System and method for creating, searching and executing artificial intelligence agents

1 . A system comprising: a first set of one or more processors and a first memory coupled to the first set of processors comprising instructions executable by the first set of processors, the first set of processors being operable when executing the instructions to:

create a data storage service, a search engine module, an admin module, a creation module, and an execution module

and the admin module to process a request to create an authorization for a namespace comprising:

receiving a namespace;

generating a Universally Unique Identifier (UUID) as an authorization key; and

storing the namespace and authorization key in the data storage service

and the creation module to process a request to create an artificial intelligence agent (agent), comprising:

receiving a namespace and an agent specification wherein the agent specification is in a machine-readable, textual format comprising:

an agent name;

an agent title comprising a formatted string literal (f-string);

an ordered list of one or more task units wherein each task unit comprises:

a task name; and

an ordered list of one or more command units wherein each command unit comprises:

 a command type; and

 a command spec comprising an f-string that evaluates to a code block

and storing the namespace, agent name and agent specification in the data storage service

and the search engine module to process a request to index an agent, comprising:

receiving a namespace and an agent descriptor wherein the agent descriptor comprises:

an agent name;

a description of the agent;

one or more filters wherein each filter comprises a name and a value;

one or more classes in one or more classification schemes;

one or more keywords;

one or more named lists;

one or more string patterns;

one or more example queries; and

one or more natural language entity fields wherein each entity field comprises:

an entity type;

a minimum and a maximum number of occurrences of the entity type; and

one or more classes in a classification scheme that a value of the entity type can have

and indexing the agent descriptor and mapping it to the agent name

and the search engine module to process a request to search for agents comprising:

receiving a namespace, a search query and one or more inputs wherein each input comprises a key and a value;

matching the namespace, search query and inputs to one or more agent names; and

returning the matched agent names and titles

and the execution module to process a request to search for agents comprising:

receiving an authorization key, a search query and one or more inputs wherein each input comprises a key and a value;

reading a namespace associated with the authorization key in the data storage service;

forwarding the namespace and the request to the search engine module and receiving matched agents and titles; and

returning the matched agents and titles

and the execution module to process a request to retrieve an agent specification comprising:

receiving an authorization key and an agent name;

reading a namespace associated with the authorization key in the data storage service;

reading an agent specification associated with the namespace and agent name in the data storage service; and

returning the agent specification.

2 . The system of claim 1 additionally comprising: a second set of one or more processors and a second memory coupled to the second set of processors comprising instructions executable by the second set of processors, the second set of processors being operable when executing the instructions to:

create a client module and the client module reading an authorization key

and the client module to process a search request comprising:

receiving a search query and one or more inputs wherein each input comprises a name and a value;

sending the authorization key, search query and inputs to the execution module and receiving matched agent names and titles;

selecting an agent name that occurs first in matched agent names and titles;

sending a request to the execution module to retrieve an agent specification with the selected agent name and receiving a selected agent specification;

creating an agent state object in the second memory;

adding the search query, inputs, matched agent names and titles to the agent state object;

executing the selected agent specification comprising:

executing a task unit that occurs first in the ordered list of task units in the selected agent specification comprising:

executing each command unit in the task unit sequentially in the order they occur in the task unit wherein executing a command unit comprises:

 executing the command spec comprising:

 evaluating the f-string specified in the command unit to a first code block; and

 executing the first code block

and returning the agent state object.

3 . The system of claim 2 wherein a command unit additionally specifies a variable unit comprising:

an ordered list of one or more variables wherein each variable comprises:

a variable name; and

a value specified by a f-string

and the client module executes the command unit by additionally executing the variable unit comprising:

evaluating each variable in the variable unit in the order specified in the variable unit wherein evaluating a variable comprises:

evaluating the value specified by a f-string to a first value; and

storing the first value in the agent state object keyed by the variable name.

4 . The system of claim 2 wherein a command spec alternately specifies:

an f-string that specifies an Application Programming Interface (API) service endpoint;

a protocol type and protocol subtype used to communicate with the API service;

one or more pairs of names and f-strings specifying headers included in request to the API service;

one or more pairs of names and f-strings specifying parameters included in request to the API service; and

a response name

and the client module alternately executes the command spec comprising:

evaluating all f-strings in the API service endpoint, headers and parameters;

using the protocol type and protocol subtype to send a request to a server at the API service endpoint with evaluated values of f-strings in the API service endpoint, headers and parameters and receiving a response; and

storing the response in the agent state object keyed by the response name.

5 . The system of claim 2 wherein a command spec alternately specifies a conditional execution unit comprising:

a f-string that evaluates to either true or false;

a loop value that evaluates to either true or false;

a first list of ordered commands; and

a second list of ordered commands

and the client module alternately executes the command spec comprising:

executing a first step comprising evaluating the f-string to a first value;

executing a second step comprising:

if the first value is true, executing the first list of commands in the order specified; and

if the first value is false, executing the second list of commands in the order specified and ending the execution of command spec; and

if the loop value is true, repeating the first step and second step until the first value evaluates to false; and

if the loop value is false, ending the execution of the command spec.

6 . The system of claim 2 wherein a command spec alternately specifies an agent state object to be stored, and the client module alternately executes the command spec comprising:

creating an object name by generating a UUID;

serializing the agent state object; and

sending a request to the execution module to store an object with the authorization key, object name and serialized agent state object.

7 . The system of claim 2 wherein a command spec alternately specifies that an agent state object be restored, and the client module alternately executes the command spec comprising:

sending a request to the execution module to retrieve an object with the authorization key and an object name and receiving an object;

deserializing the received object; and

restoring the agent state object from the deserialized object.

8 . The system of claim 2 wherein a command spec alternately specifies a list of one or more additional inputs required to continue executing an agent wherein an additional input comprises an input name and a hint, and the client module alternately executes the command spec comprising storing the list of additional inputs in the agent state object.

9 . The system of claim 2 wherein the execution module additionally handles a request to store an event comprising:

receiving an authorization key, an event name and a list of parameters wherein each parameter contains a name and a value;

reading a namespace associated with the authorization key in the data storage service; and

storing the namespace, event name and the list of parameters in the data storage service

and a command spec alternately specifies an event name and a list of parameters, and the client module alternately executes the command spec by sending a request to the execution module to store an event with the authorization key, the event name and the list of parameters.

10 . The system of claim 2 wherein the execution module additionally handles a request to call a Large Language Model (LLM) comprising:

receiving an authorization key, a model name, an input and a list of parameters wherein each parameter comprises a name and a value;

reading a namespace associated with the authorization key in the data storage service; and

calling a first LLM service with the model name and input and receiving an output; and

returning the output

and a command spec alternately specifies:

an LLM model name;

an f-string that specifies an input;

a list of parameters; and

a response name

and the client module alternately executes the command spec comprising:

evaluating the f-string to an input value;

sending a request to the execution module to call an LLM model comprising the authorization key, LLM model name, evaluated input value and the list of parameters and receiving a value; and

storing the value in the agent state object keyed by value of the response name.

11 . A method comprising, by a first computing system:

creating a data storage service, a search engine module, an admin module, a creation module, and an execution module

and the admin module processing a request to create an authorization for a namespace comprising:

receiving a namespace;

generating a Universally Unique Identifier (UUID) as an authorization key; and

storing the namespace and authorization key in the data storage service

and the creation module processing a request to create an artificial intelligence agent (agent), comprising:

receiving a namespace and an agent specification wherein the agent specification is in a machine-readable, textual format comprising:

an agent name;

an agent title comprising a formatted string literal (f-string);

an ordered list of one or more task units wherein each task unit comprises:

a task name; and

an ordered list of one or more command units wherein each command unit comprises:

 a command type; and

 a command spec comprising an f-string that evaluates to a code block

and storing the namespace, agent name and agent specification in the data storage service

and the search engine module processing a request to index an agent, comprising:

receiving a namespace and an agent descriptor wherein the agent descriptor comprises:

an agent name;

a description of the agent;

one or more filters wherein each filter comprises a name and a value;

one or more classes in one or more classification schemes;

one or more keywords;

one or more named lists;

one or more string patterns;

one or more example queries; and

one or more natural language entity fields wherein each entity field comprises:

an entity type;

a minimum and a maximum number of occurrences of the entity type; and

one or more classes in a classification scheme that a value of the entity type can have

and indexing the agent descriptor and mapping it to the agent name

and the search engine module processing a request to search for agents comprising:

receiving a namespace, a search query and one or more inputs wherein each input comprises a key and a value;

matching the namespace, search query and inputs to one or more agent names; and

returning the matched agent names and titles

and the execution module processing a request to search for agents comprising:

receiving an authorization key, a search query and one or more inputs wherein each input comprises a key and a value;

reading a namespace associated with the authorization key in the data storage service;

forwarding the namespace and the request to the search engine module and receiving matched agents and titles; and

returning the matched agents and titles

and the execution module processing a request to retrieve an agent specification comprising:

receiving an authorization key and an agent name;

reading a namespace associated with the authorization key in the data storage service;

reading an agent specification associated with the namespace and agent name in the data storage service; and

returning the agent specification.

12 . The method of claim 11 , further comprising a second computing system:

creating a client module and the client module reading an authorization key

and the client module processing a search request comprising:

receiving a search query and one or more inputs wherein each input comprises a name and a value;

sending the authorization key, search query and inputs to the execution module and receiving matched agent names and titles;

selecting an agent name that occurs first in matched agent names and titles;

sending a request to the execution module to retrieve an agent specification with the selected agent name and receiving a selected agent specification;

creating an agent state object in the second memory;

adding the search query, inputs, matched agent names and titles to the agent state object;

executing the selected agent specification comprising:

executing a task unit that occurs first in the ordered list of task units in the selected agent specification comprising:

executing each command unit in the task unit sequentially in the order they occur in the task unit wherein executing a command unit comprises:

 executing the command spec comprising:

 evaluating the f-string specified in the command unit to a first code block; and

 executing the first code block

and returning the agent state object.

13 . The method of claim 12 , wherein a command unit additionally specifying a variable unit comprising:

an ordered list of one or more variables wherein each variable comprises:

a variable name; and

a value specified by a f-string

and the client module executing the command unit by additionally executing the variable unit comprising:

evaluating each variable in the variable unit in the order specified in the variable unit wherein evaluating a variable comprises:

evaluating the value specified by a f-string to a first value; and

storing the first value in the agent state object keyed by the variable name.

14 . The method of claim 12 , wherein a command spec alternately specifying:

an f-string that specifies an Application Programming Interface (API) service endpoint;

a protocol type and protocol subtype used to communicate with the API service;

one or more pairs of names and f-strings specifying headers included in request to the API service;

one or more pairs of names and f-strings specifying parameters included in request to the API service; and

a response name

and the client module alternately executing the command spec comprising:

evaluating all f-strings in the API service endpoint, headers and parameters;

using the protocol type and protocol subtype to send a request to a server at the API service endpoint with evaluated values of f-strings in the API service endpoint, headers and parameters and receiving a response; and

storing the response in the agent state object keyed by the response name.

15 . The method of claim 12 , wherein a command spec alternately specifying a conditional execution unit comprising:

a f-string that evaluates to either true or false;

a loop value that evaluates to either true or false;

a first list of ordered commands; and

a second list of ordered commands

and the client module alternately executing the command spec comprising:

executing a first step comprising evaluating the f-string to a first value;

executing a second step comprising:

if the first value is true, executing the first list of commands in the order specified; and

if the first value is false, executing the second list of commands in the order specified and ending the execution of command spec; and

if the loop value is true, repeating the first step and second step until the first value evaluates to false; and

if the loop value is false, ending the execution of the command spec.

16 . The method of claim 12 , wherein a command spec alternately specifying an agent state object to be stored, and the client module alternately executing the command spec comprising:

creating an object name by generating a UUID;

serializing the agent state object; and

sending a request to the execution module to store an object with the authorization key, object name and serialized agent state object.

17 . The method of claim 12 , wherein a command spec alternately specifying that an agent state object be restored, and the client module alternately executing the command spec comprising:

sending a request to the execution module to retrieve an object with the authorization key and an object name and receiving an object;

deserializing the received object; and

restoring the agent state object from the deserialized object.

18 . The method of claim 12 , wherein a command spec alternately specifying a list of one or more additional inputs required to continue executing an agent wherein an additional input comprises an input name and a hint, and the client module alternately executing the command spec comprising storing the list of additional inputs in the agent state object.

19 . The method of claim 12 , wherein the execution module additionally handling a request to store an event comprising:

receiving an authorization key, an event name and a list of parameters wherein each parameter contains a name and a value;

reading a namespace associated with the authorization key in the data storage service; and

storing the namespace, event name and the list of parameters in the data storage service

and a command spec alternately specifying an event name and a list of parameters, and the client module alternately executing the command spec by sending a request to the execution module to store an event with the authorization key, the event name and the list of parameters.

20 . The method of claim 12 , wherein the execution module additionally handling a request to call a Large Language Model (LLM) comprising:

receiving an authorization key, a model name, an input and a list of parameters wherein each parameter comprises a name and a value;

reading a namespace associated with the authorization key in the data storage service; and

calling a first LLM service with the model name and input and receiving an output; and

returning the output

and a command spec alternately specifying:

an LLM model name;

an f-string that specifies an input;

a list of parameters; and

a response name

and the client module alternately executing the command spec comprising:

evaluating the f-string to an input value;

sending a request to the execution module to call an LLM model comprising the authorization key, LLM model name, evaluated input value and the list of parameters and receiving a value; and

storing the value in the agent state object keyed by value of the response name.