An AI intelligent hardware matching APP customer service fault guidance type RAG interaction method, system, device and medium

Abstract

The application discloses an AI intelligent hardware matching APP customer service fault guiding type RAG interaction method, system, device and medium, mainly relates to the technical field of customer service fault guiding, and is used to solve the problems that the existing scheme lacks personal adaptability in fault guiding mode, the matching degree of RAG search result and user demand is low, and the dynamic mapping of guiding granularity and user behavior is not established. It comprises the following steps: obtaining a preset commonly used scene from a scene characteristic dimension, and obtaining a preset preference guiding word from an interaction habit dimension; determining the current guiding granularity type according to the current conversation behavior characteristic, the preset historical preference granularity proportion data and the preset preference guiding word; when the fault characteristic belongs to a preset high-frequency fault, directly extracting the guiding information corresponding to the current guiding granularity type from the current device; when the fault characteristic does not belong to the preset high-frequency fault, extracting the guiding information corresponding to the current guiding granularity type from the preset RAG cloud database.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of customer service fault guidance, in particular to an AI intelligent hardware matching APP customer service fault guidance type RAG interaction method, system, device and medium. BACKGROUND

[0002] With the segmentation of the AI intelligent hardware market, products (such as AI toys and smart desktop companion devices) that are oriented to users and focus on emotional companionship have gradually become consumer hotspots. Such products need to rely on matching APPs to realize device management, function setting and fault consultation. However, the customer service system of the current AI intelligent hardware matching APP has the following significant defects when serving female users:

[0003] Firstly, the fault guidance mode lacks personal adaptability. The existing customer service system adopts a guidance mode of "technical terms leading + long process text explanation", for example, directly prompting "check whether the device firmware version is v2.1 or above" and "confirm that the APP has obtained the Bluetooth permission (the location permission of the Android terminal needs to be turned on)". However, the understanding ability of some users for technical terms is weak, and complex text processes are easy to cause operation confusion, resulting in a decrease in the willingness to solve the fault.

[0004] Secondly, the matching degree of the RAG search result and the user demand is low. The conventional RAG system only pushes general solutions based on the fault keywords (such as "the device cannot be connected to the APP") input by the user, without considering the differences in the use scenarios of different users - for example, for the same "connection fault", an office user may operate using mobile phone traffic during commuting, and a home user may use home WiFi. The general solution cannot cover the scenario-based demand, and the user needs to repeatedly screen effective information.

[0005] Thirdly, the dynamic mapping of the guidance granularity and the user behavior is not established. The existing RAG system only searches for solutions based on the fault keywords, without combining the current interactive behavior (such as input speed, term use, whether to ask questions) or historical preferences of the user to dynamically judge the demand of the user for "conciseness" or "companionship" of the guidance, resulting in mismatch between the service supply and the user expectation. SUMMARY

[0006] The application provides an AI intelligent hardware matching APP customer service fault guidance type RAG interaction method, system, device and medium to solve the problems of the existing solutions that the fault guidance mode lacks personal adaptability, the matching degree of the RAG search result and the user demand is low, and the dynamic mapping of the guidance granularity and the user behavior is not established.

[0007] In a first aspect, the application provides an AI intelligent hardware matching APP customer service fault guidance type RAG interaction method, which comprises:

[0008] The system collects current user data through the backend of the AI ​​smart hardware's accompanying APP to build a structured preference feature library. The structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension.

[0009] Based on the number of each historical fault in the fault history dimension, determine the current user's preset high-frequency faults; obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance, and preset detailed accompanying guidance.

[0010] Obtain user fault consultation information, and extract current session behavior characteristics and fault characteristics from the user fault consultation information;

[0011] Based on a database of preset user history guidance granularity, determine the percentage of preset historical preference granularity data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension;

[0012] Based on the current conversation behavior characteristics, the percentage of preset historical preference granularity data, and the preset preference guidance words, determine the current guidance granularity type;

[0013] When the fault characteristics belong to the preset high-frequency faults, the boot information corresponding to the current boot granularity type is directly extracted from the current device; when the fault characteristics do not belong to the preset high-frequency faults, the boot information corresponding to the current boot granularity type is extracted from the preset RAG cloud database.

[0014] In one implementation of this application, the user's current usage data is collected through the background of the AI ​​smart hardware's accompanying APP to construct a structured preference feature library, specifically including:

[0015] The system collects current user usage data through the background of the AI ​​smart hardware's accompanying APP.

[0016] Extract interaction habit dimension, scenario feature dimension, and fault history dimension from the current user's usage data;

[0017] Among them, the interaction habit dimension includes at least: preset preference guide words, and the preset preference guide words are preset guide words that appear more frequently than a preset threshold in the usage data within a preset time period;

[0018] The scene feature dimension includes at least: preset common scenes, and the preset common scenes are the preset scenes that appear most frequently in the data within a preset time period;

[0019] The fault history dimension should include at least: all historical faults that occurred within a preset time period.

[0020] In one implementation of this application, the preset high-frequency faults for the current user are determined based on the number of each historical fault in the fault history dimension, specifically including:

[0021] Based on the number of each historical fault in the fault history dimension, historical faults whose number exceeds a preset threshold are identified as preset high-frequency faults for the current user.

[0022] In one implementation of this application, before obtaining boot information of several boot granularities for preset high-frequency faults from a preset RAG cloud database, the method further includes:

[0023] Obtain boot information for each fault at several boot granularities;

[0024] And upload it to the preset RAG cloud database.

[0025] In one implementation of this application, user fault inquiry information is obtained, and current session behavior features and fault features are extracted from the user fault inquiry information, specifically including:

[0026] Extract text length, technical terminology density, operational intent data, sentiment data, and fault characteristics from user fault consultation information;

[0027] Among them, the current conversation behavior characteristics include at least text length, technical terminology density, operational intent data, and sentiment data, and the sentiment data is divided into: containing preset urgency words and not containing preset urgency words.

[0028] In one implementation of this application, the preset historical preference granularity percentage data is determined based on a preset user historical guidance granularity selection database, specifically including:

[0029] The proportion of each type of guidance granularity is determined from the preset user history guidance granularity selection database;

[0030] The guiding granularity and the corresponding proportion are used as preset historical preference granularity proportion data.

[0031] In one implementation of this application, the current guidance granularity type is determined based on current conversation behavior characteristics, preset historical preference granularity percentage data, and preset preference guidance words, specifically including:

[0032] Retrieve user-uploaded preset length-assignment data table, preset density-assignment data table, preset tendency-assignment data table, and preset guidance-assignment data table;

[0033] Based on the preset length-assignment data table, determine the specific numerical value of the text length in each conversation behavior feature and assign it to the preset length value in each type of guidance granularity;

[0034] Based on the preset density-assignment data table, determine the specific values ​​of the technical term density in each conversation behavior feature and assign them to the preset density values ​​in each type of guidance granularity.

[0035] Based on the preset tendency-assignment data table, determine the preset tendency assignment of the emotional tendency data in each conversation behavior feature at each type of guidance granularity;

[0036] Based on the preset guidance-assignment data table, determine the preset guidance assignment for each preset preference guidance word at each type of guidance granularity;

[0037] The percentage of each guidance granularity in the preset historical preference granularity percentage data is used as the preset preference percentage for each type of guidance granularity.

[0038] Accumulate the preset length assignment, preset density assignment, preset tendency assignment, preset guidance assignment, and preset preference ratio for each type; determine the guidance granularity type corresponding to the maximum value as the current guidance granularity type.

[0039] Secondly, this application provides an AI-powered smart hardware-supported APP customer service fault-guided RAG interactive system, the system comprising:

[0040] The module is used to collect current user usage data through the background of the AI ​​smart hardware's accompanying APP and build a structured preference feature library; the structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension;

[0041] The acquisition module is used to determine the current user's preset high-frequency faults based on the number of each historical fault in the fault history dimension; and to obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance, and preset detailed accompanying guidance.

[0042] The extraction module is used to obtain user fault consultation information, extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity to determine the preset historical preference granularity proportion data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension; determine the current guidance granularity type based on the current session behavior features, the preset historical preference granularity proportion data, and the preset preference guidance words.

[0043] The information module is used to directly extract the guidance information corresponding to the current guidance granularity type from the current device when the fault characteristics belong to the preset high-frequency faults; and to extract the guidance information corresponding to the current guidance granularity type from the preset RAG cloud database when the fault characteristics do not belong to the preset high-frequency faults.

[0044] Thirdly, this application provides an AI-powered smart hardware-supported APP customer service fault-guided RAG interactive device, the device comprising:

[0045] processor;

[0046] And a memory that stores executable code, which, when executed, causes the processor to execute an AI smart hardware-supporting APP customer service fault-guided RAG interaction method as described above.

[0047] Fourthly, this application provides a non-volatile computer storage medium storing computer instructions, which, when executed, implement an AI smart hardware-supporting APP customer service fault-guided RAG interaction method as described above.

[0048] As can be seen from the above technical solutions, this application has the following advantages:

[0049] I. Improve the personal adaptability of fault booting:

[0050] By constructing a structured preference feature library covering interaction habits, scenario characteristics, and fault history, the system can accurately capture users' personalized needs. For example, for users with weak understanding of technical terminology, the system can automatically identify their preference for brief guidance in their historical interactions and prioritize pushing "preset brief guidance" information to avoid operational confusion caused by complex textual processes. At the same time, by combining the user's current conversation behavior characteristics (such as input speed and frequency of terminology use), the system dynamically adjusts the level of detail in the guidance content to ensure that information delivery matches the user's cognitive ability, significantly reducing the problem of decreased user willingness to solve faults due to comprehension barriers.

[0051] II. Optimize the contextual matching of RAG search results:

[0052] By pre-setting multi-granularity guidance information (brief, general, and detailed support) in the RAG cloud database and combining it with user scenario characteristics (such as using mobile data during commutes or using home WiFi), the system can push differentiated solutions for the same fault (such as "connection failure"). For example, working users may receive a brief prompt to "quickly check network signal," while home users will receive detailed steps to "restart the router and check WiFi settings." This scenario-based adaptation reduces the burden on users to repeatedly sift through information and improves the efficiency and accuracy of troubleshooting.

[0053] III. Achieving dynamic mapping between guidance granularity and user behavior:

[0054] By using a database of preset historical guidance granularity and current session behavior characteristics (such as whether follow-up questions are asked and the frequency of terminology use), the system determines in real time whether the user's guidance needs to be "simple" or "companion-like." For example, when a user quickly enters keywords related to a problem without asking follow-up questions, the system automatically pushes a "preset brief guidance"; if the user frequently uses technical terms or actively asks follow-up questions, it switches to a "preset detailed companion guidance," providing step-by-step instructions. This dynamic mapping mechanism ensures a precise match between service provision and user expectations, avoiding the user experience mismatch problem caused by fixed guidance granularity in existing solutions. Attached Figure Description

[0055] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a fault-guided RAG interaction method for AI smart hardware and its supporting APP customer service, provided in an embodiment of this application.

[0057] Figure 2 This is a schematic diagram of the internal structure of an AI smart hardware-supporting APP customer service fault-guided RAG interactive system provided in an embodiment of this application.

[0058] Figure 3 This is a schematic diagram of the internal structure of an AI smart hardware-supporting APP customer service fault guidance RAG interactive device provided in this application embodiment. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.

[0061] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0062] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0063] This embodiment provides an AI-powered smart hardware-supporting APP customer service fault-guided RAG interaction method, such as... Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps:

[0064] Step 110: Collect current user usage data through the AI ​​smart hardware's accompanying APP backend to build a structured preference feature library.

[0065] The structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension.

[0066] In some embodiments, the user's current usage data is collected through the background of the AI ​​smart hardware's accompanying APP to construct a structured preference feature library, specifically including:

[0067] The system collects current user usage data through the background of the AI ​​smart hardware's accompanying APP.

[0068] Extract interaction habit dimension, scenario feature dimension, and fault history dimension from the current user's usage data;

[0069] Among them, the interaction habit dimension includes at least: preset preference guide words, and the preset preference guide words are preset guide words that appear more frequently than a preset threshold in the usage data within a preset time period;

[0070] The scene feature dimension includes at least: preset common scenes, and the preset common scenes are the preset scenes that appear most frequently in the data within a preset time period;

[0071] The fault history dimension should include at least: all historical faults that occurred within a preset time period.

[0072] Understandably, this step involves collecting user data through the AI-powered smart hardware's accompanying app to construct a structured preference feature library encompassing three dimensions: interaction habits, scenario characteristics, and fault history. In the interaction habits dimension, the system identifies preset guiding words that occur more frequently than a threshold within a preset time period, directly reflecting the user's high-frequency operation patterns. The scenario characteristics dimension captures the most frequently occurring preset scenarios within a preset time period, clarifying the user's typical usage environment. The fault history dimension records all historical faults within a preset time period, comprehensively preserving device anomaly information. Thus, the feature library accurately characterizes user behavior patterns, providing data support for subsequent personalized services, while also facilitating fault retrospective analysis and improving problem localization efficiency.

[0073] Step 120: Determine the current user's preset high-frequency faults based on the number of each historical fault in the fault history dimension; obtain several guidance granularities of guidance information for the preset high-frequency faults from the preset RAG cloud database.

[0074] The preset RAG cloud database contains several levels of boot information for each fault; the boot granularity types are: preset brief boot, preset regular boot, and preset detailed accompanying boot.

[0075] To further explain, the default brief guide (efficient direct access) requires only key steps or self-service links;

[0076] Pre-set standard guidance: requires step-by-step operation but does not require emotional language; uses non-technical language.

[0077] Pre-set detailed companionship guidance (delicate companionship): includes emotional encouragement, progress prompts, and graphic annotations.

[0078] For example, the "Device firmware version is too low" (preset brief boot) can be associated with the "Device hasn't been updated in a long time" label, and related to the scenario (such as "first boot after a long period of inactivity").

[0079] The "APP permissions not enabled" is matched with the "APP has not obtained permission to connect to the device" label, further subdivided into scenarios (such as "first use of a newly installed APP" and "permissions reset after a phone system update").

[0080] Each scenario tag is associated with "preset detailed companion guidance" (such as "Hi, let's first check if the APP has permission to connect to the device~") and "scenario-based solution summary" (such as "In a home WiFi environment, first check if the router is on the 2.4G band").

[0081] Based on the number of each historical fault in the fault history dimension, the preset high-frequency faults for the current user are determined, specifically including:

[0082] Based on the number of each historical fault in the fault history dimension, historical faults whose number exceeds a preset threshold are identified as preset high-frequency faults for the current user.

[0083] Understandably, the system identifies high-frequency user faults through fault history and provides multi-granular guidance information based on a pre-set RAG cloud database. Pre-set brief guidance uses key steps or self-service links for quick responses, suitable for scenarios requiring efficient resolution; pre-set regular guidance uses step-by-step operations and non-technical language, facilitating user workflow; and pre-set detailed support guidance incorporates emotional encouragement, progress indicators, and graphic annotations to enhance the user experience in complex scenarios. For example, for high-frequency faults such as "device firmware version too low" or "APP permissions not enabled," the system can associate specific usage scenarios (such as prolonged device inactivity or after a system update) to provide tagged guidance and scenario-based solution summaries, helping users more accurately pinpoint the problem. This tiered guidance mechanism directly adapts to different fault types and user needs, reducing operational confusion while improving problem-solving efficiency through structured information display.

[0084] Before retrieving boot information at several boot granularities for preset high-frequency faults from the preset RAG cloud database, the method further includes:

[0085] Obtain boot information for each fault at several boot granularities;

[0086] And upload it to the preset RAG cloud database.

[0087] Understandably, by pre-collecting and uploading multi-granularity guidance information to a pre-defined RAG cloud database, structured support is provided for handling high-frequency faults. Specifically, the system pre-sets three types of information for each fault: brief guidance, regular guidance, and detailed support guidance. Brief guidance focuses on key steps or self-service links for quick response; regular guidance provides step-by-step instructions using non-technical language to lower the barrier to understanding; and detailed support guidance incorporates emotional encouragement, progress prompts, and graphic annotations to improve the user experience in complex scenarios. For example, for faults such as "device firmware version too low" or "APP permissions not enabled," the guidance information can be associated with specific usage scenarios (such as prolonged device inactivity or after a system update) and matched with tagged solutions. This tiered guidance mechanism can directly adapt to different fault types and user needs, reduce operational confusion, and improve problem-solving efficiency through structured information display.

[0088] Step 130: Obtain user fault consultation information, extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity, and determine the preset historical preference granularity percentage data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension; determine the current guidance granularity type based on the current session behavior features, the preset historical preference granularity percentage data, and the preset preference guidance words.

[0089] Understandably, determining the granularity of fault guidance through multi-dimensional data fusion improves the accuracy of user consultation responses. Specifically, the system extracts current session behavior characteristics (such as operation frequency and dwell time) and fault characteristics (such as error codes and phenomenon descriptions) from user fault consultation information. It then combines this with historical preference data from the database (such as the user's past selection ratio for brief / standard / detailed guidance), and obtains preset common scenarios (such as "nighttime use" and "multi-user sharing") from the scenario feature dimension, and extracts preset preference guidance words (such as "quick solution" and "step-by-step guidance") from the interaction habit dimension. Based on this data, the system dynamically matches the required guidance granularity for the current session: if the user's historical preference for brief guidance is high and the current fault characteristics are simple, it prioritizes providing efficient and direct key steps; if the scenario is complex or the fault characteristics are ambiguous, it combines step-by-step explanations of standard guidance with emotional support of detailed guidance. This mechanism directly adapts to users' personalized needs, reduces information redundancy, and improves problem-solving efficiency through structured feature matching.

[0090] Obtain user fault inquiry information, and extract current session behavior features and fault features from the user fault inquiry information, specifically including:

[0091] Extract text length, technical terminology density, operational intent data, sentiment data, and fault characteristics from user fault consultation information;

[0092] Among them, the current conversation behavior characteristics include at least text length, technical terminology density, operational intent data, and sentiment data, and the sentiment data is divided into: containing preset urgency words and not containing preset urgency words.

[0093] Understandably, multi-dimensional feature extraction improves the accuracy of fault consultation responses. Specifically, the system extracts text length, technical terminology density, operational intent data, and sentiment data (including and excluding urgency words) from user fault consultation information, while simultaneously combining this with fault characteristics (such as error codes and symptom descriptions) for comprehensive analysis. For example, if the text is short and contains urgency words, the system can determine that the user needs a quick solution and prioritize providing concise guidance on key steps; if the technical terminology density is high and the operational intent is clear, it matches step-by-step instructions with standard guidance; if the sentiment indicates user confusion or anxiety, it activates detailed guidance with emotional support. This mechanism directly adapts to the user's current state, reduces information redundancy, and improves problem-solving efficiency through structured feature matching, ensuring that the guidance content highly matches the user's needs.

[0094] Based on a pre-defined database of user historical guidance granularity selections, the percentage of pre-defined historical preference granularity data is determined, specifically including:

[0095] The proportion of each type of guidance granularity is determined from the preset user history guidance granularity selection database;

[0096] The guiding granularity and the corresponding proportion are used as preset historical preference granularity proportion data.

[0097] Understandably, by pre-setting a database of users' historical guidance granularity selections, the system systematically quantifies users' historical preferences for guidance granularity, providing a precise basis for subsequent fault response decisions. Specifically, the system extracts the percentage data of each type of guidance granularity (such as brief guidance, regular guidance, and detailed guidance) from the database and uses this percentage as the pre-set historical preference granularity percentage data. For example, if the percentage of brief guidance in a user's historical selections reaches 70%, the system can infer that they prefer efficient and direct solutions, prioritizing key steps or self-service links in subsequent fault consultations; if the percentage of detailed guidance increases significantly, it indicates that the user needs more nuanced, supportive guidance, and the system can automatically match elements such as emotional encouragement and progress prompts.

[0098] Based on current conversation behavior characteristics, preset historical preference granularity percentage data, and preset preference guidance words, the current guidance granularity type is determined, specifically including:

[0099] Retrieve user-uploaded preset length-assignment data table, preset density-assignment data table, preset tendency-assignment data table, and preset guidance-assignment data table;

[0100] Based on the preset length-assignment data table, determine the specific numerical value of the text length in each conversation behavior feature and assign it to the preset length value in each type of guidance granularity;

[0101] Based on the preset density-assignment data table, determine the specific values ​​of the technical term density in each conversation behavior feature and assign them to the preset density values ​​in each type of guidance granularity.

[0102] Based on the preset tendency-assignment data table, determine the preset tendency assignment of the emotional tendency data in each conversation behavior feature at each type of guidance granularity;

[0103] Based on the preset guidance-assignment data table, determine the preset guidance assignment for each preset preference guidance word at each type of guidance granularity;

[0104] The percentage of each guidance granularity in the preset historical preference granularity percentage data is used as the preset preference percentage for each type of guidance granularity.

[0105] Accumulate the preset length assignment, preset density assignment, preset tendency assignment, preset guidance assignment, and preset preference ratio for each type; determine the guidance granularity type corresponding to the maximum value as the current guidance granularity type.

[0106] Understandably, the system achieves dynamic adaptation of fault guidance granularity through multi-dimensional data fusion and weighted calculation. The system first obtains user-uploaded preset length-assignment data tables, preset density-assignment data tables, preset tendency-assignment data tables, and preset guidance-assignment data tables, and quantifies and assigns values ​​to text length, technical terminology density, sentiment tendency data, and preset preference guidance words, respectively. For example, longer text with higher technical terminology density may match the preset length and preset density values ​​for detailed guidance; when the sentiment tendency contains urgency-related words, the tendency assignment leans towards concise guidance; preset preference guidance words such as "quick solution" correspond to the preset guidance assignment for concise guidance. Simultaneously, the system uses the preset historical preference granularity ratio data (such as the proportion of users' past selections of concise guidance) as the preset preference ratio. Finally, by accumulating the preset length assignment, preset density assignment, preset tendency assignment, preset guidance assignment, and preset preference ratio of each type, the system determines the guidance granularity type corresponding to the maximum value.

[0107] Step 140: When the fault characteristic belongs to the preset high-frequency fault, directly extract the boot information corresponding to the current boot granularity type from the current device; when the fault characteristic does not belong to the preset high-frequency fault, extract the boot information corresponding to the current boot granularity type from the preset RAG cloud database.

[0108] Understandably, classifying fault characteristics enables differentiated acquisition of guidance information, improving response efficiency and accuracy. When a fault characteristic belongs to a preset high-frequency fault, the system directly extracts information at the corresponding guidance granularity from the current device, avoiding cloud retrieval delays and ensuring rapid resolution of high-frequency issues. When a fault characteristic does not belong to a preset high-frequency fault, the system extracts information at the corresponding guidance granularity from a preset RAG cloud database, utilizing rich cloud resources to cover long-tail issues. For example, for high-frequency faults such as "device firmware version too low," the brief guidance stored locally on the device can immediately provide key steps; while for low-frequency faults such as "APP permissions not enabled," the detailed companion guidance in the cloud database can provide emotional encouragement and graphic annotations.

[0109] As described above, this embodiment, by constructing a structured preference feature library covering interaction habits, scenario characteristics, and fault history, can accurately capture users' personalized needs. For example, for users with weak understanding of technical terminology, the system can automatically identify their preference for brief guidance in their historical interactions and prioritize pushing "preset brief guidance" information to avoid operational confusion caused by complex textual processes. At the same time, combined with the user's current conversational behavior characteristics (such as input speed and frequency of terminology use), the system dynamically adjusts the level of detail in the guidance content to ensure that information delivery matches the user's cognitive ability, significantly reducing the problem of decreased user willingness to solve faults due to comprehension barriers.

[0110] By pre-setting multi-granularity guidance information (brief, general, and detailed support) in the RAG cloud database and combining it with user scenario characteristics (such as using mobile data during commutes or using home WiFi), the system can push differentiated solutions for the same fault (such as "connection failure"). For example, working users may receive a brief prompt to "quickly check network signal," while home users will receive detailed steps to "restart the router and check WiFi settings." This scenario-based adaptation reduces the burden on users to repeatedly sift through information and improves the efficiency and accuracy of troubleshooting.

[0111] By using a database of preset historical guidance granularity and current session behavior characteristics (such as whether follow-up questions are asked and the frequency of terminology use), the system determines in real time whether the user's guidance needs to be "concise" or "companion-like." For example, when a user quickly enters keywords related to a problem without asking follow-up questions, the system automatically pushes a "preset brief guidance"; if the user frequently uses technical terms or actively asks follow-up questions, it switches to a "preset detailed companion guidance," providing step-by-step instructions. This dynamic mapping mechanism ensures a precise match between service provision and user expectations, avoiding the user experience mismatch problem caused by fixed guidance granularity in existing solutions. In addition, this application... Figure 2 This application provides an AI-powered smart hardware-supporting APP customer service fault-guided RAG interactive system. For example... Figure 2As shown in the embodiments of this application, the system mainly includes:

[0112] Module 210 is used to collect current user usage data through the AI ​​smart hardware's accompanying APP backend and build a structured preference feature library; the structured preference feature library includes: interaction habit dimension, scene feature dimension and fault history dimension.

[0113] The acquisition module 220 is used to determine the current user's preset high-frequency faults based on the number of each historical fault in the fault history dimension; and to obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance and preset detailed accompanying guidance.

[0114] The extraction module 230 is used to obtain user fault consultation information, extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity, and determine the preset historical preference granularity proportion data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension; and determine the current guidance granularity type based on the current session behavior features, the preset historical preference granularity proportion data, and the preset preference guidance words.

[0115] Information module 240 is used to directly extract the guidance information corresponding to the current guidance granularity type from the current device when the fault feature belongs to the preset high-frequency fault; and to extract the guidance information corresponding to the current guidance granularity type from the preset RAG cloud database when the fault feature does not belong to the preset high-frequency fault.

[0116] The above are method embodiments of this application. Based on the same inventive concept, this application also provides an AI smart hardware-supporting APP customer service fault guidance RAG interactive device. Figure 3 As shown, the device includes: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to execute an AI smart hardware-supporting APP customer service fault-guided RAG interaction method as described in the above embodiment.

[0117] Specifically, the server collects current user data through the AI ​​smart hardware's accompanying APP backend to build a structured preference feature library; the structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension;

[0118] Based on the number of each historical fault in the fault history dimension, determine the current user's preset high-frequency faults; obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance, and preset detailed accompanying guidance.

[0119] Obtain user fault consultation information, and extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity to determine the preset historical preference granularity proportion data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension;

[0120] Based on the current session behavior characteristics, the percentage of preset historical preference granularity data, and the preset preference guidance words, the current guidance granularity type is determined. When the fault characteristics belong to the preset high-frequency faults, the guidance information corresponding to the current guidance granularity type is directly extracted from the current device. When the fault characteristics do not belong to the preset high-frequency faults, the guidance information corresponding to the current guidance granularity type is extracted from the preset RAG cloud database.

[0121] In addition, this application embodiment also provides a non-volatile computer storage medium storing executable instructions, which, when executed, implement the AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method as described above.

[0122] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault-guided RAG interaction method for AI smart hardware paired with an APP customer service system, characterized in that, The method includes: The system collects current user data through the backend of the AI ​​smart hardware's accompanying APP to build a structured preference feature library. The structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension. Based on the number of each historical fault in the fault history dimension, determine the current user's preset high-frequency faults; obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance, and preset detailed accompanying guidance. Obtain user fault consultation information, and extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity to determine the preset historical preference granularity proportion data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension; Based on the current session behavior characteristics, the percentage of preset historical preference granularity data, and the preset preference guidance words, the current guidance granularity type is determined. When the fault characteristics belong to the preset high-frequency faults, the guidance information corresponding to the current guidance granularity type is directly extracted from the current device. When the fault characteristics do not belong to the preset high-frequency faults, the guidance information corresponding to the current guidance granularity type is extracted from the preset RAG cloud database.

2. The AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 1, characterized in that, By collecting current user usage data through the backend of the AI ​​smart hardware's accompanying APP, a structured preference feature library is constructed, specifically including: The system collects current user usage data through the background of the AI ​​smart hardware's accompanying APP. Extract interaction habit dimension, scenario feature dimension, and fault history dimension from the current user's usage data; Among them, the interaction habit dimension includes at least: preset preference guide words, and the preset preference guide words are preset guide words that appear more frequently than a preset threshold in the usage data within a preset time period; The scene feature dimension includes at least: preset common scenes, and the preset common scenes are the preset scenes that appear most frequently in the data within a preset time period; The fault history dimension should include at least: all historical faults that occurred within a preset time period.

3. The AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 1, characterized in that, Based on the number of each historical fault in the fault history dimension, the preset high-frequency faults for the current user are determined, specifically including: Based on the number of each historical fault in the fault history dimension, historical faults whose number exceeds a preset threshold are identified as preset high-frequency faults for the current user.

4. The AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 1, characterized in that, Before obtaining boot information at several boot granularities for preset high-frequency faults from the preset RAG cloud database, the method further includes: Obtain boot information for each fault at several boot granularities; And upload it to the preset RAG cloud database.

5. The AI-powered smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 1, characterized in that, Obtain user fault inquiry information, and extract current session behavior features and fault features from the user fault inquiry information, specifically including: Extract text length, technical terminology density, operational intent data, sentiment data, and fault characteristics from user fault consultation information; Among them, the current conversation behavior characteristics include at least text length, technical terminology density, operational intent data, and sentiment data, and the sentiment data is divided into: containing preset urgency words and not containing preset urgency words.

6. The AI-powered smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 1, characterized in that, Based on a pre-defined database of user historical guidance granularity selections, the percentage of pre-defined historical preference granularity data is determined, specifically including: The proportion of each type of guidance granularity is determined from the preset user history guidance granularity selection database; The guiding granularity and the corresponding proportion are used as preset historical preference granularity proportion data.

7. The AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method according to claim 6, characterized in that, Based on current conversation behavior characteristics, preset historical preference granularity percentage data, and preset preference guidance words, the current guidance granularity type is determined, specifically including: Retrieve user-uploaded preset length-assignment data table, preset density-assignment data table, preset tendency-assignment data table, and preset guidance-assignment data table; Based on the preset length-assignment data table, determine the specific numerical value of the text length in each conversation behavior feature and assign it to the preset length value in each type of guidance granularity; Based on the preset density-assignment data table, determine the specific values ​​of the technical term density in each conversation behavior feature and assign them to the preset density values ​​in each type of guidance granularity. Based on the preset tendency-assignment data table, determine the preset tendency assignment of the emotional tendency data in each conversation behavior feature at each type of guidance granularity; Based on the preset guidance-assignment data table, determine the preset guidance assignment for each preset preference guidance word at each type of guidance granularity; The percentage of each guidance granularity in the preset historical preference granularity percentage data is used as the preset preference percentage for each type of guidance granularity. Accumulate the preset length assignment, preset density assignment, preset tendency assignment, preset guidance assignment, and preset preference ratio for each type; determine the guidance granularity type corresponding to the maximum value as the current guidance granularity type.

8. A fault-guided RAG interactive system for AI-powered smart hardware and its accompanying APP customer service, characterized in that: The system includes: The module is used to collect current user usage data through the background of the AI ​​smart hardware's accompanying APP and build a structured preference feature library; the structured preference feature library includes: interaction habit dimension, scene feature dimension, and fault history dimension; The acquisition module is used to determine the current user's preset high-frequency faults based on the number of each historical fault in the fault history dimension; and to obtain guidance information of several guidance granularities for the preset high-frequency faults from the preset RAG cloud database; wherein, the preset RAG cloud database contains guidance information of several guidance granularities corresponding to each fault; the guidance granularity types are divided into: preset brief guidance, preset regular guidance, and preset detailed accompanying guidance. The extraction module is used to obtain user fault consultation information, extract current session behavior features and fault features from the user fault consultation information; select a database based on the preset user historical guidance granularity to determine the preset historical preference granularity proportion data; obtain preset common scenarios from the scenario feature dimension and preset preference guidance words from the interaction habit dimension; determine the current guidance granularity type based on the current session behavior features, the preset historical preference granularity proportion data, and the preset preference guidance words. The information module is used to directly extract the guidance information corresponding to the current guidance granularity type from the current device when the fault characteristics belong to the preset high-frequency faults; and to extract the guidance information corresponding to the current guidance granularity type from the preset RAG cloud database when the fault characteristics do not belong to the preset high-frequency faults.

9. A fault-guided RAG interactive device for AI-powered smart hardware and its accompanying APP customer service, characterized in that, The device includes: processor; And a memory storing executable code, which, when executed, causes the processor to execute the AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method as described in any one of claims 1-7.

10. A non-volatile computer storage medium, characterized in that, It stores computer instructions, which, when executed, implement the AI ​​smart hardware-supporting APP customer service fault-guided RAG interaction method as described in any one of claims 1-7.

Images