Converting human-readable queries to machine-readable queries

1 . A device, comprising:

one or more processors and memory storing instructions that, when executed by the one or more processors, cause the device to:

receive, via a user interface, a natural language prompt from a user;

generate, by a large language model (LLM), a vector representation of the natural language prompt;

retrieve, from a vector database, schema information associated with a monitoring database, the schema information including vector embeddings of database entities;

convert, by the LLM and using the vector representation of the prompt and the vector embeddings of the schema information, the natural language prompt into a structured query language (SQL) query;

execute the SQL query against the monitoring database;

receive monitoring results from the monitoring database based on executing the SQL query;

detect, by the device, an execution error resulting from a schema mismatch or schema change in the monitoring database;

in response to detecting the execution error, execute a schema query to generate updated schema information from the monitoring database, and store the updated schema information in the vector database;

regenerate, by the LLM, the SQL query using the updated schema information; and

convert the monitoring results into a natural language response, using the LLM, for output to the user via the user interface.

2 . The device of claim 1 , wherein the one or more processors are configured to:

convert the natural language prompt into an initial SQL query using the LLM and vector embeddings corresponding to previous schema information;

detect an execution error indicative of schema mismatch or change based on executing the initial SQL query; and

in response to detecting the execution error, automatically update the schema information by transmitting a schema query to the monitoring database and updating the vector database with new vector embeddings.

3 . The device of claim 1 , wherein, to convert the natural language prompt into the SQL query, the one or more processors are configured to:

map extracted information from the natural language prompt to database entities represented by vector embeddings of the schema information using the LLM; and

construct one or more SQL commands in the SQL query to handle joins, filters, or aggregations.

4 . The device of claim 1 , wherein the one or more processors are configured to:

store the schema information in a vector database; and

query the vector database to retrieve the schema information before converting the natural language prompt into the SQL query.

5 . The device of claim 1 , wherein, to execute the SQL query, the one or more processors are configured to:

schedule execution of the SQL query according to a monitoring interval indicated in the natural language prompt.

6 . The device of claim 1 , wherein the one or more processors are configured to:

transmit a command to instantiate a service ticket in an issue tracking system based on an anomaly detected in the monitoring results.

7 . A method, comprising:

receiving, via a user interface of a query system, a natural language prompt from a user;

generating, by a large language model (LLM), a vector representation of the natural language prompt;

retrieving, from a vector database, schema information associated with a monitoring database, the schema information including vector embeddings of database entities;

converting, by the LLM and using the vector representation of the prompt and the vector embeddings of the schema information, the natural language prompt into a structured query language (SQL) query;

executing, by the query system, the SQL query against the monitoring database to receive monitoring results;

detecting, by the query system, an execution error resulting from a schema mismatch or schema change in the monitoring database;

in response to detecting the execution error, automatically updating the schema information by transmitting a schema query to the monitoring database and updating the vector database with new vector embeddings;

regenerating, by the LLM, the SQL query using the updated schema information; and

converting, by the query system, the monitoring results into a natural language response, using the LLM, for output to the user via the user interface.

8 . The method of claim 7 , wherein converting the natural language prompt into the SQL query comprises:

generating, by the LLM, an initial SQL query using the vector representation of the natural language prompt and vector embeddings corresponding to a previous version of the schema information;

detecting, by the query system, an execution error based on executing the initial SQL query; and

in response to detecting the execution error, updating the schema information by transmitting a schema query to the monitoring database and updating the vector database with new vector embeddings.

9 . The method of claim 7 , wherein converting the natural language prompt into the SQL query comprises:

mapping, by the query system and using vector representations and vector embeddings, extracted information from the natural language prompt to database entities represented by the schema information; and

constructing, by the query system, one or more SQL commands in the SQL query to handle joins, filters, or aggregations.

10 . The method of claim 7 , further comprising:

transmitting, by the query system, the natural language response via a communication channel that is selected based on the natural language prompt.

11 . The method of claim 7 , further comprising:

storing, by the query system, the schema information in a vector database; and

querying, by the query system, the vector database to retrieve the schema information before converting the natural language prompt into the SQL query.

12 . The method of claim 7 , wherein executing the SQL query comprises:

scheduling, by the query system, execution of the SQL query according to a monitoring interval indicated in the natural language prompt.

13 . The method of claim 7 , further comprising:

transmitting, by the query system, a command to instantiate a service ticket in an issue tracking system based on an anomaly detected in the monitoring results.

14 . A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:

one or more instructions that, when executed by one or more processors of a device, cause the device to:

receive, via a user interface, a natural language prompt from a user;

generate, by a large language model (LLM), a vector representation of the natural language prompt;

retrieve, from a vector database, schema information associated with a monitoring database, the schema information including vector embeddings of database entities;

convert, by the LLM and using the vector representation of the prompt and the vector embeddings of the schema information, the natural language prompt into a structured query language (SQL) query;

execute the SQL query against the monitoring database to receive monitoring results;

detect, by the device, an execution error resulting from a schema mismatch or schema change in the monitoring database;

in response to detecting the execution error, automatically update the schema information by transmitting a schema query to the monitoring database and update the vector database with new vector embeddings;

regenerate, by the LLM, the SQL query using the updated schema information; and

convert the monitoring results into a natural language response, using the LLM, for output to the user via the user interface.

15 . The non-transitory computer-readable medium of claim 14 , wherein the one or more instructions cause the device to:

convert the natural language prompt into an initial SQL query using the LLM and vector embeddings corresponding to previous schema information;

detect an execution error indicative of schema mismatch or change based on executing the initial SQL query; and

in response to detecting the execution error, automatically update the schema information by transmitting a schema query to the monitoring database and update the vector database with new vector embeddings.

16 . The non-transitory computer-readable medium of claim 14 , wherein the instructions, that cause the device to convert the natural language prompt into the SQL query, cause the device to:

map extracted information from the natural language prompt to database entities represented by vector embeddings of the schema information using the LLM; and

construct one or more SQL commands in the SQL query to handle joins, filters, or aggregations.

17 . The non-transitory computer-readable medium of claim 14 , wherein the one or more instructions cause the device to:

store the schema information in a vector database; and

query the vector database to retrieve the schema information before converting the natural language prompt into the SQL query.

18 . The non-transitory computer-readable medium of claim 14 , wherein the instructions, to execute the SQL query, the one or more processors are configured to:

schedule execution of the SQL query according to a monitoring interval indicated in the natural language prompt.

19 . The non-transitory computer-readable medium of claim 14 , wherein the one or more instructions further cause the device to:

transmit a command to instantiate a service ticket in an issue tracking system based on an anomaly detected in the monitoring results.

20 . The device of claim 1 , wherein the one or more processors are configured to, prior to converting the natural language prompt into the SQL query, query the vector database to retrieve the schema information and provide both the natural language prompt and the schema information as input to the LLM.