Sample retrieval and user interface generator
A maintenance and monitoring infrastructure can use a plurality of sensors or monitors in an industrial plant to obtain machine samples from a large collection of assets (industrial machines). An artificial-intelligence (AI) assistant can conversationally interact with a technician to answer questions regarding maintenance status of the assets in the plant. The technician can conversationally request various machine samples and/or associated interactive user interfaces to examine and analyze the samples in relation to maintenance and monitoring tasks. The AI assistant can identify and retrieve relevant data and in real-time build instructions for generating user interfaces responsive to the technician's conversational requests.
1 . A method comprising:
an upstream remote maintenance and monitoring infrastructure, providing a plurality of removable and portable monitors in an industrial plant, each monitor comprising: a wireless communication module, a microcontroller, an accelerometer, and a battery module, each monitor, removably attachable to a machine in the plant and configured to collect machine samples from an attached machine in the plant, the monitor wirelessly communicating the samples to a receiver in the plant, the receiver transmitting the samples to one or more data servers in the remote maintenance and monitoring infrastructure, the samples having associated attributes;
wherein the remote maintenance and monitoring infrastructure, comprises a primary LLM agent and one or more secondary LLM agents,
instructing the primary LLM agent on an asset tree of the plant, the asset tree comprising asset identifiers, each asset identifier uniquely corresponding to a distinct machine in the plant;
instructing the primary LLM agent on availability of one or more secondary LLM agents and functionality, input and output specification of the secondary LLM agents;
instructing each secondary LLM agent on performing a task, wherein performing the task at least in part comprises interfacing with the one or more data servers of the maintenance and monitoring infrastructure and retrieving sample IDs from the data servers;
receiving, by the primary LLM agent, a conversational user request, comprising a conversational request for samples of a machine and/or multiple machines for a time period;
the primary LLM agent translating the conversational request to asset identifiers of one or more machines in the plant, based on the asset tree of the plant;
the primary LLM agent, selecting, based on the conversational request, a secondary LLM agent previously instructed in performing sample ID retrieval of the conversational request;
the primary LLM agent, providing the translated asset identifiers to the selected secondary agent;
the selected secondary LLM agent, retrieving sample IDs corresponding to the translated asset identifier;
the selected secondary LLM agent, providing the retrieved sample IDs to the primary LLM agent;
the primary LLM agent, generating instructions that when executed perform operations comprising:
interfacing with the data servers and obtaining values of samples corresponding to the retrieved sample IDs; and
generating a visualization of the sample values corresponding to the sample IDs in a frontend user interface of the maintenance and monitoring infrastructure.
2 . The method of claim 1 , wherein the sample attributes comprise one or more of sample identifiers, sample time stamps, and sample type.
3 . The method of claim 1 , wherein the secondary LLM agents comprise:
a multiple asset comparison agent, instructed to obtain, from the data servers, sample IDs of two or more machines for a selected time period, comprising a recent time period.
4 . The method of claim 1 , wherein the secondary LLM agents comprise:
a historical sample agent, instructed to obtain, from the data servers, sample IDs of a machine for a time frame, the time frame approximated based on available samples in the data servers and the time period received in the conversational request.
5 . The method of claim 1 , wherein the one or more data servers comprise endpoints configured to receive an asset identifier and a time period and return sample IDs of samples having the received asset identifier and corresponding to the received time period, wherein the selected secondary LLM agent is configured to retrieve the sample IDs by interfacing with the endpoints.
6 . The method of claim 1 , wherein generating the user interfaces comprises:
instructing the primary LLM agent with a uniform resource locator (URL) map rule, the URL map rule comprising example URLs and URL formatting rules, the URL formatting rules comprising asset identifier and sample ID locations in the example URLs, wherein executing the URL generates computer display instructions, generating the user interfaces visualizing the sample values.
7 . The method of claim 1 , wherein the primary LLM agent is further instructed on performing operations comprising:
identifying a task from the conversational request;
generating computer system instructions corresponding to the identified task;
communicating the computer instructions to one or more computer systems within the maintenance and monitoring infrastructure.
8 . A non-transitory computer storage that stores executable program instructions that, when executed by one or more computing devices, configure the one or more computing devices to perform operations comprising:
in an upstream remote maintenance and monitoring infrastructure, providing a plurality of removable and portable monitors in an industrial plant, each monitor comprising: a wireless communication module, a microcontroller, an accelerometer, and a battery module, each monitor, removably attachable to a machine in the plant and configured to collect machine samples from an attached machine in the plant, the monitor wirelessly communicating the samples to a receiver in the plant, the receiver transmitting the samples to one or more data servers in the remote maintenance and monitoring infrastructure, the samples having associated attributes;
wherein the remote maintenance and monitoring infrastructure, comprises a primary LLM agent and one or more secondary LLM agents,
instructing the primary LLM agent on an asset tree of the plant, the asset tree comprising asset identifiers, each asset identifier uniquely corresponding to a distinct machine in the plant;
instructing the primary LLM agent on availability of one or more secondary LLM agents and functionality, input and output specification of the secondary LLM agents;
instructing each secondary LLM agent on performing a task, wherein performing the task at least in part comprises interfacing with the one or more data servers of the maintenance and monitoring infrastructure and retrieving sample IDs from the data servers;
receiving, by the primary LLM agent, a conversational user request, comprising a conversational request for samples of a machine and/or multiple machines for a time period;
the primary LLM agent translating the conversational request to asset identifiers of one or more machines in the plant, based on the asset tree of the plant;
the primary LLM agent, selecting, based on the conversational request, a secondary LLM agent previously instructed in performing sample IDs retrieval of the conversational request;
the primary LLM agent, providing the translated asset identifiers to the selected secondary agent;
the selected secondary LLM agent, retrieving sample IDs corresponding to the translated asset identifier;
the selected secondary LLM agent, providing the retrieved sample IDs to the primary LLM agent;
the primary LLM agent, generating instructions that when executed perform operations comprising:
interfacing with the data servers and obtaining values of samples corresponding to the retrieved sample IDs; and
generating a visualization of the sample values corresponding to the sample IDs in a frontend user interface of the maintenance and monitoring infrastructure.
9 . The non-transitory computer storage of claim 8 , wherein the sample attributes comprise one or more of sample identifiers, sample time stamps, and sample type.
10 . The non-transitory computer storage of claim 8 , wherein the secondary LLM agents comprise:
a multiple asset comparison agent, instructed to obtain, from the data servers, sample IDs of two or more machines for a selected time period, comprising a recent time period.
11 . The non-transitory computer storage of claim 8 , wherein the secondary LLM agents comprise:
a historical sample agent, instructed to obtain, from the data servers, sample IDs of a machine for a time frame, the time frame approximated based on available samples in the data servers and the time period received in the conversational request.
12 . The non-transitory computer storage of claim 8 , wherein the one or more data servers comprise endpoints configured to receive an asset identifier and a time period and return sample IDs of samples having the received asset identifier and the time period, wherein the selected secondary LLM agent is configured to retrieve the sample IDs by interfacing with the endpoints.
13 . The non-transitory computer storage of claim 8 , wherein generating the user interfaces comprises:
instructing the primary LLM agent with a uniform resource locator (URL) map rule, the URL map rule comprising example URLs and URL formatting rules, the URL formatting rules comprising asset identifier and sample ID locations in the example URLs, wherein executing the URL generates computer display instructions, generating the user interfaces visualizing the sample values.
14 . The non-transitory computer storage of claim 8 , wherein the primary LLM agent is further instructed on performing operations comprising:
identifying a task from the conversational request;
generating computer system instructions corresponding to the identified task;
communicating the computer instructions to one or more computer systems within the maintenance and monitoring infrastructure.
15 . A system comprising one or more processors, wherein the one or more processors are configured to perform operations comprising:
an upstream remote maintenance and monitoring infrastructure, providing a plurality of removable and portable monitors in an industrial plant, each monitor comprising: a wireless communication module, a microcontroller, an accelerometer, and a battery module, each monitor, removably attachable to a machine in the plant and configured to collect machine samples from an attached machine in the plant, the monitor wirelessly communicating the samples to a receiver in the plant, the receiver transmitting the samples to one or more data servers in the remote maintenance and monitoring infrastructure, the samples having associated attributes;
wherein the remote maintenance and monitoring infrastructure, comprises a primary LLM agent and one or more secondary LLM agents,
instructing the primary LLM agent on an asset tree of the plant, the asset tree comprising asset identifiers, each asset identifier uniquely corresponding to a distinct machine in the plant;
instructing the primary LLM agent on availability of one or more secondary LLM agents and functionality, input and out specification of the secondary LLM agents;
instructing each secondary LLM agent on performing a task, wherein performing the task at least in part comprises interfacing with the one or more data servers of the maintenance and monitoring infrastructure and retrieving sample IDs from the data servers;
receiving, by the primary LLM agent, a conversational user request, comprising a conversational request for samples of a machine and/or multiple machines for a time period;
the primary LLM agent translating the conversational request to asset identifiers of one or more machines in the plant, based on the asset tree of the plant;
the primary LLM agent, selecting, based on the conversational request, a secondary LLM agent previously instructed in performing sample ID retrieval of the conversational request;
the primary LLM agent, providing the translated asset identifiers to the selected secondary agent;
the selected secondary LLM agent, retrieving sample IDs corresponding to the translated asset identifier;
the selected secondary LLM agent, providing the retrieved sample IDs to the primary LLM agent;
the primary LLM agent, generating instructions that when executed perform operations comprising:
interfacing with the data servers and obtaining values of samples corresponding to the retrieved sample IDs; and
generating a visualization of the sample values corresponding to the sample IDs in a frontend user interface of the maintenance and monitoring infrastructure.
16 . The system of claim 15 , wherein the sample attributes comprise one or more of sample identifiers, sample time stamps, and sample type.
17 . The system of claim 15 , wherein the secondary LLM agents comprise:
a multiple asset comparison agent, instructed to obtain, from the data servers, sample IDs of two or more machines for a selected time period, comprising a recent time period.
18 . The system of claim 15 , wherein the secondary LLM agents comprise:
a historical sample agent, instructed to obtain, from the data servers, sample IDs of a machine for a time frame, the time frame approximated based on available samples in the data servers and the time period received in the conversational request.
19 . The system of claim 15 , wherein the one or more data servers comprise endpoints configured to receive an asset identifier and a time period and return sample IDs of samples having the received asset identifier and the time period, wherein the selected secondary LLM agent is configured to retrieve the sample IDs by interfacing with the endpoints.
20 . The system of claim 15 , wherein generating the user interfaces comprises:
instructing the primary LLM agent with a uniform resource locator (URL) map rule, the URL map rule comprising example URLs and URL formatting rules, the URL formatting rules comprising asset identifier and sample ID locations in the example URLs, wherein executing the URL generates computer display instructions, generating the user interfaces visualizing the sample values.