Intelligent agent personal private assistant method and device and integrated electronic equipment

Abstract

The invention provides an agent personal private assistant method and device and integrated electronic equipment, and the method comprises the steps: a localized deployment process of a large model: deploying the large model on an all-in-one machine, and carrying out the physical isolation; in the process of creating the personal private assistant of the intelligent agent, user exclusive data is collected to serve as a data source of an RAG of a large model, and collection means that in the process of interaction between the all-in-one machine and the user, at least one of life information, historical memory information and special sensitive information of the user is recorded in real time and is permanently memorized; the calling response process of the personal private assistant of the intelligent agent comprises a daily calling process and a special calling process; the daily calling process is that the user calls life information, historical memory information or special sensitive information through interaction with the all-in-one machine. According to the invention, a mode of'all-in-one machine localization deployment of a large model + physical isolation 'is adopted, and a permanent memory function of a computer system is utilized to solve the pain point problem of memory decline of old people.

Description

Technical Field

[0001] This invention relates to the field of intelligent agents, and in particular to a method, apparatus, and integrated electronic device for an intelligent agent personal privacy assistant. Background Technology

[0002] An intelligent agent is an artificial intelligence system developed based on a large language model (LLM) that is capable of autonomously performing tasks. It refers to an entity that can autonomously perform tasks or make decisions in a specific environment. Intelligent agents can take the form of software, such as chatbots, recommendation systems, and characters in games; or they can take the form of hardware, such as self-driving cars and service robots.

[0003] Currently, intelligent agents are used as personal work or life assistants, involving only their most basic functions and generally only related to information processing workflows, without involving the operation of hardware tools. However, their applications are also very wide-ranging, including: ① Automated office work: such as DeepSeekAgent and Microsoft Copilot, which can help automatically organize emails, generate reports, schedule appointments, etc., improving work efficiency. ② Programming assistants: such as AutoGPT and DeepSeek-Coder, which can help automatically complete code writing, debugging, and optimization. ③ Financial investment: intelligent agents can act as investment advisors, providing investment strategies and monitoring market risks.

[0004] From a technical perspective, current intelligent personal assistants can be summarized as: a large underlying language model + Retrieval-Augmented Generation (RAG) technology + flexible dialogue interaction methods.

[0005] From an engineering perspective, current intelligent personal assistant products, such as the simplest smart speakers or more advanced ones like Apple's Siri and Amazon's Alexa, connect to cloud servers (essentially "large language models") via the internet and interact with users through voice. Besides basic functions like information retrieval, timed reminders, and music playback, they can proactively communicate with users, provide personalized suggestions based on their daily habits and current context, and even automatically complete complex online tasks such as scheduling and online shopping.

[0006] However, current intelligent agents have the following drawbacks:

[0007] 1. Existing intelligent personal assistants rely on internet access to large-scale cloud models, posing significant data security risks. This is particularly concerning for sensitive personal data such as bank card passwords, raising concerns about user safety.

[0008] 2. The existing data resources of intelligent personal assistants are mainly general "public domain" knowledge, with insufficient personalized "private domain" knowledge, resulting in insufficient fit with the users they serve and weak personal exclusivity.

[0009] 3. Existing intelligent personal assistants lack continuous collection of users' private data, making it difficult to form long-term memories; in daily use, they also lack adaptive and convenient operation designs for key customer groups such as the elderly. Summary of the Invention

[0010] The technical problem to be solved by this invention is to provide a method, device and integrated electronic device for a smart personal privacy assistant. It adopts the mode of "local deployment of a large model of the all-in-one machine + physical isolation (no network connection)", takes the protection of user privacy data as the highest priority, uses the permanent memory function of the computer system to solve the pain point of memory decline in the elderly, and forms a user's personal local knowledge base.

[0011] In a first aspect, the present invention provides a method for a smart agent's personal privacy assistant, comprising:

[0012] The local deployment process of large models: Deploy the large model on the all-in-one machine, complete the initial setup, and physically isolate the all-in-one machine from the Internet;

[0013] The process of creating a personal intelligent assistant: After the all-in-one device is put into use, user-specific data is collected as the data source for the RAG of the large model. This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanently storing this information. During continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information, forming the user's personal local knowledge base. This knowledge base is then used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on the user's provided historical information and can be recalled at any time. The special sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords.

[0014] The call response process of the intelligent personal privacy assistant includes a daily call process and a special call process. The daily call process is when the user interacts with the all-in-one machine to call up life information, historical memory information or special sensitive information. The special call process is when the user does not interact with the all-in-one machine within a set time period, it is inferred that the user has lost self-awareness, and the emergency response plan is executed.

[0015] Furthermore, the interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal; the daily access process also verifies the user's identity before allowing access to life information, historical memory information, or special sensitive information.

[0016] Furthermore, when the special sensitive information is invoked, a process for recognizing people and the surrounding environment is added; when an unsafe environment is detected, the user is prompted to change to a safe environment; when a safe environment is detected, the special sensitive information is output.

[0017] Furthermore, based on the continuous accumulation of personal data in the local knowledge base, the large model is continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

[0018] Secondly, the present invention provides an intelligent agent personal privacy assistant device, comprising:

[0019] Local deployment module for large models: used to deploy large models on the all-in-one machine, complete the initial setup, physically isolate the all-in-one machine, and not connect it to the Internet;

[0020] The module for creating a personal intelligent assistant: After the all-in-one device is put into use, it collects user-specific data as the data source for the RAG of the large model. This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanently storing this information. During continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information, forming the user's personal local knowledge base. This knowledge base is then used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on the user's provided historical information and can be recalled at any time. The special sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords.

[0021] The intelligent personal privacy assistant's call response module includes a daily call process and a special call process. The daily call process is when the user interacts with the all-in-one machine to call up life information, historical memory information, or special sensitive information. The special call process is when the user does not interact with the all-in-one machine within a set time period, it is inferred that the user has lost consciousness, and the emergency response plan is invoked for execution.

[0022] Furthermore, the interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal; the daily access process also verifies the user's identity before allowing access to life information, historical memory information, or special sensitive information.

[0023] Furthermore, when the special sensitive information is invoked, a process for recognizing people and the surrounding environment is added; when an unsafe environment is detected, the user is prompted to change to a safe environment; when a safe environment is detected, the special sensitive information is output.

[0024] Furthermore, based on the continuous accumulation of personal data in the local knowledge base, the large model is continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

[0025] Thirdly, the present invention provides an integrated electronic device with a large model deployed thereon, which is physically isolated after initial setup and can perform the following processes:

[0026] The process of creating a personal intelligent assistant: After deployment, user-specific data is collected as the data source for the large model's RAG (Real-Time Analysis). This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and sensitive information during interaction, and permanent memory storage. During continuous interaction, the user's personal data is continuously accumulated and automatically categorized. Historical memory information is then cross-validated to form the user's personal local knowledge base, which is used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on user-provided historical information and can be recalled at any time. The sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset sensitive keywords.

[0027] The invocation and response process of the intelligent agent personal privacy assistant includes a daily invocation process and a special invocation process. The daily invocation process involves interacting with the user to invoke life information, historical memory information, or special sensitive information. The special invocation process occurs when no interaction with the user is achieved within a user-defined time period. If the user is not deemed to have lost their autonomy, the emergency response plan will be invoked and executed.

[0028] Furthermore, the interaction method includes voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal.

[0029] The daily access process also first verifies the user's identity, and only after the verification is successful is it allowed to access life information, historical memory information or special sensitive information;

[0030] When the aforementioned special sensitive information is invoked, a surrounding environment and people recognition step is added; when the environment is identified as unsafe, the user is prompted to change to a safe environment; when the environment is identified as safe, the special sensitive information is output.

[0031] Based on the continuous accumulation of personal data in the local knowledge base, the large model is also continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

[0032] One or more technical solutions provided by this invention have at least the following technical effects or advantages:

[0033] 1. By adopting the "all-in-one machine local deployment large model + physical isolation (i.e., no network connection)" model, the highest priority is to protect users' private data, which can minimize the risk of leakage of personal private data.

[0034] 2. By utilizing the permanent memory function of computer systems, we can address the pain point of memory decline in the elderly and create a user-exclusive local knowledge base. Based on this local knowledge base, we continuously "feed" the local large model to train and optimize it, resulting in an intelligent personal assistant that becomes more personalized and private with each use. This avoids the problems of existing intelligent personal assistants lacking long-term memory and strong personal exclusivity.

[0035] 3. It adopts voice interaction and other methods to ensure ease of use, and uses an engineering technology architecture of "lightweight data acquisition / interaction terminal + all-in-one platform". In the home environment, users can use the matching lightweight data acquisition / interaction terminal to ensure convenient interaction between users and smart agents anytime and anywhere in the home scenario. At present, the voice interaction method has the advantage of cost performance.

[0036] 4. It has methods for handling daily call responses and special case call responses to ensure the security and exclusivity of user private data to the greatest extent.

[0037] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] Figure 1This is a flowchart of the method in Embodiment 1 of the present invention;

[0040] Figure 2 This is a schematic diagram of the device in Embodiment 2 of the present invention;

[0041] Figure 3 This is a schematic diagram of the electronic device in Embodiment 3 of the present invention. Detailed Implementation

[0042] This application provides a method, device, and integrated electronic device for a smart personal privacy assistant. It adopts a "large-scale model with localized deployment of the all-in-one machine + physical isolation (no network connection)" model. It prioritizes the protection of user privacy data, utilizes the permanent memory function of the computer system to solve the pain point of memory decline in the elderly, and forms a user's personal local knowledge base.

[0043] The overall approach of the technical solution in this application is as follows: By adopting the "local deployment of a large model on an all-in-one machine + physical isolation" model, the risk of leakage of personal private data is minimized; the permanent memory function of the computer system is used to solve the pain point of memory decline in the elderly, forming a user's own local knowledge base, and continuously "feeding" it to the local large model, avoiding the problems of existing intelligent personal assistants lacking long-term memory and strong personal exclusivity; the engineering architecture of "lightweight data acquisition terminal + all-in-one machine platform" ensures convenient interaction between users and intelligent agents anytime and anywhere in the home scenario, and the current voice interaction-based approach has cost-effectiveness advantages and is more suitable for elderly users.

[0044] Example 1

[0045] like Figure 1 As shown, this embodiment provides a method for a smart agent's personal privacy assistant, including:

[0046] The local deployment process for large-scale models involves deploying the large-scale model on an all-in-one machine, completing initial setup, and physically isolating the machine from the internet. Prior to this, a suitable hardware configuration should be determined based on the large-scale model's operational needs and the client's final budget, resulting in a customized all-in-one machine integrating hardware and software. This all-in-one machine must have a long service life. Then, a suitable large-scale model is deployed on the all-in-one machine. Considering the needs of key customer groups such as the elderly, the all-in-one machine can be designed as a one-time purchase with a long service life, requiring no maintenance, automatically starting up when powered on, and having sufficient storage capacity. Ideally, it should be configured in one go without subsequent expansion. To control the amount of data accumulating during subsequent use, a "front-end voice interaction plus back-end text storage" approach is adopted.

[0047] The process of creating a personal intelligent assistant: After the all-in-one device is put into use, user-specific data is collected as the data source for the RAG (Retrieval Augmentation Generation) of the large model; the collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanent memory of it; during continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information to form the user's personal local knowledge base, and trains and optimizes the large model to obtain the user's personal intelligent assistant.

[0048] The "life information" mainly refers to general information related to daily life, such as clothing, food, housing, and transportation, and is excluding historical recollections and special sensitive information.

[0049] The historical memory information is automatically organized and cross-validated based on user-provided historical information, and can be recalled at any time. The automatic organization refers to the intelligent agent organizing the information according to dimensions such as timeline and keywords, forming permanent memories. Being recalled means that when the user speaks relevant keywords, the intelligent agent can provide corresponding information in response.

[0050] The special sensitive information refers to the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords;

[0051] The classification is based on the information provided by the user. First, it is determined whether the information contains sensitive words. If so, it is classified as special sensitive information. If not, it is further determined whether the information contains historical time points. If so, it is classified as historical memory information. If not, it is classified as life information.

[0052] Furthermore, based on the continuous accumulation of personal exclusive data in the local knowledge base, the large model is continuously trained and optimized, and the weighting of user personal data is strengthened in the algorithm, making the intelligent agent personal privacy assistant more distinctively user-personalized.

[0053] Furthermore, the interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal. At present, voice interaction has the advantages of high cost-effectiveness and efficiency. Of course, other interaction methods are not excluded. For example, when the cost of video recognition is greatly reduced, data collection / interaction can also incorporate video and other methods.

[0054] Among them, the automatic classification and cross-validation of personal exclusive data by the all-in-one machine mainly involves automatically cross-validating the same information obtained from different dimensions, eliminating abnormal information and retaining normal information. This is a function that the large model currently has built-in.

[0055] The invocation and response process of the intelligent agent's personal privacy assistant includes a routine invocation process and a special invocation process. The routine invocation process first verifies the user's identity before allowing access to life information, historical memory information, or special sensitive information. For example, in daily use, the user can invoke the assistant through dialogue, and a voice identity can be preset. During verification, the intelligent agent can verify the interactive voice to ensure that it only responds to the user's voice.

[0056] The daily access process involves users interacting with the all-in-one machine to access life information, historical memory information, or special sensitive information.

[0057] Example 1 (Example of collecting and retrieving local information):

[0058] After taking their medication in the morning, the user immediately tells the intelligent personal assistant on the all-in-one device, "Xiao i (example of intelligent personal assistant name), I have just taken my blood pressure medication." The intelligent personal assistant records this information and marks the specific time as 8:09 AM.

[0059] Two hours later, the user asked, "Xiao i, did I take my blood pressure medication this morning?" The AI ​​personal assistant replied, "You took your blood pressure medication at 8:09 AM."

[0060] Example 2 (Example of collecting and retrieving life information):

[0061] On April 2, 2025, the user tidied up the Goose brand down jacket he had used last winter and put it in the closet, and simultaneously told the intelligent personal privacy assistant on the all-in-one machine: "Xiao i, I put the Goose brand down jacket under the wardrobe on the left side of the master bedroom."

[0062] On November 5, 2025, a user asked, "Xiao i, where is my down jacket?" The AI ​​agent, Personal Privacy Assistant, countered, "Which one?" The user replied, "Goose brand." The AI ​​agent, Personal Privacy Assistant, replied, "Oh, you put it under the wardrobe on the left side of the master bedroom in April."

[0063] Example 3 (Example of collecting and retrieving historical memory information):

[0064] On February 1, 2025, during lunch, a user recounted to their personal privacy assistant: "Thirty years ago, I went to Xi'an alone on a business trip. After finishing my work, I went to Mount Hua. The path up the mountain was incredibly dangerous; I didn't dare look down from the top." The personal privacy assistant recorded this and noted, "Around 1995 (the recalled time may not be accurate and needs to be verified with other information), the user went to Mount Hua alone. Their impression was that the path up the mountain was very dangerous."

[0065] On June 2, 2025, a user asked the AI ​​agent personal privacy assistant: "Xiao i, I visited Mount Hua many years ago, what year was it exactly?" The AI ​​agent personal privacy assistant retrieved this historical memory information through keywords such as "many years ago" and "Mount Hua" and responded: "It was in 1995."

[0066] On September 5, 2025, the user again told the intelligent personal assistant, "Xiao i, I once went on a business trip to Xi'an and then went to Mount Hua." The intelligent personal assistant retrieved this historical memory information through keywords such as "Xi'an," "business trip," and "Mount Hua," and responded, "The mountain path was very dangerous; I didn't dare look down from the top." This provides emotional support and chat functionality, offering significant emotional value to the user.

[0067] It is evident that as more historical memory information is collected, intelligent personal assistants can leverage their familiarity with their owners' past life experiences and daily habits to provide emotional support and chat with friends, thus enhancing the user's emotional experience and meeting the emotional needs of "empty nest" elderly.

[0068] Example 4 (Example of collecting and retrieving special sensitive information):

[0069] On March 1, 2025, the user told the AI ​​personal privacy assistant: "My China Construction Bank card account number is 90002444329, and the password is 865073." The AI ​​personal privacy assistant immediately recorded this information and marked it as "Special Sensitive Information."

[0070] On March 20, 2025, a user asked the AI ​​agent personal privacy assistant: "What is the password for my China Construction Bank card with the last four digits being 4329?" The AI ​​agent personal privacy assistant replied: "The password is 865073."

[0071] The special invocation process occurs when the user does not interact with the all-in-one machine within a set time period, inferring that the user has lost consciousness, and then the emergency response plan is invoked and executed.

[0072] If a user does not interact with the system within a set time period (e.g., one year), it is inferred that the user has lost their autonomy. In this case, appropriate emergency response measures will be implemented based on the user's pre-set settings. These emergency response measures can be any of the following:

[0073] (1) All system data is retained, and the interaction permission is transferred to a family member designated by the user. This family member must have undergone identity voice authentication in advance.

[0074] (2) The system destroys some data, and transfers the usage rights of the remaining data to a family member designated by the user;

[0075] (3) The system automatically destroys all data.

[0076] Furthermore, when accessing the aforementioned special sensitive information, a surrounding environment and personnel recognition step is added. If an unsafe environment is detected, the user is prompted to change to a safer environment; if a safe environment is detected, the special sensitive information is output. The surrounding environment and personnel recognition step can be performed using voice recognition. If someone other than the user is detected, the intelligent agent prompts, "There are other people present; please leave before replying."

[0077] Therefore, in Example 4 above, when calling special sensitive information, it should be changed to:

[0078] On March 20, 2025, a user asked the AI ​​personal privacy assistant, "What is the password for my China Construction Bank card ending in 4329?" The AI ​​personal privacy assistant replied, "Several people are present; please avoid displaying sensitive information." After the user confirmed, "Other people have left; please state the password," the AI ​​personal privacy assistant replied, "The password is 865073." This ensures the security of highly sensitive information.

[0079] In addition, sensitive information may be updated. If such information is updated, it will still be recorded separately without deleting the old record. When the user subsequently accesses the information, the intelligent agent personal privacy assistant will prioritize the latest data. If the data is incorrect, it will then access the next newest data, and so on. For example, if the intelligent agent personal privacy assistant first records the password for a bank card as 865073, then records it as 865074, and then as 865075, the assistant will prioritize the third recorded password (865075). If the user reports an error, it will then access the second recorded password (865074), and so on.

[0080] Example 2

[0081] Based on the same inventive concept, this application also provides an apparatus corresponding to the method in Embodiment 1, as detailed in Embodiment 2.

[0082] like Figure 2 As shown, this embodiment provides an intelligent agent personal privacy assistant device, including:

[0083] Local deployment module for large models: used to deploy large models on the all-in-one machine, complete the initial setup, physically isolate the all-in-one machine, and not connect it to the Internet;

[0084] The module for creating a personal intelligent assistant: After the all-in-one device is put into use, it collects user-specific data as the data source for the RAG of the large model. This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanently storing this information. During continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information, forming the user's personal local knowledge base. This knowledge base is then used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on the user's provided historical information and can be recalled at any time. The special sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords.

[0085] The interaction methods include voice interaction, allowing users to perform daily interactions anytime in their home environment using a lightweight data collection / interaction terminal. Currently, voice interaction offers a strong cost-performance advantage. Future updates could significantly reduce costs for technologies such as video recognition, and video could also be added to the data collection / interaction methods.

[0086] Based on the continuous accumulation of personalized data in the local knowledge base, the large model is also continuously trained and optimized, and the algorithm strengthens the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

[0087] The intelligent agent's personal privacy assistant's call response module includes daily call processes and special call processes. The daily call process first verifies the user's identity before allowing access to personal information, historical memories, or sensitive information. For example, in daily use, the user can call the agent via dialogue, with pre-set voice identity settings. During verification, the intelligent agent can verify the interactive voice, ensuring it only responds to the user's voice.

[0088] The daily access process involves users interacting with the all-in-one machine to access life information, historical memory information, or special sensitive information.

[0089] Example 1 (Example of collecting and retrieving local information):

[0090] After taking their medication in the morning, the user immediately tells the AI ​​assistant on the all-in-one device, "Xiao i (example of AI assistant name), I have just taken my blood pressure medication." The AI ​​assistant records this information and marks the specific time as 8:09 AM.

[0091] Two hours later, the user asked, "Xiao i, did I take my blood pressure medication this morning?" The AI ​​replied, "You took your blood pressure medication at 8:09 AM."

[0092] Example 2 (Example of collecting and retrieving life information):

[0093] On April 2, 2025, the user tidied up the Goose brand down jacket used last winter and put it in the closet, simultaneously telling the AI ​​assistant on the all-in-one machine: "Xiao i, I put the Goose brand down jacket under the wardrobe on the left side of the master bedroom."

[0094] On November 5, 2025, a user asked, "Xiao i, where is my down jacket?" The AI ​​agent countered, "Which one?" The user replied, "Goose brand." The AI ​​agent responded, "Oh, you put it under the wardrobe on the left side of the master bedroom in April."

[0095] Example 3 (Example of collecting and retrieving historical memory information):

[0096] On February 1, 2025, during lunch, the user recounted to the AI ​​agent: "Thirty years ago, I went to Xi'an alone on a business trip. After finishing my work, I went to Mount Hua. The mountain path was incredibly dangerous; I didn't dare look down from the top." The AI ​​agent recorded this and noted, "Around 1995 (the recalled time may not be accurate and needs to be verified with other information), the user went to Mount Hua alone. My impression is that the mountain path was very dangerous."

[0097] On June 2, 2025, a user asked the AI ​​agent personal privacy assistant: "Xiao i, I visited Mount Hua many years ago, what year was it exactly?" The AI ​​agent personal privacy assistant retrieved this historical memory information through keywords such as "many years ago" and "Mount Hua" and responded: "It was in 1995."

[0098] On September 5, 2025, the user again told the intelligent personal assistant, "Xiao i, I once went on a business trip to Xi'an and then went to Mount Hua." The intelligent personal assistant retrieved this historical memory information through keywords such as "Xi'an," "business trip," and "Mount Hua," and responded, "The mountain path was very dangerous; I didn't dare look down from the top." This provides emotional support and chat functionality, offering significant emotional value to the user.

[0099] It is evident that as more historical memory information is collected, intelligent personal assistants can leverage their familiarity with their owners' past life experiences and daily habits to provide emotional support and chat with friends, thus enhancing the user's emotional experience and meeting the emotional needs of "empty nest" elderly.

[0100] Example 4 (Example of collecting and retrieving special sensitive information):

[0101] On March 1, 2025, the user told the AI ​​agent: "My China Construction Bank card account number is 90002444329, and the password is 865073." The AI ​​agent immediately recorded this and marked it as "Special Sensitive Information."

[0102] On March 20, 2025, a user asked the AI ​​agent, "What is the password for my China Construction Bank card with the last four digits being 4329?" The AI ​​agent replied, "The password is 865073."

[0103] The special invocation process occurs when the user does not interact with the all-in-one machine within a set time period, inferring that the user has lost consciousness, and then the emergency response plan is invoked and executed.

[0104] If a user does not interact with the system within a set time period (e.g., one year), it is inferred that the user has lost their autonomy. In this case, appropriate emergency response measures will be implemented based on the user's pre-set settings. These emergency response measures can be any of the following:

[0105] (1) All system data is retained, and the interaction permission is transferred to a family member designated by the user. This family member must have undergone identity voice authentication in advance.

[0106] (2) The system destroys some data, and transfers the usage rights of the remaining data to a family member designated by the user;

[0107] (3) The system automatically destroys all data.

[0108] Furthermore, when accessing the aforementioned special sensitive information, a surrounding environment and personnel recognition step is added. If an unsafe environment is detected, the user is prompted to change to a safer environment; if a safe environment is detected, the special sensitive information is output. The surrounding environment and personnel recognition step can be performed using voice recognition. If someone other than the user is detected, the intelligent agent prompts, "There are other people present; please leave before replying."

[0109] Therefore, in Example 4 above, when calling special sensitive information, it should be changed to:

[0110] On March 20, 2025, a user asked the AI ​​agent, "What is the password for my China Construction Bank card ending in 4329?" The AI ​​agent replied, "Several people are present; please avoid displaying sensitive information." After the user confirmed, "Other people have left; please state the password," the AI ​​agent replied, "The password is 865073." This ensures the security of highly sensitive information.

[0111] In addition, sensitive information may be updated. If such information is updated, it will still be recorded separately without deleting the old record. When the user subsequently accesses the system, the intelligent agent personal privacy assistant will prioritize the latest data. If it is incorrect, it will then retrieve the next newest data, and so on. For example, if the intelligent agent personal privacy assistant first records the password for a bank card as 865073, then records it as 865074, and then as 865075, the system will prioritize the third recorded password (865075). If the user reports an error, it will then retrieve the second recorded password (865074), and so on.

[0112] Since the apparatus described in Embodiment 2 of the present invention is an apparatus used to implement the method of Embodiment 1 of the present invention, those skilled in the art can understand the specific structure and variations of the apparatus based on the method described in Embodiment 1 of the present invention, and therefore will not be described again here. All apparatuses used in the method of Embodiment 1 of the present invention fall within the scope of protection of the present invention.

[0113] Example 3

[0114] Based on the same inventive concept, this application provides an electronic device embodiment corresponding to Embodiment 1, as detailed in Embodiment 3.

[0115] like Figure 3 As shown, this embodiment provides an integrated electronic device with a large model deployed on it. After initial setup, it is physically isolated, not connected to the internet, and can perform the following processes:

[0116] The process of creating a personal intelligent assistant: After deployment, user-specific data is collected as the data source for the large model's RAG (Real-Time Analysis). This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and sensitive information during interaction, and permanent memory storage. During continuous interaction, the user's personal data is continuously accumulated and automatically categorized. Historical memory information is then cross-validated to form the user's personal local knowledge base, which is used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on user-provided historical information and can be recalled at any time. The sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset sensitive keywords.

[0117] The invocation and response process of the intelligent agent personal privacy assistant includes a daily invocation process and a special invocation process. The daily invocation process involves interacting with the user to invoke life information, historical memory information, or special sensitive information. The special invocation process occurs when no interaction with the user is achieved within a user-defined time period. If the user is not deemed to have lost their autonomy, the emergency response plan will be invoked and executed.

[0118] Furthermore, the interaction method includes voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal.

[0119] The daily access process also first verifies the user's identity, and only after the verification is successful is it allowed to access life information, historical memory information or special sensitive information;

[0120] When the aforementioned special sensitive information is invoked, a surrounding environment and people recognition step is added; when the environment is identified as unsafe, the user is prompted to change to a safe environment; when the environment is identified as safe, the special sensitive information is output.

[0121] Based on the continuous accumulation of personalized data in the local knowledge base, the large model is also continuously trained and optimized, and the algorithm strengthens the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

[0122] Since the electronic device described in this embodiment is the device used to implement the method in Embodiment 1 of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in Embodiment 1 of this application. Therefore, how the electronic device implements the method in the embodiment of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiment of this application falls within the scope of protection of this application.

[0123] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for providing a personal privacy assistant to an intelligent agent, characterized in that: include: The local deployment process of large models: Deploy the large model on the all-in-one machine, complete the initial setup, and physically isolate the all-in-one machine from the Internet; The process of creating a personal intelligent assistant: After the all-in-one device is put into use, user-specific data is collected as the data source for the RAG of the large model. This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanently storing this information. During continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information, forming the user's personal local knowledge base. This knowledge base is then used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on the user's provided historical information and can be recalled at any time. The special sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords. The call response process of the intelligent personal privacy assistant includes a daily call process and a special call process. The daily call process is when the user interacts with the all-in-one machine to call up life information, historical memory information or special sensitive information. The special call process is when the user does not interact with the all-in-one machine within a set time period, it is inferred that the user has lost self-awareness, and the emergency response plan is executed.

2. The method according to claim 1, characterized in that: The interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal. The daily access process also requires user identity verification before accessing life information, historical memory information, or special sensitive information is allowed.

3. The method according to claim 1, characterized in that: When the special sensitive information is invoked, a step of recognizing people in the surrounding environment is added; when the environment is identified as unsafe, the user is prompted to change to a safe environment; when the environment is identified as safe, the special sensitive information is output.

4. The method according to claim 1, characterized in that: Based on the continuous accumulation of personal data in the local knowledge base, the large model is also continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

5. A smart agent personal privacy assistant device, characterized in that: include: Local deployment module for large models: used to deploy large models on the all-in-one machine, complete the initial setup, physically isolate the all-in-one machine, and not connect it to the Internet; The module for creating a personal intelligent assistant: After the all-in-one device is put into use, it collects user-specific data as the data source for the RAG of the large model. This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and special sensitive information during the interaction between the all-in-one device and the user, and permanently storing this information. During continuous interaction, the all-in-one device continuously accumulates the user's personal data and automatically performs processing including classification and cross-validation of historical memory information, forming the user's personal local knowledge base. This knowledge base is then used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on the user's provided historical information and can be recalled at any time. The special sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset special sensitive keywords. The intelligent personal privacy assistant's call response module includes a daily call process and a special call process. The daily call process is when the user interacts with the all-in-one machine to call up life information, historical memory information, or special sensitive information. The special call process is when the user does not interact with the all-in-one machine within a set time period, it is inferred that the user has lost consciousness, and the emergency response plan is invoked for execution.

6. The apparatus according to claim 5, characterized in that: The interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal. The daily access process also requires user identity verification before accessing life information, historical memory information, or special sensitive information is allowed.

7. The apparatus according to claim 5, characterized in that: When the special sensitive information is invoked, a step of recognizing people in the surrounding environment is added; when the environment is identified as unsafe, the user is prompted to change to a safe environment; when the environment is identified as safe, the special sensitive information is output.

8. The apparatus according to claim 5, characterized in that: Based on the continuous accumulation of personal data in the local knowledge base, the large model is also continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

9. An integrated electronic device, characterized in that: A large model is deployed, physically isolated after initial setup, not connected to the internet, and capable of the following processes: The process of creating a personal intelligent assistant: After deployment, user-specific data is collected as the data source for the large model's RAG (Real-Time Analysis). This collection refers to the real-time recording of at least one of the user's life information, historical memory information, and sensitive information during interaction, and permanent memory storage. During continuous interaction, the user's personal data is continuously accumulated and automatically categorized. Historical memory information is then cross-validated to form the user's personal local knowledge base, which is used to train and optimize the large model, resulting in the user's personal intelligent assistant. The historical memory information is automatically organized based on user-provided historical information and can be recalled at any time. The sensitive information is the user's private information, which is judged and specially labeled by the large model based on preset sensitive keywords. The invocation and response process of the intelligent agent personal privacy assistant includes a daily invocation process and a special invocation process. The daily invocation process involves interacting with the user to invoke life information, historical memory information, or special sensitive information. The special invocation process occurs when no interaction with the user is achieved within a user-defined time period. If the user is not deemed to have lost their autonomy, the emergency response plan will be invoked and executed.

10. The integrated electronic device according to claim 9, characterized in that: The interaction methods include voice interaction, allowing users to perform daily interactions at any time in their home environment through a lightweight data collection / interaction terminal. The daily access process also first verifies the user's identity, and only after the verification is successful is it allowed to access life information, historical memory information or special sensitive information; When the aforementioned special sensitive information is invoked, a surrounding environment and people recognition step is added; when the environment is identified as unsafe, the user is prompted to change to a safe environment; when the environment is identified as safe, the special sensitive information is output. Based on the continuous accumulation of personal data in the local knowledge base, the large model is also continuously trained and optimized to strengthen the weighting of user personal data, making the intelligent agent personal privacy assistant more distinctively personalized.

Images