding JSON strings can also be used as examples to update Prompt B. This makes the interaction of the large model form a complete closed-loop process. As the number of large-model interactions increases, the large model's ability to understand the configuration intent will continue to improve.

[0110] Figure 6B The following schematically illustrates a closed-loop update prompt of a configuration information generation method based on a large model according to another embodiment of the present disclosure.

[0111] like Figure 6B As shown, in embodiment 600B, the difference from embodiment 600A is that the content description and the intent description can be matched based on similarity or a large model. For example, intent keywords can be extracted from the content description and the intent description respectively, and intent matching can be performed based on the similarity between the intent keywords. An intent recognition prompt can also be constructed based on the content description and the intent description, and the intent matching prompt is input into the large model to obtain an output intent matching result, thereby reducing the impact of subjective judgment based on professional experience on the configuration result, further improving the intelligence of the configuration process and improving configuration efficiency.

[0112] In some embodiments, in response to determining that the configuration content does not match the configuration intent, the target text is used as an error example, and the large model is used to correct the first configuration information until second configuration information that matches the configuration intent is generated.

[0113] For example, the intent description and the JSON string generated in that round can be used as an error example 604 to update PromptA, reminding the large model not to generate the same configuration information in the next configuration generation. This enables the large model to self-correct its generated content. Similarly, for configurations that meet the user's intent, the intent description and the JSON string generated in that round can be used as a correct example 603 to update PromptA, improving the large model's understanding of configuration intent and enhancing the accuracy of configuration generation.

[0114] In some embodiments, the first configuration information generated by the large model may be format-checked before being processed using the large model to generate the target text. For example, the first configuration information may be format-checked; and in response to determining that the format of the first configuration information does not match the target format, the first configuration information may be used as an error example and corrected using the large model until third configuration information in the target format is generated.

[0115] Figure 6C The following schematically illustrates a closed-loop update Prompt of a configuration information generation method based on a large model according to another embodiment of the present disclosure.

[0116] like Figure 6C As shown, the difference between embodiment 600C and embodiment 600B is that a format verification process for the json string is added.

[0117] When it is determined that the json string matches the target format, a content description is generated based on the json string and Prompt B by calling the large model 310 .

[0118] When it is determined that the JSON string does not match the target format, the JSON string generated in the current round and the intent description can be used as an error example to update Prompt A, reminding the large model not to generate the configuration information in the error example when generating the configuration next time.

[0119] Similarly, for a JSON string whose format matches the target format, Prompt A can be updated with the intent description and the JSON string generated in this round as a correct example.

[0120] By adding a format verification step, the probability of matching content description and intent for malformed JSON strings is reduced, further improving configuration generation efficiency.

[0121] Figure 7 The flowchart of the program running method according to the embodiment of the present disclosure is schematically shown.

[0122] like Figure 7 As shown, the method 700 includes operation S710 and operation S720.

[0123] In operation S710 , in response to receiving an execution instruction for a target program, target configuration information is acquired.

[0124] In operation S720, the target configuration information is loaded into the target program, and the target program is restarted.

[0125] According to an embodiment of the present disclosure, the target configuration information may be generated using the method described above. The target configuration information may be stored in a database. When a run instruction for a target program is received, the target configuration information may be retrieved from the database. The target configuration information may then be loaded into the target program, and the target program may be restarted.

[0126] After multiple rounds of interaction, the large model's ability to understand configuration intent is continuously improved, thereby realizing intelligent configuration for any scenario, reducing the impact of professional experience on configuration quality and efficiency, and further improving the efficiency of program operation.

[0127] Figure 8 The block diagram of the apparatus for generating configuration information based on a large model according to an embodiment of the present disclosure is schematically shown.

[0128] like Figure 8 As shown, the generating device 800 may include a first generating module 810 , a second generating module 820 and a determining module 830 .

[0129] The first generating module 810 is configured to generate first configuration information in a target format based on an initial text using a large model, wherein the initial text is used to describe configuration intentions for a target program.

[0130] The second generating module 820 is used to generate a target text based on the first configuration information using the large model; wherein the target text is used to describe the configuration content of the first configuration information.

[0131] The determination module 830 is configured to determine that the first configuration information is the target configuration information in response to determining that the configuration content matches the configuration intention.

[0132] According to an embodiment of the present disclosure, the first generation module may include: a first parsing submodule and a first generation submodule. The first parsing submodule is configured to parse the initial text to obtain a first logical hierarchical relationship between fields used to represent the configuration intent. The first generation submodule is configured to generate first configuration information based on the first logical hierarchical relationship and the initial text using a large model.

[0133] According to an embodiment of the present disclosure, the first generation submodule may include: a first acquisition unit and a first output unit. The first acquisition unit is configured to acquire first prompt text that matches the first logical hierarchical relationship; the first prompt text includes a configuration rule and a first reference example for indicating the logical hierarchical relationship between fields in the configuration information of the target format. The first output unit is configured to input the initial text and the first prompt text into the large model to output the first configuration information.

[0134] According to an embodiment of the present disclosure, the first generating submodule may include: a splitting unit, a second acquiring unit, a second output unit, and a combining unit.

[0135] A splitting unit is used to split the initial text into multiple sub-texts according to the first logical hierarchical relationship; wherein each sub-text is used to describe the single-layer logical relationship of the configuration intention. A second acquisition unit is used to obtain each second prompt text that matches each single-layer logical relationship; wherein the second prompt text includes a configuration rule and a second reference example for indicating the single-layer logical relationship between each field in the configuration information of the target format. A second output unit is used to input the initial text and each second prompt text into the large model to obtain the output configuration sub-information. A combination unit is used to combine each configuration sub-information according to the first logical hierarchical relationship to generate the first configuration information.

[0136] According to an embodiment of the present disclosure, the second generation module may include: a second parsing submodule and a second generation submodule. The second parsing submodule is configured to parse the first configuration information to obtain a second logical hierarchical relationship between fields in the first configuration information. The second generation submodule is configured to generate a target text based on the second logical hierarchical relationship and the first configuration information using a large model.

[0137] According to an embodiment of the present disclosure, the second generation submodule may include: a second acquisition unit and a second output unit. The second acquisition unit is configured to acquire third prompt text that matches the second logical hierarchical relationship; the third prompt text includes configuration rules and a third reference example for indicating the logical hierarchical relationship between fields in the configuration information of the target format. The second output unit is configured to input the first configuration information and the third prompt text into the large model to output the target text.

[0138] According to an embodiment of the present disclosure, the device 800 may also include a first correction module for, in response to determining that the configuration content does not match the configuration intent, taking the target text as an error example and using a large model to correct the first configuration information until second configuration information matching the configuration intent is generated.

[0139] According to an embodiment of the present disclosure, the apparatus 800 may further include a second correction module.

[0140] The second correction module is configured to, in response to determining that the format of the first configuration information does not match the target format, take the first configuration information as an error example and correct the first configuration information using the large model until third configuration information in the target format is generated.

[0141] According to an embodiment of the present disclosure, the apparatus 800 may further include an extraction module and an intention reinforcement module.

[0142] The extraction module is used to extract intent keywords from the initial text. The intent enhancement module is used to enhance the intent of the initial text based on the intent keywords using a large model to generate enhanced intent text.

[0143] Figure 9 The block diagram schematically shows a program execution device according to an embodiment of the present disclosure.

[0144] like Figure 9 As shown, the apparatus 900 may include an acquisition module 910 and an operation module 920 .

[0145] The acquisition module 910 is configured to acquire target configuration information in response to receiving a running instruction for a target program; wherein the target configuration information is generated based on the large model-based configuration information generation method described above.

[0146] The running module is used to load the target configuration information into the target program and restart the target program.

[0147] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0148] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described above.

[0149] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method described above.

[0150] According to an embodiment of the present disclosure, a computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements the method described above.

[0151] Figure 10 The structural block diagram of an artificial intelligence agent according to an embodiment of the present disclosure is schematically shown.

[0152] In the embodiments of the present disclosure, inspired by the von Neumann structure in modern computer theory, such as Figure 10 As shown, the AI ​​agent 1000 may include three core modules: an input module 1010, an output module 1020, and a processing module 1030. The processing module 1030 may include a control unit 1031, a storage unit 1032, and an operation unit 1033.

[0153] Input module 1010 is responsible for receiving or sensing information such as queries, requests, instructions, signals, or data from the outside world (e.g., users or the external environment) and converting it into a format that AI agent 1000 can understand and process. Input module 1010 is the primary link for AI agent 1000 to interact with the outside world. It enables AI agent 1000 to efficiently and accurately obtain necessary "sensory" information from the outside world and respond to this information.

[0154] In an example, the input information received by the input module 1010 may be the initial text described above.

[0155] In this example, processing module 1030 is the core support for AI agent 1000's ability to handle complex tasks. Processing module 1030 can determine a target task based on the input information received by input module 1010, determine a large model based on the target task, and execute the large model-based configuration information generation method described above by calling the large model to output target configuration information.

[0156] In the example, the control unit 1031 in the processing module 1030 will continuously interact with the storage unit 1032, the computing unit 1033, and / or the output module 1020 during operation. However, it should be noted that in the embodiment of the present disclosure, the control unit 1031 acts as a single initiator to initiate communication with the storage unit 1032, the computing unit 1033, and / or the output module 1020, and there is no communication coupling between the storage unit 1032, the computing unit 1033, and the output module 1020.

[0157] In this example, the performance of the control unit 1031 can be closely related to the large model on which the AI ​​agent 1000 is based. To fully utilize the capabilities of the large language model, the internal structure of the control unit 1031 can be designed to be highly configurable and scalable to cope with various types of tasks and requirements in real scenarios.

[0158] The storage unit 1032 may be responsible for memorizing information such as historical conversations, event flows, etc. The configuration information, target text, and data resources generated in each round as described above may be included in the storage unit 1032 .

[0159] In the example, after the AI ​​agent 1000 obtains the configuration generation request, the AI ​​agent 1000 can use the intent recognition model to determine the configuration intent from the initial text. The configuration intent can be stored in the storage unit 1032. The AI ​​agent 1000 can retrieve relevant data resources from the storage unit 1032 and feed it back to the control unit 1031. Then, the control unit 1031 can use the fed-back data resources to obtain the configuration data corresponding to the initial text. It can also retrieve relevant text data from the storage unit 1032 and feed it back to the control unit 1031. Then, the control unit 1031 can use the returned text data to obtain the target text. And pass the target text and configuration data to the output module 1020.

[0160] The computing unit 1033 can be viewed as a predefined tool library, and the renderer and presentation controls mentioned above can be included in the computing unit 1033 .

[0161] In the example, when the AI ​​agent 1000 needs to render multiple output data, it can call the relevant renderer and display control from the operation unit 1033 and feed it back to the control unit 1031. Then, the control unit 1031 can use the feedback renderer and display control to render the first search result and pass the first search result to the output module 1020. It can be understood that although the large language model has excellent language understanding and generation capabilities, it is the same as a human. Without the help of any tools, the tasks that can be solved are very limited. When the AI ​​agent 1000 is given the ability to call tools, it can achieve tasks such as completing mathematical operations with the help of a calculator, completing data analysis with the help of Python, and completing prediction tasks with the help of a search engine.

[0162] In an example, the output module 1020 may output the target configuration information and target text described above.

[0163] The AI ​​agent 1000 according to the embodiment of the present disclosure can simply and effectively improve the level of intelligence, and enhance flexibility and versatility.

[0164] Figure 11A schematic block diagram of an example electronic device 1110 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0165] like Figure 11 As shown, device 1110 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. RAM 1103 may also store various programs and data required for the operation of device 1100. Computing unit 1101, ROM 1102, and RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to bus 1104.

[0166] Various components in device 1100 are connected to I / O interface 1105, including an input unit 1106, such as a keyboard and mouse; an output unit 1107, such as various types of displays and speakers; a storage unit 1108, such as a magnetic disk and optical disk; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0167] Computing unit 1101 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Computing unit 1101 performs the various methods and processes described above, such as the large-model-based configuration information generation method or program execution method. For example, in some embodiments, the large-model-based configuration information generation method or program execution method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed onto device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by computing unit 1101, one or more steps of the large-model-based configuration information generation method or program execution method described above can be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to execute the configuration information generating method or the program running method based on the large model in any other appropriate manner (for example, by means of firmware).

[0168] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-a-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0169] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0170] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0171] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0172] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0173] A computer system may include clients and servers. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a server, a server in a distributed system, or a server in conjunction with a blockchain.

[0174] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0175] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for generating configuration information based on a large model, comprising: Generate first configuration information in a target format based on an initial text using the large model, wherein the initial text is used to describe a configuration intention for a target program; Using the large model, generating a target text based on the first configuration information; wherein the target text is used to describe the configuration content of the first configuration information; and In response to determining that the configuration content matches the configuration intention, the first configuration information is determined to be target configuration information for running the target program.

2. The method according to claim 1, wherein The step of generating the first configuration information based on the initial text using the large model includes: Parsing the initial text to obtain a first logical hierarchical relationship between the fields used to represent the configuration intent; and The first configuration information is generated based on the first logical hierarchical relationship and the initial text using the large model.

3. The method according to claim 2, wherein: The generating the first configuration information by using the large model based on the first logical hierarchical relationship and the initial text includes: Obtaining a first prompt text that matches the first logical hierarchical relationship; wherein the first prompt text includes a configuration rule and a first reference example for indicating a logical hierarchical relationship between fields in the configuration information of the target format; and The initial text and the first prompt text are input into the large model, and the first configuration information is output.

4. The method according to claim 2, wherein: The generating the first configuration information by using the large model based on the first logical hierarchical relationship and the initial text includes: Splitting the initial text into a plurality of sub-texts according to the first logical hierarchical relationship; wherein each sub-text is used to describe a single-layer logical relationship of the configuration intention; Acquire each second prompt text matching each single-layer logical relationship; wherein the second prompt text includes a configuration rule and a second reference example for indicating the single-layer logical relationship between each field in the configuration information of the target format; Inputting the initial text and each of the second prompt texts into the large model to obtain output configuration sub-information; and The configuration sub-information is combined according to the first logical hierarchical relationship to generate the first configuration information.

5. The method according to claim 1, wherein Generating a target text based on the first configuration information using the large model includes: Parsing the first configuration information to obtain a second logical hierarchical relationship between fields in the first configuration information; and The target text is generated by utilizing the large model based on the second logical hierarchical relationship and the first configuration information.

6. The method according to claim 5, wherein: The generating the target text by using the large model based on the second logical hierarchical relationship and the first configuration information includes: Obtaining a third prompt text that matches the second logical hierarchical relationship; wherein the third prompt text includes a configuration rule and a third reference example for indicating a logical hierarchical relationship between fields in the configuration information of the target format; and The first configuration information and the third prompt text are input into the large model to obtain the output target text.

7. The method according to any one of claims 1 to 6, wherein: The method further comprises: In response to determining that the configuration content does not match the configuration intent, the target text is used as an error example, and the first configuration information is corrected using the large model until second configuration information matching the configuration intent is generated.

8. The method according to any one of claims 1 to 6, wherein: The method further comprises: In response to determining that the format of the first configuration information does not match the target format, the first configuration information is used as an error example and corrected using the large model until third configuration information in the target format is generated.

9. The method according to any one of claims 1 to 6, further comprising: extracting intent keywords from the initial text; as well as Based on the intention keywords, the large model is used to enhance the intention of the initial text to generate an enhanced intention text.

10. A program running method, comprising: In response to receiving a run instruction for a target program, obtaining target configuration information; wherein the target configuration information is generated based on the method according to any one of claims 1 to 9; and The target configuration information is loaded into the target program, and the target program is restarted and run.

11. A configuration information generating device based on a large model, comprising: A first generating module is configured to generate first configuration information in a target format based on an initial text using a large model, wherein the initial text is used to describe a configuration intention for a target program; a second generating module, configured to generate a target text based on the first configuration information using the large model; wherein the target text is used to describe the configuration content of the first configuration information; and A determination module is configured to, in response to determining that the configuration content matches the configuration intention, determine that the first configuration information is target configuration information for running the target program.

12. The device according to claim 11, wherein The first generation module includes: A first parsing submodule is configured to parse the initial text to obtain a first logical hierarchical relationship between the fields used to represent the configuration intent; and The first generating submodule is configured to generate the first configuration information based on the first logical hierarchical relationship and the initial text by using the large model.

13. The device according to claim 12, wherein The first generation submodule includes: a first acquisition unit, configured to acquire a first prompt text matching the first logical hierarchical relationship; wherein the first prompt text includes a configuration rule and a first reference example for indicating a logical hierarchical relationship between fields in the configuration information of the target format; and The first output unit is used to input the initial text and the first prompt text into the large model and output the first configuration information.

14. The device according to claim 12, wherein The first generation submodule includes: a splitting unit, configured to split the initial text into a plurality of sub-texts according to the first logical hierarchical relationship; wherein each sub-text is used to describe a single-layer logical relationship of the configuration intention; A second acquisition unit is configured to acquire second prompt texts matching the single-layer logical relationships; wherein the second prompt texts include a configuration rule and a second reference example for indicating the single-layer logical relationship between the fields in the configuration information of the target format; A second output unit is configured to input the initial text and each of the second prompt texts into the large model to obtain output configuration sub-information; and A combining unit is configured to combine the configuration sub-information according to the first logical hierarchical relationship to generate the first configuration information.

15. The device according to claim 11, wherein The second generation module includes: a second parsing submodule, configured to parse the first configuration information to obtain a second logical hierarchical relationship between fields in the first configuration information; and The second generating submodule is used to generate the target text based on the second logical hierarchical relationship and the first configuration information by using the large model.

16. The device according to claim 15, wherein The second generation submodule includes: a second acquisition unit, configured to acquire a third prompt text matching the second logical hierarchical relationship; wherein the third prompt text includes a configuration rule and a third reference example for indicating a logical hierarchical relationship between fields in the configuration information of the target format; and The second output unit is used to input the first configuration information and the third prompt text into the large model to obtain the output target text.

17. The device according to any one of claims 11 to 16, further comprising: A first correction module is used to, in response to determining that the configuration content does not match the configuration intent, take the target text as an error example and use the large model to correct the first configuration information until second configuration information that matches the configuration intent is generated.

18. The device according to any one of claims 11 to 16, further comprising: A second correction module is configured to, in response to determining that the format of the first configuration information does not match the target format, take the first configuration information as an error example and correct the first configuration information using the large model until third configuration information in the target format is generated.

19. The device according to any one of claims 11 to 16, further comprising: An extraction module, configured to extract intended keywords from the initial text; as well as The intention reinforcement module is used to enhance the intention of the initial text based on the intention keyword using the large model to generate an enhanced intention text.

20. A program running device, comprising: an acquisition module, configured to acquire target configuration information in response to receiving a run instruction for a target program; wherein the target configuration information is generated based on the method according to any one of claims 1 to 9; and The running module is used to load the target configuration information into the target program and restart the target program.

21. An artificial intelligence agent, comprising: An input module, used for receiving input information; a processing module, configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and execute the method according to any one of claims 1 to 10 by calling the large model to obtain output information; as well as An output module is used to output the output information obtained by the processing module.

22. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

23. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-10.

24. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 10.

Images