20 results about "Service framework" patented technology

The invention discloses a real-time translation recognition system under a cloud service framework, belongs to the technical field of real-time translation, and solves the problems that an existing translation system is insufficient in real-time performance, poor in scene adaptability, weak in privacy protection, slow in model evolution and the like. The dynamic model management engine obtains adaptive slices from a model slice factory according to scenes, equipment states and network quality and distributes the adaptive slices to edges, and the adaptive slices are distributed to a cloud-side collaborative reasoning system; the cloud-side collaborative reasoning system comprises a cloud-side collaborative reasoning system, a cloud-side collaborative reasoning system, a cloud-side collaborative reasoning system and a cloud-side collaborative reasoning system; the multi-modal perception engine fuses audio, images and dialogue history to generate a structured context vector and improve translation context fitting degree, the cloud edge cooperation engine takes an edge model as a core, processes different complexity tasks in combination with a cloud end, and constructs a data closed loop by incremental learning and a federation engine to realize model optimization and privacy protection; according to the system, the real-time performance, accuracy and safety are improved through cloud edge collaboration, dynamic adaptation and continuous learning, and the system is suitable for multi-scene real-time translation.

The present disclosure provides a service implementation method and device based on a heterogeneous service framework, equipment and medium, wherein the application service implementation method based on the heterogeneous service framework comprises: accessing a probe on an application side, and adding a configuration file of the probe in a start script on the application side, the configuration file being pre-provided with an address of an application service of a data plane of the heterogeneous service framework, and the probe performing data interaction with the data plane based on the address of the application service. Through the embodiments of the present disclosure, the underlying code logic of the registration, discovery, calling, routing, flow control and the like of a plurality of service providers and consumers can be dynamically modified and reset to a unified implementation architecture, thereby achieving unified management of the heterogeneous service framework in a fully automated manner without invasion of the application development party, without introducing any new potential fault nodes, and improving the reliability and fusion of the heterogeneous service framework.

The invention relates to a model service operation method, computer equipment, a readable storage medium and a program product, and relates to the technical field of model services. The method comprises the following steps: in response to an operation instruction for target model service, obtaining a target model service template of the target model service from the operation instruction; matching the target model service template with each candidate model service framework to obtain a matching result between the target model service template and each candidate model service framework; screening out a target model service framework of which the matching result meets a matching condition from the plurality of candidate model service frameworks; and running the target model service framework based on the template content of the target model service template to obtain a running result of the target model service. By adopting the method, the operation efficiency of the template service can be improved.

The application discloses a method for constructing a dynamic and expandable single-agent runtime framework, and belongs to the technical field of artificial intelligence and large language model agents. The framework comprises a dynamic and expandable runtime framework module, a memory system module, a skill system module and a tool system module. The runtime framework module adopts a three-layer separation architecture and runtime pool management, supports LLM provider hot switching and a double-axis scheduling mechanism; the memory system module realizes five types of memory division, time-controllable context compression and experience sedimentation driven by a minimum subtask stage boundary; the skill system module realizes skill discovery through double-layer search, skill creation through an A / B evaluation cycle, and skill evolution through a closed loop using statistical tracking; and the tool system module adopts a four-level exposure model, a three-level permission control and a two-level dynamic routing mechanism. The application solves the problems of poor expandability and uncontrollable compression timing of existing frameworks, and realizes efficient, controllable and continuously evolving single-agent services.

The invention provides a micro-service platform management and control method for lossless management of different micro-service framework applications, belongs to the technical field of micro-service management and control, and aims to solve the problems that the difference between an existing Java micro-service framework operation mechanism such as Apache Dubbo, SpringCloud and HSF and a governance model is large, cross-framework interconnection and intercommunication are difficult, service governance is scattered, and codes need to be modified. A JavaAgent plug-in is loaded when a micro-service JVM is started, a framework core component is enhanced in a class loading stage, service registration, discovery and calling processes are intercepted, the framework core component is converted into a unified data model and a communication protocol, and cross-framework intercommunication and unified management are realized in combination with a unified management and control center comprising a registration center, a configuration center, a monitoring center and a service management center. According to the method, business codes and framework source codes do not need to be modified, lossless management of the micro-service is realized, and the management and control efficiency and the system stability are improved.

This invention discloses an automated tuning method and system for large language model service configurations, belonging to the fields of artificial intelligence and cloud computing technology. The invention includes an offline analysis module for performance profiling of the large language model (LLM) service system under a finite configuration set and for collecting training data; an online search module for exploring and evaluating the performance of different configurations in the configuration space based on a Monte Carlo tree search algorithm; and a simulator module for receiving the configurations generated by the online search module and predicting the performance metrics of the configurations under a target workload based on a logical step simulation model. This invention effectively solves the problems of low search efficiency and poor performance of existing automatic tuning methods in existing large language model service frameworks (such as vLLM, SGLang, etc.) due to the huge number of configuration parameters (reaching a petabyte-scale search space) and the complex dependencies and mutual exclusion relationships between parameters.

The invention discloses a Nodejs-based progressive service framework, and relates to the technical field of computers, and the Nodejs-based progressive service framework comprises a double-layer architecture layer which is used for separating extended functions of core functions; the double-layer architecture layer comprises a core package and a framework template, the core package comprises operation architecture codes and general functions, the framework template appoints a system file structure and a function writing mode, the core package pre-integrates Modbus, SNMP and MQTT protocol analysis engines, and the plug-and-play equipment access capability is achieved through the standardized adaptation layer; an API (Application Program Interface) service framework and a communication management and data processing component are arranged in the core packet; the equipment access layer is used for automatically identifying the type of equipment when the equipment is dynamically or statically accessed, and carrying out conversion and access through the protocol adaptation layer; an operation and maintenance management layer; a modular design layer; an extension layer; according to the invention, a common equipment protocol analysis engine is preset in the core packet, and the equipment access process is simplified through the standardized adaptation layer, so that the complexity of development and integration is reduced.

In general, certain embodiments described herein relate to a method for performing inferences on noisy compressed data. The method includes receiving, from a client, the noisy compressed data at a machine learning system. The noisy compressed data is transmitted over a network with the machine learning system being a machine learning as a service configuration. An amount of noise in the noisy compressed data is based on a privacy level. The method further includes classifying, by a classifier of the machine learning system, the noisy compressed data by executing an inference on the noisy compressed data to obtain classification results, and transmitting the classification results to the client via a network. The inference is executed by a trained model, which is trained on noisy compressed training data. By transmitting noisy compressed data, the privacy of actual data is preserved.

The invention relates to a privacy protection-oriented cryptography credential model watermark service method and system, and the core of the invention is to initiate a privacy protection-based model watermark, i.e., a service framework: a service provider does not need to access private data and model information of a model owner at all, and uses a proxy model pool and a public data set of the service provider; and generating a watermark pair with strong mobility and robustness through iterative optimization of a watermark trigger image generator and a plurality of task agent models. According to the method, the digital signature and the credible timestamp are combined for the first time, a verifiable watermark voucher is generated, and the generation time, the service source and the integrity of the watermark are ensured not to be tampered and non-repudiated, so that ownership disputes such as model overwriting attacks and the like are effectively coped with. According to the method, three bottlenecks of privacy disclosure, poor mobility and insufficient legal evidence force in the existing watermarking technology are fundamentally solved, and the method is particularly suitable for ownership protection of image processing models.

The application provides a blockchain smart contract service framework design method based on a hash multi-way tree and an application system, relates to the technical field of blockchain smart contracts, and forms a hash multi-way tree on the blockchain by connecting a plurality of node smart contracts to each other with each node smart contract as a tree node, realizes the construction of a blockchain smart contract framework based on the hash multi-way tree, realizes the scalability and the integrated multi-way tree structure on the blockchain, internally builds an access mechanism based on roles and permissions for each tree node of the hash multi-way tree, checks whether a user has specific roles and permissions, sets a user access interface externally, provides a serialization engine and a deserialization engine, realizes two functional domains of data management and contract management, and can effectively improve the scalability, compatibility, access speed, maintainability, data confidentiality and data integrity of a DApp.

The invention relates to a text processing method and apparatus, a device, a medium and a program product. The method comprises the steps of obtaining an identifier of a target model and a target dialogue text carried by a model reasoning request; under the condition that the first type of dialogue template is built in the target model, converting the target dialogue text into an input sequence by utilizing the first type of dialogue template built in the target model; under the condition that the target model is not internally provided with the first type of dialogue template and the second type of dialogue template supports the target model, converting the target dialogue text into an input sequence by utilizing the second type of dialogue template; and under the condition that the target model is not internally provided with the first type of dialogue template and the second type of dialogue template does not support the target model, converting the target dialogue text into an input sequence by utilizing the first type of dialogue template predefined outside the target model. Therefore, the target model service platform can support the dialogue service of models adopting different dialogue templates under the target model service framework.

The invention discloses a method for assisting anti-fraud algorithm iteration based on an AI agent, and the method comprises the steps: building a rear-end service framework through a network interface based on an MCP technology; constructing a system cue word for guiding the LLM to generate an output conforming to expectation; testing the capability of the LLM for performing complex task planning according to the natural language instruction until a set requirement is met; the back-end service framework analyzes the input fraud-related request, splices the input fraud-related request with a system cue word and then transmits the spliced fraud-related request and the system cue word to the LLM; calling of all tools in the back-end service framework is automatically planned through the LLM according to the received content; and generating a fraud identification result through the LLM according to an output result of each tool in the back-end service framework. According to the method, fraud identification is performed through an automatic incremental training process and a plurality of identification models, so that the identification capability of the identification models is enhanced, the fraud identification precision is greatly improved, and fraud information is effectively intercepted.

The application provides a government affair question and answer method and product based on knowledge enhanced large language model, and relates to the technical field of artificial intelligence. The current government affair question and answer system mainly adopts the mode of manual reply, which is low in efficiency, and the system based on RAG lacks query discrimination mechanism, cannot identify intention difference and utilize dialogue context, and affects the quality of retrieved knowledge. In view of the defects existing in the related technologies, the application provides a government affair question and answer method based on knowledge enhanced large model, and specifically, in view of the problem of low efficiency of manual reply, a complete government hotline service framework containing intelligent problem shunting and structured feedback is designed; in view of the problem of poor quality of retrieved knowledge of the traditional RAG system, a multi-stage semantic enhancement retrieval method is designed, and cross-domain knowledge is supplemented through network query, so that accurate response and service closed loop of government consultation are realized.

The application relates to the technical field of microservices, and discloses a service framework switching method, device, equipment and storage medium, the method comprises the following steps: determining a mixed deployment distributed microservice framework and a control plane and a data plane of a service mesh framework, wherein the control plane comprises a control module for managing the distributed microservice framework and the service mesh framework, and the data plane comprises data in the distributed microservice framework and the service mesh framework; acquiring monitoring data in the distributed microservice framework and the service mesh framework according to the data plane; switching between the distributed microservice framework and the service mesh framework according to the monitoring data; in the above manner, the monitoring data of the distributed microservice framework and the service mesh framework is acquired, then the current service framework and the target service framework are switched through the monitoring data, so that the business instance does not need to be stopped during the switching process of the service framework, and the smooth running of the stateful service can be ensured.

The invention belongs to the technical field of large language model services, and particularly relates to a dynamic reasoning load balancing service method for a doctor-patient dialogue model. According to the method, a collaborative optimization end-to-end service framework is constructed, and the core of the collaborative optimization end-to-end service framework is a concurrent sensing dynamic routing mechanism; the mechanism can monitor the amount of concurrent requests of doctor-patient dialogues in real time, and intelligently route the requests to heterogeneous reasoning replica clusters with different parallel (such as tensor parallel and replica parallel) strategies, so that the problem that when an existing service architecture faces a low-concurrency and high-concurrency mixed load, the concurrency of the doctor-patient dialogues is influenced is effectively solved. The problem of resource waste or request blocking caused by adopting static single configuration is solved, and elastic and efficient utilization of computing resources is achieved. Experimental results prove that the effect of the method is remarkably better than that of the existing technical method in the aspects of improving the overall throughput of a doctor-patient dialogue system, reducing the average waiting time of the users and supporting simultaneous online consultation of larger-scale users.

The present specification provides a service project security protection and safe execution method, device, medium and product. The service project security protection method comprises: obtaining a target service project written in an interpretive language, containing decoupled service framework source code and algorithm logic source code; performing obfuscation encryption processing on the algorithm logic source code to generate a ciphertext file and a runtime loader; packaging the ciphertext file and the runtime loader into an algorithm software package carrying a version identifier, and uploading to a private software warehouse; generating a service framework software package containing the service framework source code and dependency configuration information; the algorithm software package is configured to be able to be pulled by a package management tool according to the dependency configuration information to be integrated and deployed with the service framework source code; and the runtime loader is configured to be loaded into a running environment where the service framework source code is located in a service starting stage, and to perform dynamic decryption and de-obfuscation processing on the ciphertext file for execution in a service running process.

The application discloses an online adjustment method and device based on a micro-service framework general development specification architecture, which comprises the following steps: according to a detected online architecture adjustment instruction, determining an architecture project to be adjusted and the project structure of the project, a first editing requirement for the project and a second editing requirement for the structure corresponding to the project; then, according to the project structure, the first and second editing requirements, combining the architecture development specification, performing an online architecture adjustment operation on the project, and generating a target structure file matched with a preset data structure according to the obtained adjustment result and the target template data of the completed configuration, the target structure file being used to be displayed to a target user through an online preview editing module to trigger the target user to perform a structure viewing and / or structure editing operation on the target structure file through the online preview editing module. It can be seen that the application can reduce the calling complexity of the development architecture and improve the usability of the development architecture.

The invention discloses an AI behavior dynamic reprogramming system and method based on the on-demand creation of a contextualized sandbox.The system comprises a contextualized sandbox module, a behavior monitoring module, a dynamic reprogramming module and a supporting framework module, the contextualized sandbox module is used for generating an isolated simulation environment according to the demand, dynamic storage and recovery of the environment state are achieved, and the behavior monitoring module is used for monitoring the dynamic reprogramming module; the behavior monitoring module is used for collecting environment perception, decision logic and behavior results of an intelligent agent in real time and generating memory flow data, the dynamic reprogramming module analyzes memory flow and generates a behavior adjustment strategy based on a large language model, and the supporting framework module is used for providing a micro-service framework and supporting creation and management of a sandbox; according to the method, the influence range of malicious behaviors is controlled in a virtual environment through sandbox isolation, the system crash risk is reduced, the sandbox creation time is shortened through resource pooling and template multiplexing, fine-grained instruction injection is supported through a dynamic adjustment strategy, and the AI task completion rate is increased.

Embodiments of the present application provide a multi-level topology graph generation method and device, electronic equipment and storage medium, which are applied to a cloud native service framework. The cloud native service framework includes a micro-service platform. Original process information and network connection information of a local service end process are obtained through the micro-service platform. The original process information is subjected to regular matching processing, and the original process information that is successfully matched is labeled to obtain service information for a service. The service information is input into a preset service feature library to obtain a service group corresponding to the service information and service group information of the service group. The service group information includes a target service group label associated with the service group. Service groups with the same type label of the target service group label are aggregated to obtain a business system corresponding to the service group and system attribute information of the business system. A multi-level topology graph is generated according to the network connection information, the service information, the service group information and the system attribute information.