Large model knowledge base content gray release method, system and device and medium
By creating grayscale versions in the knowledge base and setting an identifier field, the problem of the knowledge base being unable to be released in a grayscale manner was solved, enabling effective version management and rollback, and ensuring the stability of the output of large models.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI SHUHE INFORMATION TECH CO LTD
- Filing Date
- 2025-12-26
- Publication Date
- 2026-05-12
AI Technical Summary
In existing technologies, knowledge bases cannot be released in a canary manner, which affects the output results of large models and makes it impossible to roll back to the previous version.
By creating a grayscale version of the knowledge base, setting its status to offline, obtaining the slice content to be released in grayscale, setting the grayscale identifier field, adjusting the knowledge base to form a grayscale version, and completing the grayscale release by verifying the RAG effect.
The knowledge base can be released in a gray-scale manner, which facilitates version management, supports version rollback, and ensures the stability and reliability of the output results of large models.
Smart Images
Figure CN122018959A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of big data technology and relates to a method for gray-scale release of knowledge base, particularly a method, system, device and medium for gray-scale release of content in a large model knowledge base. Background Technology
[0002] Large Language Models (LLMs) are deep learning models trained on massive amounts of text data, enabling them to generate natural language text or understand the meaning of language text. These models can provide in-depth knowledge and language production on a wide range of topics through training on large datasets. The core idea is to learn patterns and structures of natural language through large-scale unsupervised training, mimicking human language cognition and generation processes to some extent. A knowledge base is a core component providing external knowledge support for large language models; essentially, it dynamically supplements the model's basis for generating answers through structured or unstructured data sets.
[0003] In the process of generating answers using large language models, retrieval-augmented generation (RAG) techniques are generally employed to improve the accuracy of the generated answers. RAG is an artificial intelligence framework that combines information retrieval and text generation, aiming to enhance the accuracy and reliability of large language models (LLMs) in professional question answering. Its core consists of two stages: Retrieval stage: transforming the user's question into a vector and retrieving relevant real-time information from external knowledge bases (such as databases or document collections); Generation stage: inputting the retrieval results as context into the LLM to generate more accurate and evidence-based answers.
[0004] In existing technologies, the knowledge base used in retrieval enhancement generation techniques typically does not support the release of gray-scale versions of knowledge base changes. In particular, for unverified changes, saving and applying them directly may have an adverse effect on the online RAG performance. Furthermore, if the RAG performance of the knowledge base is poor due to changes, it is impossible to roll back to the previous version, thus affecting the output results of the large model. Summary of the Invention
[0005] The purpose of this application is to provide a method, system, device, and medium for canary release of large model knowledge base content, in order to solve the problem in the prior art that the knowledge base cannot be canary released, which affects the output results of large models.
[0006] Firstly, this application provides a method for canary release of content in a large model knowledge base, the method comprising: When a gray-scale release of a knowledge base is required, create a gray-scale version of the knowledge base and set its status to "not online". Obtain the slice content to be published in grayscale, set the grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in the vector database; Read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly; Verify the RAG effect of the grayscale version knowledge base, and complete the grayscale release based on the verification results.
[0007] In one implementation of the first aspect, the grayscale identifier field includes "released", "add" (for grayscale addition), and "remove" (for grayscale deletion).
[0008] In one implementation of the first aspect, the step of reading the grayscale version of the knowledge base and the standard slice content, adjusting the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configuring the grayscale identifier field of the standard slice content accordingly, includes: When the gray version of the knowledge base corresponds to newly added data, the standard slice content is added to the knowledge base to form the gray version knowledge base, and the gray identifier field of the standard slice content is set to add to be added in gray version. When the gray version of the knowledge base corresponds to deleted data, the standard slice content is deleted, and the gray-scale identifier field of the standard slice content is set to be removed in gray-scale deletion to form the gray-scale version knowledge base; When the grayscale version of the knowledge base corresponds to modified data, the initial slice data before modification is obtained from the knowledge base, the grayscale identifier field of the initial slice data is set to be removed during grayscale, the standard slice content is added to the knowledge base, and the grayscale identifier field of the standard slice content is set to be added during grayscale, so as to form the grayscale version knowledge base.
[0009] In one implementation of the first aspect, verifying the RAG effect of the grayscale version knowledge base includes: When a regular user of the knowledge base verifies the RAG effect, the official RAG interface is called, and the slice content of the official version of the knowledge base is retrieved through the first search condition for effect verification. When the owner of the knowledge base verifies the RAG effect, the grayscale RAG interface is invoked, and the slice content of the grayscale version of the knowledge base is retrieved through the second search condition to verify the RAG effect.
[0010] In one implementation of the first aspect, the step of completing the canary release based on the verification results includes: When the verification results meet the preset requirements, the gray release verification is determined to be successful. The current official version of the gray version knowledge base is set to the historical version, and the gray version of the knowledge base is set to the updated official version. Modify the grayscale identifier field of the standard slice content that is to be added in grayscale to "released", and delete the initial slice data that is to be removed in grayscale.
[0011] In one implementation of the first aspect, the step of completing the canary release based on the verification results further includes: When the verification effect does not meet the preset requirements, it is determined that the gray release verification fails, the gray release is canceled, the standard slice content whose gray-scale identifier character is to be added in the gray release is deleted, and the initial slice data whose gray-scale identifier field is to be removed in the gray release is modified to be released, and the gray-scale version is deleted.
[0012] In one implementation of the first aspect, when the verification effect does not meet the preset requirements, the verification is determined to be unsuccessful. Based on the grayscale version, the content of the standard slice is readjusted and then re-released in grayscale.
[0013] Secondly, this application also provides a canary release system for large model knowledge base content, the system comprising: The gray-scale version creation module is used to create gray-scale versions of the knowledge base and set their status to not online when a gray-scale release of the knowledge base is required. The content acquisition module is used to acquire the slice content to be released in grayscale, set the grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in the vector database. The version configuration module is used to read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly. The verification and release module is used to verify the RAG effect of the grayscale version knowledge base and complete the grayscale release based on the verification results.
[0014] Thirdly, the present invention also discloses an electronic device, the electronic device comprising: a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory to cause the electronic device to perform the grayscale release method of the large model knowledge base content as described in any one of claims 1 to 7.
[0015] Fourthly, the present invention also discloses a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed, implements the above-mentioned gray-scale release method for the content of the large model knowledge base.
[0016] As described above, the gray-scale release method, system, device, and medium for the large model knowledge base content described in this application have the following beneficial effects: When a canary release of a knowledge base is required, standard slice content is generated by obtaining the slice content to be released and setting a canary identifier field on the slice content. This facilitates adjustments to the slice content after the canary release, thereby enabling canary release of knowledge base document content through data visibility. It also facilitates canary release after verifying the release effect and allows for easy cancellation of canary release and version rollback, thus achieving effective management of knowledge base versions. Attached Figure Description
[0017] Figure 1 The flowchart shown is a method for grayscale release of large model knowledge base content as described in an embodiment of this application.
[0018] Figure 2 This diagram illustrates the execution process of the gray-scale release method for the large model knowledge base content described in this application embodiment.
[0019] Figure 3 The diagram shown represents the structural block diagram of the grayscale release system for the large model knowledge base content described in this application embodiment. Detailed Implementation
[0020] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.
[0021] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0022] See Figures 1 to 3 The following embodiments of this application provide a method, system, device, and medium for canary release of large model knowledge base content. When canary release of the knowledge base is required, the slice content to be canary released is obtained, and a canary identifier field is set on the slice content to form standard slice content. This facilitates the adjustment of the slice content after subsequent canary release, thereby realizing the canary release of knowledge base document content through data visibility. It also facilitates the canary release after the canary release is verified, and the canary release can be canceled to achieve version rollback, thereby achieving effective management of knowledge base versions.
[0023] The technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.
[0024] like Figure 1 As shown in the figure, this embodiment provides a method for canary release of content in a large model knowledge base, the method including the following steps: S101. When it is necessary to release the knowledge base in a gray-scale manner, create a gray-scale version of the knowledge base and set its status to not online.
[0025] refer to Figure 2 In this embodiment, when it is necessary to release the knowledge base in a gray-scale manner, a gray-scale version of the knowledge base is first created and set to be not online, so as to facilitate the subsequent release of the gray-scale version of the knowledge base.
[0026] S102. Obtain the slice content to be published in grayscale, set the grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in the vector database.
[0027] When performing a canary release, the slice content to be released is first obtained. After setting the corresponding canary identifier field for the slice content, a standard slice content is formed. The standard slice content is then stored in a vector database, which facilitates the subsequent use of the standard slice content to release canary versions of the knowledge base.
[0028] In some embodiments, the grayscale identifier field includes "released", "add" (to be added in grayscale), and "remove" (to be deleted in grayscale). By configuring the grayscale identifier field corresponding to the standard slice content and setting the above three enumeration values, it is convenient to set the status of the slice content to be released in grayscale release in advance according to the enumeration value of the grayscale identifier field. This facilitates subsequent changes to the slice content based on data visibility, and makes it convenient to implement version grayscale release of the knowledge base based on the grayscale identifier field of the slice content.
[0029] S103. Read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly.
[0030] In this embodiment, after establishing a grayscale version of the knowledge base, the grayscale version of the knowledge base is read, and the standard slice content is obtained from the vector database. The content in the knowledge base is then adjusted based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base. At the same time, the grayscale identifier field of the standard slice content is configured so that the grayscale version of the standard slice content can be rolled back according to the configured grayscale identifier field.
[0031] In some other embodiments, the step of reading the grayscale version of the knowledge base and the standard slice content, adjusting the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configuring the grayscale identifier field of the standard slice content accordingly includes: When the gray version of the knowledge base corresponds to newly added data, the standard slice content is added to the knowledge base to form the gray version knowledge base, and the gray identifier field of the standard slice content is set to add to be added in gray version. When the gray version of the knowledge base corresponds to deleted data, the standard slice content is deleted, and the gray-scale identifier field of the standard slice content is set to be removed in gray-scale deletion to form the gray-scale version knowledge base; When the grayscale version of the knowledge base corresponds to modified data, the initial slice data before modification is obtained from the knowledge base, the grayscale identifier field of the initial slice data is set to be removed during grayscale, the standard slice content is added to the knowledge base, and the grayscale identifier field of the standard slice content is set to be added during grayscale, so as to form the grayscale version knowledge base.
[0032] In this embodiment, when configuring based on a grayscale version, the operation of determining the slice data content in the knowledge base is first performed based on the grayscale version, so as to adjust and process the knowledge base according to the slice data content.
[0033] Specifically, when the gray version of the knowledge base corresponds to newly added data, the standard slice content is directly added to the knowledge base to form the gray version knowledge base. Simultaneously, the gray-scale identifier field of the standard slice content is set to "add" for later gray-scale addition, facilitating quick identification of the newly added slice content based on the gray-scale identifier field and enabling version rollback. When the gray version of the knowledge base corresponds to deleted data, the corresponding standard slice content is directly deleted from the knowledge base, and its gray-scale identifier field is set to "remove" for later gray-scale deletion, thus obtaining the gray version knowledge base. When the gray version of the knowledge base corresponds to modified data, the initial slice data before modification is first retrieved from the knowledge base, and the gray-scale identifier field of the initial slice data is set to "remove" for later gray-scale deletion. The standard slice content is then added to the knowledge base, and its gray-scale identifier field is set to "add" for later gray-scale addition, thus forming the final gray version knowledge base.
[0034] During the process of changing the knowledge base, the slice data of the knowledge base is not directly modified. Instead, the modification operation is changed to delete the slice data before the modification and add the slice data after the modification. This allows for the modification of slice data, so as to facilitate the rapid release and rollback of the gray version of the knowledge base.
[0035] S104. Verify the RAG effect of the grayscale version knowledge base, and complete the grayscale release based on the verification results.
[0036] In some other embodiments, verifying the RAG effect of the grayscale version knowledge base includes: When a regular user of the knowledge base verifies the RAG effect, the official RAG interface is called, and the slice content of the official version of the knowledge base is retrieved through the first search condition for effect verification. When the owner of the knowledge base verifies the RAG effect, the grayscale RAG interface is called, and the slice content of the grayscale version knowledge base is retrieved through the second search condition to verify the RAG effect. The first search condition is that the grayscale identifier field is either "released" or "to be removed in grayscale," and the second search condition is that the grayscale identifier field is either "released" or "to be added in grayscale."
[0037] In this embodiment, after the gray-scale release of the knowledge base, the RAG effect of the gray-scale version knowledge base is verified so as to determine whether the gray-scale release can be completed based on the verified RAG effect.
[0038] Specifically, since different users verify the RAG effect of the knowledge base in different ways, when ordinary users of the knowledge base verify the RAG effect, they directly call the official RAG interface to recall the slice content of the official version of the knowledge base, and then verify the RAG effect based on the recalled slice content; while when the knowledge base owner verifies the RAG effect, they call the grayscale RAG interface and use the second search condition to recall the slice content of the grayscale version of the knowledge base to verify the RAG effect.
[0039] It should be noted that the RAG effect verification based on the recalled slice content uses existing technology. This solution does not involve any improvement to the RAG effect verification method; the existing verification method is sufficient, and will not be elaborated here.
[0040] In some embodiments, completing the canary release based on the verification results includes: When the verification results meet the preset requirements, the gray release verification is determined to be successful. The current official version of the gray version knowledge base is set to the historical version, and the gray version of the knowledge base is set to the updated official version. Modify the grayscale identifier field of the standard slice content that is to be added in grayscale to "released", and delete the initial slice data that is to be removed in grayscale.
[0041] In this embodiment, when the verification effect meets the preset requirements, the gray-scale release verification is determined to be successful. Then, the current version of the gray-scale version knowledge base is set to the historical version, and the gray-scale version is set to the updated official version. At the same time, the gray-scale identifier field of the standard slice content to be added in the gray-scale is modified to "released", and the initial slice data to be removed in the gray-scale is deleted. This completes the change process of the entire standard slice content and realizes the official release of the gray-scale version.
[0042] In some other embodiments, the step of completing the canary release based on the verification results further includes: When the verification effect does not meet the preset requirements, it is determined that the gray release verification fails, the gray release is canceled, the standard slice content whose gray-scale identifier character is to be added in the gray release is deleted, and the initial slice data whose gray-scale identifier field is to be removed in the gray release is modified to be released, and the gray-scale version is deleted.
[0043] In this embodiment, if the verification effect does not meet the preset requirements, it is determined that the gray release verification has failed. In this case, the gray release is canceled, and the gray-scale identifier character is deleted to add the standard slice content to be added in the gray release. The initial slice data whose gray-scale identifier field is removed is modified to be released. At the same time, the gray-scale version is deleted, thereby rolling back the knowledge base from the gray-scale version to the previous official version. This will not affect the official version of the knowledge base, ensuring that the version can be quickly rolled back when the gray-scale version update of the knowledge base is not effective, without affecting the normal use by users.
[0044] In some other embodiments, if the verification effect does not meet the preset requirements, the verification is determined to have failed. The standard slice content is then readjusted based on the grayscale version, and a new grayscale release is conducted. If grayscale verification fails, but further grayscale release is still required, the standard slice content is readjusted, and the grayscale release is completed again. The process is essentially the same as described above and will not be repeated here.
[0045] The scope of protection for the gray-scale release method of large model knowledge base content described in this application is not limited to the execution order of the steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application.
[0046] This invention also discloses a canary release system for large model knowledge base content, referencing Figure 3 The system includes: The gray version creation module 301 is used to create a gray version of the knowledge base and set its status to not online when a gray release of the knowledge base is required. The content acquisition module 302 is used to acquire the slice content to be grayscale published, set a grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in a vector database. Version configuration module 303 is used to read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly. The verification and release module 304 is used to verify the RAG effect of the grayscale version knowledge base and complete the grayscale release based on the verification results.
[0047] The gray-scale release system for large model knowledge base content described in this application can implement the gray-scale release method for large model knowledge base content described in this application. However, the implementation device for the gray-scale release method for large model knowledge base content described in this application includes, but is not limited to, the structure of the gray-scale release system for large model knowledge base content listed in this embodiment. All structural modifications and substitutions of the prior art made based on the principles of this application are included within the protection scope of this application.
[0048] It should be noted that the structure and principle of each of the above modules correspond one-to-one with the steps in the gray-scale release method of the large model knowledge base content. The specific working principle can also be referred to the introduction of the gray-scale release method of the large model knowledge base content in the foregoing embodiments, so it will not be repeated here.
[0049] The present invention also provides an electronic device, the electronic device comprising: a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-described grayscale release method for the large model knowledge base content.
[0050] The present invention also discloses a computer-readable storage medium storing a computer program that, when executed, implements the above-described method for canary release of large model knowledge base content.
[0051] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. This available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state drive (SSD)).
[0052] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.
[0053] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of this application, depending on actual needs. For example, the functional modules / units in the various embodiments of this application may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.
[0054] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0055] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.
[0056] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.
Claims
1. A method for canary release of content in a large model knowledge base, characterized in that, The method includes: When a gray-scale release of a knowledge base is required, create a gray-scale version of the knowledge base and set its status to "not online". Obtain the slice content to be published in grayscale, set the grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in the vector database; Read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly; Verify the RAG effect of the grayscale version knowledge base, and complete the grayscale release based on the verification results.
2. The method for canary release of large model knowledge base content according to claim 1, characterized in that, The grayscale identifier field includes "released", "add" (to be added during grayscale testing), and "remove" (to be deleted during grayscale testing).
3. The method for canary release of large model knowledge base content according to claim 2, characterized in that, The process of reading the grayscale version of the knowledge base and the standard slice content, adjusting the knowledge base based on the grayscale version and the standard slice content to form a grayscale version knowledge base, and configuring the grayscale identifier field of the standard slice content accordingly includes: When the grayscale version of the knowledge base corresponds to newly added data, the standard slice content is added to the knowledge base to form the grayscale version knowledge base, and the grayscale identifier field of the standard slice content is set to add to be added in grayscale. When the gray version of the knowledge base corresponds to deleted data, the standard slice content is deleted, and the gray-scale identifier field of the standard slice content is set to be removed in gray-scale deletion to form the gray-scale version knowledge base; When the grayscale version of the knowledge base corresponds to modified data, the initial slice data before modification is obtained from the knowledge base, the grayscale identifier field of the initial slice data is set to be removed during grayscale, the standard slice content is added to the knowledge base, and the grayscale identifier field of the standard slice content is set to be added during grayscale, so as to form the grayscale version knowledge base.
4. The method for canary release of large model knowledge base content according to claim 3, characterized in that, The verification of the RAG effect of the grayscale version knowledge base includes: When a regular user of the knowledge base verifies the RAG effect, the official RAG interface is called, and the slice content of the official version of the knowledge base is retrieved through the first search condition for effect verification. When the owner of the knowledge base verifies the RAG effect, the grayscale RAG interface is invoked, and the slice content of the grayscale version of the knowledge base is retrieved through the second search condition to verify the RAG effect.
5. The method for canary release of large model knowledge base content according to claim 4, characterized in that, The process of completing the canary release based on the verification results includes: When the verification results meet the preset requirements, the gray release verification is determined to be successful. The current official version of the gray version knowledge base is set to the historical version, and the gray version of the knowledge base is set to the updated official version. Modify the grayscale identifier field of the standard slice content that is to be added in grayscale to "released", and delete the initial slice data that is to be removed in grayscale.
6. The method for canary release of large model knowledge base content according to claim 4, characterized in that, The process of completing the canary release based on the verification results also includes: When the verification effect does not meet the preset requirements, it is determined that the gray release verification fails, the gray release is canceled, the standard slice content whose gray-scale identifier character is to be added in the gray release is deleted, and the initial slice data whose gray-scale identifier field is to be removed in the gray release is modified to be released, and the gray-scale version is deleted.
7. The method for canary release of large model knowledge base content according to claim 6, characterized in that, If the verification result does not meet the preset requirements, the verification is determined to be unsuccessful. The standard slice content is then readjusted based on the grayscale version and re-released in grayscale.
8. A canary release system for content in a large model knowledge base, characterized in that, The system includes: The gray-scale version creation module is used to create gray-scale versions of the knowledge base and set their status to not online when a gray-scale release of the knowledge base is required. The content acquisition module is used to acquire the slice content to be released in grayscale, set the grayscale identifier field on the standard slice content to form standard slice content, and save the standard slice content in the vector database. The version configuration module is used to read the grayscale version of the knowledge base and the standard slice content, adjust the knowledge base based on the grayscale version of the knowledge base and the standard slice content to form a grayscale version knowledge base, and configure the grayscale identifier field of the standard slice content accordingly. The verification and release module is used to verify the RAG effect of the grayscale version knowledge base and complete the grayscale release based on the verification results.
9. An electronic device, characterized in that, The electronic device includes: a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory to cause the electronic device to perform the grayscale release method of the large model knowledge base content as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed, it implements the gray-scale release method for the content of the large model knowledge base as described in any one of claims 1 to 7.