Computer artificial intelligence system based on large model

Abstract

The invention discloses a computer artificial intelligence system based on a large model, which comprises a computer artificial intelligence system body, and the computer artificial intelligence system body comprises an LLM Agent architecture platform, an optimization design platform, an intelligent storage platform and an Agent platform. According to the computer artificial intelligence system based on the large model, the optimization design platform is arranged, the reasoning speed and the cost efficiency are optimized, the performance of a multi-agent system is improved, and the system has the advantages of being high in practicability and high in practicability. Stronger adaptability and sustainability are shown in the aspects of user interaction and long-term development of the system, and higher efficiency and accuracy are shown when a specific task is processed; by setting the Agent platform, the customizable and dialogue Agents are provided, the limitation of the Agents is solved by integrating the Agents with people and tools and performing dialogue among the Agents through automatic chat, and the cost of information analysis of the Agents is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of computer artificial intelligence systems, and in particular to a computer artificial intelligence system based on a large model. Background Art

[0002] In the current information age, the development speed and influence of large language models are becoming increasingly significant. The powerful reasoning and generation capabilities of large models have become the best components for building intelligent entities. At present, the industry generally believes that applications based on large models are concentrated in two directions: RAG and Agent. Regardless of the application, designing, implementing and optimizing applications that can fully utilize the potential of large models requires a lot of effort and expertise. As developers begin to create increasingly complex artificial intelligence systems, the development process inevitably becomes more complicated.

[0003] However, traditional AI systems have the following disadvantages:

[0004] Traditional AI systems that rely solely on LLM are not enough to answer such complex questions. Although the RAG system that combines LLM with an external knowledge base can provide some help, further operations are required to fully answer this question, and the information analysis is time-consuming and costly. Summary of the invention

[0005] The purpose of the present invention is to provide a computer artificial intelligence system based on a large model to solve the problem that the traditional artificial intelligence system proposed in the above background technology only relies on LLM to answer such complex problems. Although the RAG system combining LLM with an external knowledge base can provide some help, further operations are required to fully answer this problem, and the information analysis time is long and the analysis cost is high.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a computer artificial intelligence system based on a large model, comprising a computer artificial intelligence system ontology, wherein the computer artificial intelligence system ontology comprises an LLM Agent architecture platform, an optimization design platform, an intelligent storage platform and an Agent platform;

[0007] The LLM Agent architecture platform uses a large language model as its core computing engine, so it is not limited to text generation, but can also conduct conversations, complete tasks, reason, and exhibit a certain degree of autonomous behavior;

[0008] The optimized design platform understands complex instructions, adapts to different scenarios and continuously optimizes its own behavior, showing higher efficiency and accuracy when handling specific tasks;

[0009] The smart storage platform is responsible for storing past thoughts, actions, observations, and interactions with users;

[0010] The agent platform easily configures the use and roles of LLMs in agents through advanced reasoning features, automatically solves complex tasks through group chat, artificial intelligence and supervision are implemented through agents with different participation levels and modes, and agents have native support for LLM-driven code / function execution, automatically solve tasks through code generation, execution and debugging, using the provided tools as functions.

[0011] As a preferred technical solution of the present invention, the LLM Agent architecture platform includes a user request module, an agent module, a tool module, a memory module and a planning module. The user request module receives user questions or requests. The agent module builds an agent system with a large language model as the core, acting as the main brain of the system and the core of multi-task coordination. The tool module defines the role and personality of the agent, including background, character, social environment and demographic information. The memory module manages the past behavior of the agent, and the planning module plans the future actions of artificial intelligence.

[0012] As a preferred technical solution of the present invention, the optimization design platform includes an optimization and adjustment module, a modal interaction module and a data security module. The optimization and adjustment module has rich context understanding and continuous learning capabilities. It not only processes and memorizes a large amount of interactive information, but also needs to continuously optimize execution strategies and prediction models. The modal interaction module introduces multimodal interaction and integrates multiple input and output forms such as text, images, and sounds, allowing the system to handle complex tasks and environments more naturally and effectively. The data security module enhances data security and reliability to ensure stable operation of the system.

[0013] As a preferred technical solution of the present invention, the intelligent storage platform includes a short-term memory module, a long-term memory module and a hybrid memory module. The short-term memory module focuses on the contextual information of the current situation, is short-lived and limited, and is usually implemented through learning restricted by the context window. The long-term memory module stores the historical behavior and thinking of the intelligent agent, and is implemented through external vector storage to quickly retrieve important information. The hybrid memory module integrates short-term and long-term memory, which not only optimizes the intelligent agent's understanding of the current situation, but also strengthens the use of past experience, thereby improving its ability of long-term reasoning and experience accumulation.

[0014] As a preferred technical solution of the present invention, the Agent platform includes an Assistant Agent module and a Multi-Agent session module. The Assistant Agent module calls the automatic chat between the Assistant Agent and the User Agent to build an enhanced version of the ChatGPT+Code Interpreter+ plug-in, which has customizable automation capabilities and is used in a customized environment or embedded in a larger system. The Multi-Agent session module has many benefits centered around the Agent session, including: naturally handling ambiguity, feedback, progress, and collaboration, enabling effective coding-related tasks such as using tools through back-and-forth troubleshooting, allowing users to seamlessly choose to join or exit through the chat agent, and achieving collective goals through the collaboration of multiple experts. AutoGen supports automatic chat and diverse communication modes, making it easy to orchestrate complex, dynamic workflows and experimental versatility.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. By setting up an optimized design platform, optimizing reasoning speed and cost efficiency, improving the performance of multi-agent systems, showing stronger adaptability and sustainability in user interaction and long-term system development, and showing higher efficiency and accuracy in processing specific tasks;

[0017] 2. By setting up an agent platform, customizable and conversational agents are provided, leveraging the strongest features of LLM while addressing their limitations by integrating with people and tools and conducting conversations between multiple agents through automatic chat, thereby reducing the cost of information analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the architecture of the computer artificial intelligence system of the present invention;

[0019] Figure 2 This is a schematic diagram of the architecture of the LLM Agent architecture platform of the present invention;

[0020] Figure 3 A schematic diagram of the architecture of the optimization design platform of the present invention;

[0021] Figure 4 It is a schematic diagram of the architecture of the intelligent storage platform of the present invention;

[0022] Figure 5 It is a schematic diagram of the architecture of the Agent platform of the present invention;

[0023] Figure 6 The state diagram used by the Agent module of the present invention;

[0024] Figure 7 The Multi-Agent session module of the present invention uses a state diagram. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0026] See also Figure 1-7 , the present invention provides a computer artificial intelligence system based on a large model, including a computer artificial intelligence system ontology, the computer artificial intelligence system ontology includes an LLM Agent architecture platform, an optimization design platform, an intelligent storage platform and an Agent platform;

[0027] The LLM Agent architecture platform uses a large language model as its core computing engine, which is not limited to text generation, but can also conduct conversations, complete tasks, reason, and demonstrate a certain degree of autonomous behavior;

[0028] The optimized design platform understands complex instructions, adapts to different scenarios and continuously optimizes its own behavior, showing higher efficiency and accuracy in handling specific tasks;

[0029] The intelligent storage platform is responsible for storing past thoughts, actions, observations, and interactions with users;

[0030] The Agent platform easily configures the use and roles of LLMs in Agents through advanced reasoning features. Complex tasks are automatically solved through group chats. Artificial intelligence and supervision are implemented through Agents with different levels and modes of participation. Agents have native support for LLM-driven code / function execution, automatically solving tasks through code generation, execution and debugging, using the provided tools as functions.

[0031] The LLM Agent architecture platform includes a user request module, an agent module, a tool module, a memory module and a planning module. The user request module receives user questions or requests. The agent module builds an agent system with a large language model as the core, acting as the main brain of the system and the core of multi-task coordination. The tool module defines the role and personality of the agent, including background, character, social environment and demographic information. The memory module manages the past behavior of the agent. The planning module plans the future actions of artificial intelligence.

[0032] The optimization design platform includes an optimization and adjustment module, a modal interaction module and a data security module. The optimization and adjustment module has rich contextual understanding and continuous learning capabilities. It not only processes and memorizes a large amount of interactive information, but also needs to continuously optimize execution strategies and prediction models. The modal interaction module introduces multimodal interaction and integrates multiple input and output forms such as text, images, and sounds, allowing the system to handle complex tasks and environments more naturally and effectively. The data security module enhances data security and reliability to ensure stable system operation.

[0033] The intelligent storage platform includes short-term memory modules, long-term memory modules and hybrid memory modules. The short-term memory module focuses on the contextual information of the current situation. It is short-lived and limited, and is usually achieved through learning with context window restrictions. The long-term memory module stores the historical behavior and thinking of the intelligent agent through external vector storage to quickly retrieve important information. The hybrid memory module integrates short-term and long-term memory to not only optimize the intelligent agent's understanding of the current situation, but also strengthen the use of past experience, thereby improving its ability of long-term reasoning and experience accumulation.

[0034] The Agent platform includes an Assistant Agent module and a Multi-Agent Conversation module. The Assistant Agent module invokes automatic chats between the Assistant Agent and the User Agent to build an enhanced version of the ChatGPT+Code Interpreter+ plug-in, which has customizable automation capabilities and can be used in customized environments or embedded in larger systems. The Multi-Agent Conversation module's Agent conversation-centric design has many benefits, including: naturally handling ambiguity, feedback, progress, and collaboration, enabling effective coding-related tasks such as using tools through back-and-forth troubleshooting, allowing users to seamlessly opt in or out through the chat agent, and achieving collective goals through the collaboration of multiple experts. AutoGen supports automatic chats and diverse communication modes, making it easy to orchestrate complex, dynamic workflows and experimental versatility.

[0035] Embodiment 1:

[0036] When the embodiment of the present application is in use: the computer artificial intelligence system body is driven by a large language model and is designed to autonomously handle a series of complex artificial intelligence tasks. HuggingGPT integrates the capabilities of LLM and the resources of the machine learning community, such as the combination of ChatGPT and Hugging Face, so that it can process inputs from different modalities. Specifically, LLM plays the role of the brain here. On the one hand, it decomposes tasks according to user requests, and on the other hand, it selects suitable models to perform tasks based on model descriptions. By executing these models and integrating the results into planned tasks, HuggingGPT can autonomously complete complex user requests. This process shows the complete process from task planning to model selection, to task execution, and finally to response generation. First, HuggingGPT uses ChatGPT to analyze user requests to understand their intentions and decompose them into solutions. Next, it will select Hugging The expert models that are most suitable for performing these tasks are hosted on Face. Each selected model is called and executed, and its results are fed back to ChatGPT. Finally, ChatGPT integrates the prediction results of all models to generate responses for users. HuggingGPT's working method not only expands the capabilities of traditional single-mode processing, but also provides efficient and accurate solutions in cross-domain tasks through its intelligent model selection and task execution mechanism.

[0037] Embodiment 2:

[0038] When using the embodiments of this application: Figure 6 ,A simple way to use AutoGen Assistant Agent is to call the automatic chat between Assistant Agent and User Agent. It is easy to build an enhanced version of ChatGPT+Code Interpreter+ plug-in, which has customizable automation capabilities, used in customized environments, and embedded into larger systems. Assistant Agent plays the role of an artificial intelligence assistant, such as Bing Chat; User Agent plays the role of the user and simulates the user's behavior, such as code execution. AutoGen automates the chat between the two agents while allowing human feedback or intervention. The user agent interacts seamlessly with humans and uses the tool when necessary. Combined with the instruction manual attached Figure 7,AutoGen supports automatic chatting and diverse communication modes, making it easy to orchestrate complex, dynamic workflows and experimental versatility. A special agent called "GroupChatManager" is used to support group chatting between multiple agents. GroupChatManager is a special agent that repeats the following three steps: select a speaker, ask the speaker to respond, and broadcast the selected speaker's information to all other agents. In summary, AutoGen is designed as a general infrastructure for building LLM applications. Its dialogue mode supports almost all mode types of existing LLM systems. In "static" mode, the topology of the agent remains unchanged regardless of the input. AutoGen allows flexible dialogue modes, including static and dynamic modes customized according to different application needs. Its Multi-Agent system executes the code generated by LLM, allowing human participation in the system execution process.

[0039] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A computer artificial intelligence system based on a large model, including a computer artificial intelligence system ontology, characterized in that: The computer artificial intelligence system ontology includes LLM Agent architecture platform, optimization design platform, intelligent storage platform and Agent platform; The LLM Agent architecture platform uses a large language model as its core computing engine, so it is not limited to text generation, but can also conduct conversations, complete tasks, reason, and exhibit a certain degree of autonomous behavior; The optimized design platform understands complex instructions, adapts to different scenarios and continuously optimizes its own behavior, showing higher efficiency and accuracy when handling specific tasks; The smart storage platform is responsible for storing past thoughts, actions, observations, and interactions with users; The agent platform easily configures the use and roles of LLMs in agents through advanced reasoning features, automatically solves complex tasks through group chat, artificial intelligence and supervision are implemented through agents with different participation levels and modes, and agents have native support for LLM-driven code / function execution, automatically solve tasks through code generation, execution and debugging, using the provided tools as functions.

2. A computer artificial intelligence system based on a large model according to claim 1, characterized in that: The LLM Agent architecture platform includes a user request module, an agent module, a tool module, a memory module and a planning module. The user request module receives user questions or requests. The agent module builds an agent system with a large language model as the core, acting as the main brain of the system and the core of multi-task coordination. The tool module defines the role and personality of the agent, including background, character, social environment and demographic information. The memory module manages the past behavior of the agent. The planning module plans the future actions of artificial intelligence.

3. A computer artificial intelligence system based on a large model according to claim 1, characterized in that: The optimization design platform includes an optimization and adjustment module, a modal interaction module and a data security module. The optimization and adjustment module has rich context understanding and continuous learning capabilities. It not only processes and memorizes a large amount of interactive information, but also needs to continuously optimize execution strategies and prediction models. The modal interaction module introduces multimodal interaction and integrates multiple input and output forms such as text, images, and sounds, allowing the system to handle complex tasks and environments more naturally and effectively. The data security module enhances data security and reliability to ensure stable operation of the system.

4. A computer artificial intelligence system based on a large model according to claim 1, characterized in that: The intelligent storage platform includes a short-term memory module, a long-term memory module and a hybrid memory module. The short-term memory module focuses on the contextual information of the current situation, is short-lived and limited, and is usually implemented through learning with context window restrictions. The long-term memory module stores the historical behavior and thinking of the intelligent agent, and is implemented through external vector storage to quickly retrieve important information. The hybrid memory module integrates short-term and long-term memories, which not only optimizes the intelligent agent's understanding of the current situation, but also strengthens the use of past experience, thereby improving its ability of long-term reasoning and experience accumulation.

5. The computer artificial intelligence system based on a large model according to claim 1, characterized in that: The Agent platform includes an Assistant Agent module and a Multi-Agent Conversation module. The Assistant Agent module invokes automatic chat between the Assistant Agent and the User Agent to build an enhanced version of the ChatGPT+Code Interpreter+ plug-in, which has customizable automation capabilities and can be used in a customized environment or embedded in a larger system. The Multi-Agent Conversation module has many benefits centered around Agent conversations, including: naturally handling ambiguity, feedback, progress, and collaboration, enabling effective coding-related tasks such as using tools through back-and-forth troubleshooting, allowing users to seamlessly opt in or out through the chat agent, and achieving collective goals through the collaboration of multiple experts. AutoGen supports automatic chat and diverse communication modes, making it easy to orchestrate complex, dynamic workflows and experimental versatility.

Images