CN119668593A - A zero-code AI agent development and deployment data management system and method - Google Patents
A zero-code AI agent development and deployment data management system and method Download PDFInfo
- Publication number
- CN119668593A CN119668593A CN202411765389.9A CN202411765389A CN119668593A CN 119668593 A CN119668593 A CN 119668593A CN 202411765389 A CN202411765389 A CN 202411765389A CN 119668593 A CN119668593 A CN 119668593A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- models
- data stream
- consistency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Computer And Data Communications (AREA)
Abstract
The invention relates to the field of agent development, in particular to a zero-code AI agent development and deployment data management system and method, comprising the following steps: the system comprises a data flow test module, a model communication module, a feedback processing module, a project migration module and a secondary development module, wherein the data flow test module is used for performing connectivity test to obtain a conversion interface, the model communication module is used for connecting the conversion interface in a forward direction to generate a conversion queue, the feedback processing module is used for sharing input data flow among models, the project migration module is used for adjusting the sharing data proportion of the models, and the secondary development module is used for developing a new model according to the service providing proportion of the models.
Description
Technical Field
The invention relates to the field of agent development, in particular to a zero-code AI agent development and deployment data management system and method.
Background
The development of the zero-code AI intelligent agent is a technology for quickly constructing the usable AI intelligent agent by using an existing AI development platform and tool and through model identification configuration and customization operation under the condition of no need of writing complex codes. The agent deployed by zero crossing codes can perform natural language dialogue, image recognition, predictive analysis and other tasks.
The zero-code intelligent agent realizes interaction with a user through processing the data flow by the intelligent model, so that the zero-code intelligent agent needs to be matched and fitted for a big data model for development, but is limited by the functions of a development platform, and the models are independent of each other, so that a developer is limited in innovation and customization in the process of developing the intelligent agent by the zero code, and the user can only select one of the interaction models of the intelligent agent, and cannot realize complex or personalized requirements.
In addition, the developer needs to develop a new model aiming at user feedback after deploying the model, but most users do not know the working principle of the model, and only can infer the requirement of the user on the model function through the use process of the user, which requires the developer to conduct additional investigation work and is not beneficial to secondary development of the intelligent agent model by the developer.
Disclosure of Invention
The invention aims to provide a zero-code AI intelligent agent development and deployment data management system and method, so as to solve the problems in the background technology.
In order to solve the technical problems, the invention provides a zero-code AI intelligent agent development and deployment data management system, which comprises a data flow test module, a model communication module, a feedback processing module, a project migration module and a secondary development module;
The data flow test module is used for generating random vector data flows, performing data connectivity test on each calling interface between the development models, taking the data flow corresponding to the highest data connectivity as a characteristic data flow, testing the time delay of the calling interface between the models, and taking the interface with the lowest time delay as a conversion interface between the models;
the model communication module is used for calculating the consistency of the characteristic data streams of all models, arranging all development models according to a consistency descending order, connecting conversion interfaces among the models in a forward direction to form a model conversion queue, developing a matched processing program based on the model conversion queue, and deploying the program on a service platform;
The feedback processing module is used for acquiring input data streams provided by users, calculating the consistency of the input data streams on the characteristic data streams of each model, selecting processing models on a model queue according to a consistency sequence, directing a directional pointer to the model with the largest consistency, enabling the sum of communication delay of a conversion interface between the models to be smaller than a preset threshold, sharing the input data streams according to a consistency proportion in the processing models, and outputting reply data streams to the users on a service platform through model processing;
The project migration module is used for acquiring feedback data streams of users, when the consistency between the feedback data streams and the standard forward data streams is smaller than the consistency between the two model characteristic data streams, increasing the sharing data proportion of a unit of a subsequent model in the model conversion queue, otherwise, reducing the sharing data proportion of a unit, reacquiring the feedback data streams of the users until the sharing data proportion of the preamble model is reduced to a threshold value, closing a conversion interface of the preamble model, and moving a directional pointer by one bit;
The secondary development module is used for outputting the position of the directional pointer after the user goes offline and the data sharing proportion of each model in the model conversion queue in the service process, obtaining the service weight of each model according to the accumulated multiplication of the sharing proportion, the sharing duration and the time attenuation coefficient, developing a new model according to the service weight of the model after the development period is finished, and deploying the agent on the service platform again.
Further, the data flow testing module comprises a communication testing unit and a time delay testing unit;
The connectivity test unit is used for generating a random vector data stream, testing connectivity between every two models, and storing the data stream generated when the connectivity is highest;
the time delay test unit is used for testing the calling time delay of each interface between the models and marking the conversion interface with the lowest time delay.
Further, the model communication module comprises a consistent arrangement unit and an interface connection unit;
the consistent arrangement unit is used for calculating the consistency of the characteristic data flow among the models and arranging all the models according to the consistency sequence;
The interface connection unit is used for connecting all models in one way through a conversion interface among the models to form a model conversion queue, and setting an initial pointer of the queue.
Further, the feedback processing module comprises an input and output unit, a model deployment unit and a proportional access unit;
the input/output unit is used for acquiring input content of a user on the intelligent agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and converting a reply data stream fed back by the server into natural language output;
The model deployment unit is used for deploying the model conversion queue in the server, receiving the input data stream and outputting the reply data stream;
The proportional access unit is used for selecting a model on the model conversion queue according to the consistency of the input data stream and the characteristic data stream, and receiving reply data of the model in proportion.
Further, the project migration module comprises a pointer sliding unit, a model conversion unit and a proportion expansion unit;
the pointer sliding unit is used for judging the reply effect of the preamble reply data according to the feedback data flow of the user, and converting the reply model according to the height of the reply effect;
the model conversion unit is used for moving the directional pointer, adjusting the data sharing proportion of the reply model and starting or closing the model on the server;
The proportion expansion unit is used for calculating the adjustment quantity of the sharing proportion, so that the model selection mode is quickly matched with the user demand.
Further, the secondary development module comprises a weight output unit and a redeployment unit;
the weight output unit is used for outputting the service weight of each model in the service process after the user goes off line from the intelligent agent platform;
The redeployment unit is used for giving the development direction of the new model in the development period and redeploying the new model to the intelligent agent platform.
A zero code AI agent development and deployment data management method comprises the following steps:
S1, selecting an interface with the lowest time delay between development models as a conversion interface, generating a random data stream, performing data connectivity test on the conversion interface of the development model, and recording the data stream generated when the connectivity is highest as a characteristic data stream of the model;
s2, obtaining the consistency of characteristic data streams of all models by using a data stream comparison method, arranging all development models according to a consistency descending order, and connecting the models as a queue through a conversion interface to form a model conversion queue for deployment in a server;
S3, providing an input data stream for an intelligent agent by a user, calculating the consistency of the input data stream and each model characteristic data stream, taking a model with highest consistency as an initial model, sharing the input data in the initial model and a subsequent model of an initial model queue according to a consistency proportion, and processing to obtain output data;
S4, providing a feedback data stream for output data by a user, adjusting the sharing proportion of the input data of each model in the model conversion queue according to the consistency between the current feedback data stream and the historical feedback data stream, repeatedly executing until the sharing proportion of the initial model is lower than a threshold value, and replacing the initial model in a server;
And S5, accumulating the data sharing proportion, the sharing duration and the time attenuation coefficient of each model in the service process after the user is offline to obtain the service weight of the model, outputting the proportion of all the service weights of the model in the development period after the development period is finished, and redeploying the model.
Further, step S1 includes:
S11, performing ping instruction test on all interfaces for developing the model, acquiring feedback time delay of the interfaces, and taking the interface with the minimum time delay in the model as a conversion interface of the model, wherein the conversion interface is used for signal transmission and data sharing between the models;
S12, connecting every two models through a conversion interface, generating random data flow applicable to an intelligent agent by using natural language generating software, inputting the random data flow into a first model, obtaining output from a second model, calculating the ratio between the data quantity of output data and the data quantity of input data, and recording the connectivity of the first model;
S13, repeatedly generating a random data stream for a times, wherein a is preset test times, recording the random data stream generated when connectivity of each model is highest in the test process, and taking the random data stream as a characteristic data stream of the model.
Further, step S2 includes:
S21, calculating the consistency of the characteristic data streams of each model according to a data stream comparison method, wherein the data stream comparison method comprises the following steps:
Wherein Q represents the consistency of the characteristic data stream of the current model, n represents the number of models, m represents the data amount of the characteristic data stream, C1 j and C1 j+1 represent the j and j+1th data points in the characteristic data stream of the current model, and Ci j and Ci j+1 represent the j and j+1th data points in the characteristic data stream of the i-th model;
s22, arranging all models in a descending order according to the consistency coefficient, and linearly connecting the models through conversion interfaces in the models according to the arrangement order to form a model conversion queue;
S23, deploying the model conversion queue in a server, acquiring input content of a user on a front-end agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and outputting an output data stream fed back by the server through the agent platform after the input data is processed by the server.
Further, step S3 includes:
s31, the user interacts with the intelligent agent, the front end of the intelligent agent converts interaction input of the user into a data stream format and sends the data stream format to the server to serve as an input data stream of the user, and the consistency of the input data stream and characteristic data streams of all models is calculated, and the model with the highest consistency is used as an initial model;
S32, setting pointers pointing to initial models in a queue, selecting c models after the initial models in the queue as sharing models, wherein c is the maximum value of the inequality T1 +T2.+ Tc < TR, wherein T1, T2 and Tc respectively represent communication time delays between the initial models and the subsequent adjacent models, and TR is a preset time delay threshold;
S33, calculating an input data sharing proportion yc of each sharing model, wherein yc=qc/ΣQ, wherein Qc represents the consistency of the feature data stream of the sharing model and the input data stream, ΣQ represents the sum of the consistency of all the feature data streams of the model and the input data stream;
S34, splitting input data according to a proportion, respectively sending the split input data into a sharing model for processing, and summarizing the output data processed by each model by using a Cluster data integration tool to serve as feedback data to be sent to the front-end intelligent agent.
Further, step S4 includes:
S41, the front-end agent acquires feedback data of a user, and adjusts the data sharing proportion according to the following method:
S41-1, if no historical feedback data exists, turning to S41-2, and if the historical feedback data exists, turning to S41-3;
S41-2, calculating the consistency between the feedback data stream and the standard forward data stream, if the consistency is smaller than that of the initial model and the follow-up model in the queue, reducing the data sharing proportion of the initial model by a preset unit, distributing the reduced proportion in the rest sharing models, otherwise, increasing the data sharing proportion of the initial model by a preset unit, wherein the increased proportion is provided by the rest sharing models;
S41-3, calculating the consistency between the current feedback data stream and the last feedback data stream, if the consistency is smaller than that between the initial model and the subsequent model in the queue, reducing the data sharing proportion of the initial model by u units, wherein u= (QB-QN)/QE is adopted, QB represents the consistency between the current feedback data stream and the last feedback data stream, QN represents the average value of the consistency between the historical feedback data stream, QE represents the consistency between the initial model and the subsequent model in the queue, and otherwise, increasing the data sharing proportion of the initial model by u units;
Step S42, executing step S41 once when the feedback data stream of the user is obtained each time, and marking the model with the highest data sharing proportion in the rest sharing models as a new starting model if the data sharing proportion of the starting model is lower than a preset threshold after the execution is finished.
Further, step S5 includes:
S51, after interaction between the user and the intelligent agent is finished, calculating service weight of each model:
Wherein Rc represents the service weight of the model, k represents the user interaction times, yv represents the data sharing proportion of the model in the v-th interaction, tv represents the duration of the v-th interaction of the user, and h represents a preset time attenuation coefficient;
and S52, after the development period is finished, calculating the sum of service weights of all models in the development period, developing a new model according to the proportion of calculation results, and redeploying in the intelligent agent server.
Compared with the prior art, the invention has the following beneficial effects:
1. According to the invention, data connectivity tests can be carried out on various models by utilizing random data streams, the data streams corresponding to the highest data connectivity are used as characteristic data streams, the models are arranged according to the vector consistency of the characteristic data streams of the models, and the isolated models are connected with each other in a forward direction to generate more accurate output content by sharing the data, so that the data processing capability can be improved, and the accuracy, the robustness and the operation efficiency of the intelligent agent system are enhanced.
2. According to the method and the system for the data flow analysis, the data flow can be provided for the intelligent agent according to the user, the consistency of the input data flow and the characteristic data flow is analyzed, the input data flow is shared among the models through the conversion interface according to the consistency proportion, the type and the calling proportion of the models can be adjusted according to the user preference, the user experience can be improved, the output content of the intelligent agent is continuously optimized, and a good performance optimization curve is obtained in the service process of the intelligent agent.
3. After the user is offline, the connection weight of the model is output according to the current model access proportion, and the model with the highest connection weight is developed into a new model in each development period, so that the process of investigation by the user is omitted, the development progress of the intelligent body can be accelerated, the humanized interaction of the intelligent body is enhanced, and the high competitiveness and high efficiency of the development process are ensured.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a schematic diagram of a zero code AI agent development and deployment data management system of the present invention;
Fig. 2 is a schematic diagram of steps of a zero code AI agent development and deployment data management method of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to FIG. 1, the invention provides a technical scheme that a zero code AI intelligent agent development and deployment data management system comprises a data flow test module, a model communication module, a feedback processing module, a project migration module and a secondary development module;
The data flow test module is used for generating random vector data flows, performing data connectivity test on each calling interface between the development models, taking the data flow corresponding to the highest data connectivity as a characteristic data flow, testing the time delay of the calling interface between the models, and taking the interface with the lowest time delay as a conversion interface between the models;
The data flow testing module comprises a communication testing unit and a time delay testing unit;
The connectivity test unit is used for generating a random vector data stream, testing connectivity between every two models, and storing the data stream generated when the connectivity is highest;
the time delay test unit is used for testing the calling time delay of each interface between the models and marking the conversion interface with the lowest time delay.
The model communication module is used for calculating the consistency of the characteristic data streams of all models, arranging all development models according to a consistency descending order, connecting conversion interfaces among the models in a forward direction to form a model conversion queue, developing a matched processing program based on the model conversion queue, and deploying the program on a service platform;
the model communication module comprises a consistent arrangement unit and an interface connection unit;
the consistent arrangement unit is used for calculating the consistency of the characteristic data flow among the models and arranging all the models according to the consistency sequence;
The interface connection unit is used for connecting all models in one way through a conversion interface among the models to form a model conversion queue, and setting an initial pointer of the queue.
The feedback processing module is used for acquiring input data streams provided by users, calculating the consistency of the input data streams on the characteristic data streams of each model, selecting processing models on a model queue according to a consistency sequence, directing a directional pointer to the model with the largest consistency, enabling the sum of communication delay of a conversion interface between the models to be smaller than a preset threshold, sharing the input data streams according to a consistency proportion in the processing models, and outputting reply data streams to the users on a service platform through model processing;
The feedback processing module comprises an input/output unit, a model deployment unit and a proportional access unit;
the input/output unit is used for acquiring input content of a user on the intelligent agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and converting a reply data stream fed back by the server into natural language output;
The model deployment unit is used for deploying the model conversion queue in the server, receiving the input data stream and outputting the reply data stream;
The proportional access unit is used for selecting a model on the model conversion queue according to the consistency of the input data stream and the characteristic data stream, and receiving reply data of the model in proportion.
The project migration module is used for acquiring feedback data streams of users, when the consistency between the feedback data streams and the standard forward data streams is smaller than the consistency between the two model characteristic data streams, increasing the sharing data proportion of a unit of a subsequent model in the model conversion queue, otherwise, reducing the sharing data proportion of a unit, reacquiring the feedback data streams of the users until the sharing data proportion of the preamble model is reduced to a threshold value, closing a conversion interface of the preamble model, and moving a directional pointer by one bit;
The project migration module comprises a pointer sliding unit, a model conversion unit and a proportion expansion unit;
the pointer sliding unit is used for judging the reply effect of the preamble reply data according to the feedback data flow of the user, and converting the reply model according to the height of the reply effect;
the model conversion unit is used for moving the directional pointer, adjusting the data sharing proportion of the reply model and starting or closing the model on the server;
The proportion expansion unit is used for calculating the adjustment quantity of the sharing proportion, so that the model selection mode is quickly matched with the user demand.
The secondary development module is used for outputting the position of the directional pointer after the user goes offline and the data sharing proportion of each model in the model conversion queue in the service process, obtaining the service weight of each model according to the accumulated multiplication of the sharing proportion, the sharing duration and the time attenuation coefficient, developing a new model according to the service weight of the model after the development period is finished, and deploying the agent on the service platform again.
The secondary development module comprises a weight output unit and a redeployment unit;
the weight output unit is used for outputting the service weight of each model in the service process after the user goes off line from the intelligent agent platform;
The redeployment unit is used for giving the development direction of the new model in the development period and redeploying the new model to the intelligent agent platform.
As shown in fig. 2, a zero code AI agent development and deployment data management method includes the following steps:
S1, selecting an interface with the lowest time delay between development models as a conversion interface, generating a random data stream, performing data connectivity test on the conversion interface of the development model, and recording the data stream generated when the connectivity is highest as a characteristic data stream of the model;
the step S1 comprises the following steps:
S11, performing ping instruction test on all interfaces for developing the model, acquiring feedback time delay of the interfaces, and taking the interface with the minimum time delay in the model as a conversion interface of the model, wherein the conversion interface is used for signal transmission and data sharing between the models;
S12, connecting every two models through a conversion interface, generating random data flow applicable to an intelligent agent by using natural language generating software, inputting the random data flow into a first model, obtaining output from a second model, calculating the ratio between the data quantity of output data and the data quantity of input data, and recording the connectivity of the first model;
S13, repeatedly generating a random data stream for a times, wherein a is preset test times, recording the random data stream generated when connectivity of each model is highest in the test process, and taking the random data stream as a characteristic data stream of the model.
S2, obtaining the consistency of characteristic data streams of all models by using a data stream comparison method, arranging all development models according to a consistency descending order, and connecting the models as a queue through a conversion interface to form a model conversion queue for deployment in a server;
The step S2 comprises the following steps:
S21, calculating the consistency of the characteristic data streams of each model according to a data stream comparison method, wherein the data stream comparison method comprises the following steps:
Wherein Q represents the consistency of the characteristic data stream of the current model, n represents the number of models, m represents the data amount of the characteristic data stream, C1 j and C1 j+1 represent the j and j+1th data points in the characteristic data stream of the current model, and Ci j and Ci j+1 represent the j and j+1th data points in the characteristic data stream of the i-th model;
s22, arranging all models in a descending order according to the consistency coefficient, and linearly connecting the models through conversion interfaces in the models according to the arrangement order to form a model conversion queue;
S23, deploying the model conversion queue in a server, acquiring input content of a user on a front-end agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and outputting an output data stream fed back by the server through the agent platform after the input data is processed by the server.
S3, providing an input data stream for an intelligent agent by a user, calculating the consistency of the input data stream and each model characteristic data stream, taking a model with highest consistency as an initial model, sharing the input data in the initial model and a subsequent model of an initial model queue according to a consistency proportion, and processing to obtain output data;
The step S3 comprises the following steps:
s31, the user interacts with the intelligent agent, the front end of the intelligent agent converts interaction input of the user into a data stream format and sends the data stream format to the server to serve as an input data stream of the user, and the consistency of the input data stream and characteristic data streams of all models is calculated, and the model with the highest consistency is used as an initial model;
S32, setting pointers pointing to initial models in a queue, selecting c models after the initial models in the queue as sharing models, wherein c is the maximum value of the inequality T1 +T2.+ Tc < TR, wherein T1, T2 and Tc respectively represent communication time delays between the initial models and the subsequent adjacent models, and TR is a preset time delay threshold;
S33, calculating an input data sharing proportion yc of each sharing model, wherein yc=qc/ΣQ, wherein Qc represents the consistency of the feature data stream of the sharing model and the input data stream, ΣQ represents the sum of the consistency of all the feature data streams of the model and the input data stream;
S34, splitting input data according to a proportion, respectively sending the split input data into a sharing model for processing, and summarizing the output data processed by each model by using a Cluster data integration tool to serve as feedback data to be sent to the front-end intelligent agent.
S4, providing a feedback data stream for output data by a user, adjusting the sharing proportion of the input data of each model in the model conversion queue according to the consistency between the current feedback data stream and the historical feedback data stream, repeatedly executing until the sharing proportion of the initial model is lower than a threshold value, and replacing the initial model in a server;
The step S4 includes:
S41, the front-end agent acquires feedback data of a user, and adjusts the data sharing proportion according to the following method:
S41-1, if no historical feedback data exists, turning to S41-2, and if the historical feedback data exists, turning to S41-3;
S41-2, calculating the consistency between the feedback data stream and the standard forward data stream, if the consistency is smaller than that of the initial model and the follow-up model in the queue, reducing the data sharing proportion of the initial model by a preset unit, distributing the reduced proportion in the rest sharing models, otherwise, increasing the data sharing proportion of the initial model by a preset unit, wherein the increased proportion is provided by the rest sharing models;
S41-3, calculating the consistency between the current feedback data stream and the last feedback data stream, if the consistency is smaller than that between the initial model and the subsequent model in the queue, reducing the data sharing proportion of the initial model by u units, wherein u= (QB-QN)/QE is adopted, QB represents the consistency between the current feedback data stream and the last feedback data stream, QN represents the average value of the consistency between the historical feedback data stream, QE represents the consistency between the initial model and the subsequent model in the queue, and otherwise, increasing the data sharing proportion of the initial model by u units;
Step S42, executing step S41 once when the feedback data stream of the user is obtained each time, and marking the model with the highest data sharing proportion in the rest sharing models as a new starting model if the data sharing proportion of the starting model is lower than a preset threshold after the execution is finished.
And S5, accumulating the data sharing proportion, the sharing duration and the time attenuation coefficient of each model in the service process after the user is offline to obtain the service weight of the model, outputting the proportion of all the service weights of the model in the development period after the development period is finished, and redeploying the model.
The step S5 comprises the following steps:
S51, after interaction between the user and the intelligent agent is finished, calculating service weight of each model:
Wherein Rc represents the service weight of the model, k represents the user interaction times, yv represents the data sharing proportion of the model in the v-th interaction, tv represents the duration of the v-th interaction of the user, and h represents a preset time attenuation coefficient;
and S52, after the development period is finished, calculating the sum of service weights of all models in the development period, developing a new model according to the proportion of calculation results, and redeploying in the intelligent agent server.
In the embodiment, 3 models are arranged in the intelligent agent server, the characteristic data flows are [2,4,1,3,5], [1,4,2,2,1] and [5,3,3,1,4], the consistency is 0.18, 0.06 and 0.23 respectively, the models are arranged according to the sequence of the model 3, the model 1 and the model 2 to form a model queue, wherein the model 3 is used as a starting model, the sharing proportion obtained by a user according to the provided input data flow in the interaction process with the intelligent agent is 0.5, 0.3 and 0.2, and after the user feedback, the consistency of the feedback data flow is reduced, and the sharing proportion is redistributed to be 0.4, 0.36 and 0.24.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be noted that the above-mentioned embodiments are merely preferred embodiments of the present invention, and the present invention is not limited thereto, but may be modified or substituted for some of the technical features thereof by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A zero code AI agent development and deployment data management method, the method comprising the steps of:
S1, selecting an interface with the lowest time delay between development models as a conversion interface, generating a random data stream, performing data connectivity test on the conversion interface of the development model, and recording the data stream generated when the connectivity is highest as a characteristic data stream of the model;
S2, obtaining the consistency of characteristic data streams of all models by using a data stream comparison method, arranging all development models according to a consistency descending order, and connecting the models as a queue through a conversion interface to form a model conversion queue for deployment in a server;
S3, providing an input data stream for an intelligent agent by a user, calculating the consistency of the input data stream and each model characteristic data stream, taking a model with highest consistency as an initial model, sharing the input data in the initial model and a subsequent model of an initial model queue according to a consistency proportion, and processing to obtain output data;
S4, providing a feedback data stream for output data by a user, adjusting the sharing proportion of the input data of each model in the model conversion queue according to the consistency between the current feedback data stream and the historical feedback data stream, repeatedly executing until the sharing proportion of the initial model is lower than a threshold value, and replacing the initial model in a server;
And S5, accumulating the data sharing proportion, the sharing duration and the time attenuation coefficient of each model in the service process after the user is offline to obtain the service weight of the model, outputting the proportion of all the service weights of the model in the development period after the development period is finished, and redeploying the model.
2. The method for managing development and deployment data of a zero-code AI agent as set forth in claim 1, wherein the step S1 comprises:
S11, performing ping instruction test on all interfaces for developing the model, acquiring feedback time delay of the interfaces, and taking the interface with the minimum time delay in the model as a conversion interface of the model, wherein the conversion interface is used for signal transmission and data sharing between the models;
S12, connecting every two models through a conversion interface, generating random data flow applicable to an intelligent agent by using natural language generating software, inputting the random data flow into a first model, obtaining output from a second model, calculating the ratio between the data quantity of output data and the data quantity of input data, and recording the connectivity of the first model;
S13, repeatedly generating a random data stream for a times, wherein a is preset test times, recording the random data stream generated when connectivity of each model is highest in the test process, and taking the random data stream as a characteristic data stream of the model.
3. The method for managing development and deployment data of a zero-code AI agent as set forth in claim 2, wherein the step S2 comprises:
S21, calculating the consistency of the characteristic data streams of each model according to a data stream comparison method, wherein the data stream comparison method comprises the following steps:
Wherein Q represents the consistency of the characteristic data stream of the current model, n represents the number of models, m represents the data amount of the characteristic data stream, C1 j and C1 j+1 represent the j and j+1th data points in the characteristic data stream of the current model, and Ci j and Ci j+1 represent the j and j+1th data points in the characteristic data stream of the i-th model;
s22, arranging all models in a descending order according to the consistency coefficient, and linearly connecting the models through conversion interfaces in the models according to the arrangement order to form a model conversion queue;
S23, deploying the model conversion queue in a server, acquiring input content of a user on a front-end agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and outputting an output data stream fed back by the server through the agent platform after the input data is processed by the server.
4. The method for developing and deploying data of a zero-code AI agent of claim 3 wherein step S3 comprises:
s31, the user interacts with the intelligent agent, the front end of the intelligent agent converts interaction input of the user into a data stream format and sends the data stream format to the server to serve as an input data stream of the user, and the consistency of the input data stream and characteristic data streams of all models is calculated, and the model with the highest consistency is used as an initial model;
S32, setting pointers pointing to initial models in a queue, selecting c models after the initial models in the queue as sharing models, wherein c is the maximum value of the inequality T1 +T2.+ Tc < TR, wherein T1, T2 and Tc respectively represent communication time delays between the initial models and the subsequent adjacent models, and TR is a preset time delay threshold;
S33, calculating an input data sharing proportion yc of each sharing model, wherein yc=qc/ΣQ, wherein Qc represents the consistency of the feature data stream of the sharing model and the input data stream, ΣQ represents the sum of the consistency of all the feature data streams of the model and the input data stream;
S34, splitting input data according to a proportion, respectively sending the split input data into a sharing model for processing, and summarizing the output data processed by each model by using a Cluster data integration tool to serve as feedback data to be sent to the front-end intelligent agent.
5. The method for developing and deploying data management of a zero-code AI agent of claim 4 wherein step S4 comprises:
S41, the front-end agent acquires feedback data of a user, and adjusts the data sharing proportion according to the following method:
S41-1, if no historical feedback data exists, turning to S41-2, and if the historical feedback data exists, turning to S41-3;
S41-2, calculating the consistency between the feedback data stream and the standard forward data stream, if the consistency is smaller than that of the initial model and the follow-up model in the queue, reducing the data sharing proportion of the initial model by a preset unit, distributing the reduced proportion in the rest sharing models, otherwise, increasing the data sharing proportion of the initial model by a preset unit, wherein the increased proportion is provided by the rest sharing models;
S41-3, calculating the consistency between the current feedback data stream and the last feedback data stream, if the consistency is smaller than that between the initial model and the subsequent model in the queue, reducing the data sharing proportion of the initial model by u units, wherein u= (QB-QN)/QE is adopted, QB represents the consistency between the current feedback data stream and the last feedback data stream, QN represents the average value of the consistency between the historical feedback data stream, QE represents the consistency between the initial model and the subsequent model in the queue, and otherwise, increasing the data sharing proportion of the initial model by u units;
Step S42, executing the step S41 once when the feedback data stream of the user is obtained each time, and marking the model with the highest data sharing proportion in the rest sharing models as a new initial model if the data sharing proportion of the initial model is lower than a preset threshold value after the execution is finished;
The step S5 comprises the following steps:
S51, after interaction between the user and the intelligent agent is finished, calculating service weight of each model:
Wherein Rc represents the service weight of the model, k represents the user interaction times, yv represents the data sharing proportion of the model in the v-th interaction, tv represents the duration of the v-th interaction of the user, and h represents a preset time attenuation coefficient;
and S52, after the development period is finished, calculating the sum of service weights of all models in the development period, developing a new model according to the proportion of calculation results, and redeploying in the intelligent agent server.
6. A zero-code AI intelligent agent development and deployment data management system is characterized by comprising a data flow test module, a model communication module, a feedback processing module, a project migration module and a secondary development module;
The data flow test module is used for generating random vector data flows, performing data connectivity test on each calling interface between the development models, taking the data flow corresponding to the highest data connectivity as a characteristic data flow, testing the time delay of the calling interface between the models, and taking the interface with the lowest time delay as a conversion interface between the models;
the model communication module is used for calculating the consistency of the characteristic data streams of all models, arranging all development models according to a consistency descending order, connecting conversion interfaces among the models in a forward direction to form a model conversion queue, developing a matched processing program based on the model conversion queue, and deploying the program on a service platform;
The feedback processing module is used for acquiring input data streams provided by users, calculating the consistency of the input data streams on the characteristic data streams of each model, selecting processing models on a model queue according to a consistency sequence, directing a directional pointer to the model with the largest consistency, enabling the sum of communication delay of a conversion interface between the models to be smaller than a preset threshold, sharing the input data streams according to a consistency proportion in the processing models, and outputting reply data streams to the users on a service platform through model processing;
The project migration module is used for acquiring feedback data streams of users, when the consistency between the feedback data streams and the standard forward data streams is smaller than the consistency between the two model characteristic data streams, increasing the sharing data proportion of a unit of a subsequent model in the model conversion queue, otherwise, reducing the sharing data proportion of a unit, reacquiring the feedback data streams of the users until the sharing data proportion of the preamble model is reduced to a threshold value, closing a conversion interface of the preamble model, and moving a directional pointer by one bit;
The secondary development module is used for outputting the position of the directional pointer after the user goes offline and the data sharing proportion of each model in the model conversion queue in the service process, obtaining the service weight of each model according to the accumulated multiplication of the sharing proportion, the sharing duration and the time attenuation coefficient, developing a new model according to the service weight of the model after the development period is finished, and deploying the agent on the service platform again.
7. The system for developing and deploying data of a zero-code AI agent of claim 6 wherein the data flow test module comprises a connectivity test unit and a latency test unit;
The connectivity test unit is used for generating a random vector data stream, testing connectivity between every two models, and storing the data stream generated when the connectivity is highest;
The time delay test unit is used for testing the calling time delay of each interface between the models and marking the conversion interface with the lowest time delay;
the model communication module comprises a consistent arrangement unit and an interface connection unit;
the consistent arrangement unit is used for calculating the consistency of the characteristic data flow among the models and arranging all the models according to the consistency sequence;
The interface connection unit is used for connecting all models in one way through a conversion interface among the models to form a model conversion queue, and setting an initial pointer of the queue.
8. The system for managing development and deployment data of zero-code AI agent of claim 7, wherein said feedback processing module comprises an input/output unit, a model deployment unit, and a proportional access unit;
the input/output unit is used for acquiring input content of a user on the intelligent agent platform, converting the input content into an input data stream, uploading the input data stream to the server, and converting a reply data stream fed back by the server into natural language output;
The model deployment unit is used for deploying the model conversion queue in the server, receiving the input data stream and outputting the reply data stream;
The proportional access unit is used for selecting a model on the model conversion queue according to the consistency of the input data stream and the characteristic data stream, and receiving reply data of the model in proportion.
9. The system for developing and deploying data of a zero-code AI agent of claim 8 wherein the project migration module comprises a pointer sliding unit, a model conversion unit and a scaling unit;
the pointer sliding unit is used for judging the reply effect of the preamble reply data according to the feedback data flow of the user, and converting the reply model according to the height of the reply effect;
the model conversion unit is used for moving the directional pointer, adjusting the data sharing proportion of the reply model and starting or closing the model on the server;
The proportion expansion unit is used for calculating the adjustment quantity of the sharing proportion, so that the model selection mode is quickly matched with the user demand.
10. The system for managing development and deployment data of zero-code AI agent of claim 9, wherein said secondary development module comprises a weight output unit and a redeployment unit;
the weight output unit is used for outputting the service weight of each model in the service process after the user goes off line from the intelligent agent platform;
The redeployment unit is used for giving the development direction of the new model in the development period and redeploying the new model to the intelligent agent platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411765389.9A CN119668593B (en) | 2024-12-04 | 2024-12-04 | Zero-code AI intelligent agent development and deployment data management system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411765389.9A CN119668593B (en) | 2024-12-04 | 2024-12-04 | Zero-code AI intelligent agent development and deployment data management system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119668593A true CN119668593A (en) | 2025-03-21 |
| CN119668593B CN119668593B (en) | 2025-06-24 |
Family
ID=94985302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411765389.9A Active CN119668593B (en) | 2024-12-04 | 2024-12-04 | Zero-code AI intelligent agent development and deployment data management system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119668593B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825362A (en) * | 2019-11-04 | 2020-02-21 | 广东道一信息技术股份有限公司 | Low-code application software development system and method |
| CN112685514A (en) * | 2021-01-08 | 2021-04-20 | 北京云桥智联科技有限公司 | AI intelligent customer value management platform |
| CN112947946A (en) * | 2021-02-24 | 2021-06-11 | 上海后米物联网技术有限公司 | Service, data and model deployment, sharing and distribution system for industrial field |
| WO2022222537A1 (en) * | 2021-04-24 | 2022-10-27 | 华为云计算技术有限公司 | Ai application deployment method, and related platform, cluster, medium and program product |
| WO2023103190A1 (en) * | 2021-12-06 | 2023-06-15 | 福建亿榕信息技术有限公司 | Multi-level linkage transparent sample model sharing apparatus for artificial intelligence platform |
| US20230315055A1 (en) * | 2022-04-01 | 2023-10-05 | Open Text Holdings, Inc. | Zero code stream processing engine for machine interface |
| CN118276873A (en) * | 2024-04-09 | 2024-07-02 | 深圳市云速易连科技有限公司 | Low-code development method based on low-code development platform |
-
2024
- 2024-12-04 CN CN202411765389.9A patent/CN119668593B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825362A (en) * | 2019-11-04 | 2020-02-21 | 广东道一信息技术股份有限公司 | Low-code application software development system and method |
| CN112685514A (en) * | 2021-01-08 | 2021-04-20 | 北京云桥智联科技有限公司 | AI intelligent customer value management platform |
| CN112947946A (en) * | 2021-02-24 | 2021-06-11 | 上海后米物联网技术有限公司 | Service, data and model deployment, sharing and distribution system for industrial field |
| WO2022222537A1 (en) * | 2021-04-24 | 2022-10-27 | 华为云计算技术有限公司 | Ai application deployment method, and related platform, cluster, medium and program product |
| WO2023103190A1 (en) * | 2021-12-06 | 2023-06-15 | 福建亿榕信息技术有限公司 | Multi-level linkage transparent sample model sharing apparatus for artificial intelligence platform |
| US20230315055A1 (en) * | 2022-04-01 | 2023-10-05 | Open Text Holdings, Inc. | Zero code stream processing engine for machine interface |
| CN118276873A (en) * | 2024-04-09 | 2024-07-02 | 深圳市云速易连科技有限公司 | Low-code development method based on low-code development platform |
Non-Patent Citations (1)
| Title |
|---|
| 张巍: "基于钉钉宜搭的中小企业低代码平台的应用研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 January 2024 (2024-01-15), pages 138 - 482 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119668593B (en) | 2025-06-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112445902B (en) | A method for identifying user intention in multi-round dialogue and related equipment | |
| CN113902116B (en) | A batch processing optimization method and system for deep learning model reasoning | |
| US20050278439A1 (en) | System and method for evaluating capacity of a heterogeneous media server configuration for supporting an expected workload | |
| CN113840007A (en) | Load balancing method and device | |
| CN110138883A (en) | Mixed cloud resource allocation methods and device | |
| WO2023166928A1 (en) | Power consumption estimation device, power consumption estimation system, and power consumption estimation method | |
| CN119668593B (en) | Zero-code AI intelligent agent development and deployment data management system and method | |
| KR20230087316A (en) | Apparatus and method for determining ai-based cloud service server | |
| CN108376369A (en) | Finance and money management Products Show method, equipment, system and storage medium | |
| US20050004788A1 (en) | Multi-level confidence measures for task modeling and its application to task-oriented multi-modal dialog management | |
| CN111209399A (en) | Text classification method and device and electronic equipment | |
| CN111753560B (en) | Self-adaptive translation system in parallel multi-source input mode | |
| CN113391795B (en) | A method and system for implementing adaptive mapping between application scenarios and software development kits | |
| KR102713221B1 (en) | Customized call center operation system through voice analysis-based consultation type classification | |
| Huang | The value-of-information in matching with queues | |
| CN117194047B (en) | Distributed system based on data collaboration | |
| CN101673281A (en) | Method for making travel plan | |
| CN111107003A (en) | Intelligent routing method | |
| JP6948849B2 (en) | AI service usage support system | |
| CN107766944B (en) | System and method for optimizing system function flow by utilizing API analysis | |
| Shestak et al. | Greedy Approaches to Static Stochastic Robust Resource Allocation for Periodic Sensor Driven Distributed Systems. | |
| CN117196011A (en) | Distributed network data transmission method and device | |
| CN111897618B (en) | Method and device for determining UI interface and storage medium | |
| US10873550B2 (en) | Methods for communication in a communication network for reduced data traffic | |
| Caduff et al. | Geo-Mobile query-by-sketch |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |