Telecommunications network management
A natural language query is received from an operator of a telecommunications network. A metadata request is computed from the natural language query. The metadata request is sent to a repository of metadata, the metadata describing telemetry data of the telecommunications network, the telemetry data stored in a relational database. Metadata is received from the metadata repository in response to the metadata request. Using the received metadata and a language model a relational database query is computed and the relational database is queried. A response is received from the relational database, triggering an action.
1 . A computer-implemented method comprising:
receiving a natural language query from an operator of a telecommunications network;
generating a metadata request from the natural language query;
sending the metadata request to a repository of metadata, the metadata describing telemetry data of the telecommunications network, the telemetry data stored in a relational database;
receiving metadata from the metadata repository in response to the metadata request;
using the received metadata, generating a prompt for input to a language model, wherein the prompt is usable to cause the language model to generate a relational database query for querying the relational database using a specified query language;
sending the prompt to the language model;
receiving, in response to the prompt, the relational database query using the specified query language;
querying the relational database using the relational database query; and
receiving a response from the relational database and triggering an action related to the natural language query using the received response; wherein the action is localized in the telecommunications network.
2 . The method of claim 1 , wherein the repository of metadata is separate from the relational database and wherein the relational database comprises a plurality of tables distributed over different geographical locations.
3 . The method of claim 1 , wherein sending the metadata request to a repository of metadata triggers search of an index of tables of the relational database and their column descriptions.
4 . The method of claim 1 , wherein the repository of metadata comprises a data map having locations of tables of the relational database and wherein the received metadata comprises a location of a table of the relational database.
5 . The method of claim 1 , wherein the received metadata comprises any of: a table name, a column description, a table relationship, an indication that a column contains personally identifiable information.
6 . The method of claim 1 , wherein computing the metadata request from the natural language query comprises extracting key words from the natural language query and formulating a question from the extracted key words by using any of: a template, rules, a language model.
7 . The method of claim 1 , wherein using the received metadata comprises sending a request to the language model to generate an output, wherein the request specifies a relational database query language and contains the received metadata.
8 . The method of claim 1 , wherein computing the metadata request comprises computing a request for metadata where personally identifiable information is requested to be masked.
9 . The method of claim 8 , wherein receiving the metadata comprises receiving metadata with masking, whereby personally identifiable information is masked such that, the response from the relational database comprises tokens replacing personally identifiable information.
10 . The method of claim 1 , further comprising prior to querying the relational database, presenting the relational database query as a candidate to a user and receiving user input selecting the candidate.
11 . The method of claim 1 , further comprising triggering the action by creating a configuration file using rules and the received response from the relational database, and sending the configuration file to an orchestrator of the telecommunications network.
12 . An apparatus comprising:
a processor;
a memory storing instructions that, when executed by the processor, perform operations comprising:
receiving a natural language query from an operator of a telecommunications network;
computing a metadata request from the natural language query;
sending the metadata request to a repository of metadata, the metadata describing telemetry data of the telecommunications network, the telemetry data stored in a relational database;
receiving metadata from the metadata repository in response to the metadata request;
using the received metadata, generating a prompt for input to a language model, wherein the prompt is usable to cause the language model to generate a relational database query for querying the relational database using a specified query language;
sending the prompt to the language model;
receiving, in response to the prompt, the relational database query using the specified query language;
querying the relational database using the relational database query; and
receiving a response from the relational database and triggering an action related to the natural language query using the received response.
13 . The apparatus of claim 12 , wherein the action is any of: presenting the response at a user interface, using the response to create a configuration file and sending the configuration file to an orchestrator of the telecommunications network, sending an alert to an operator of the telecommunications network.
14 . The apparatus of claim 12 , wherein the apparatus is communicatively coupled to a management node of a telecommunications network and wherein the repository of metadata is a data governance repository having a data map indicating locations of tables of the relational database distributed geographically.
15 . The apparatus of claim 12 , further comprising an index of tables of the relational database and their column descriptions.
16 . The apparatus of claim 12 , wherein the repository is updated automatically in response to changes to a schema of the relational database.
17 . A computer-implemented method comprising:
receiving a natural language query from an operator of a 5G telecommunications network;
computing a metadata request from the natural language query;
sending the metadata request to a repository of metadata, the metadata describing telemetry data of the 5G telecommunications network, the telemetry data stored in a relational database;
receiving metadata from the metadata repository in response to the metadata request;
using the received metadata, generating a prompt for input to a language model, wherein the prompt is usable to cause the language model to generate a relational database query for querying the relational database using a specified query language;
sending the prompt to the language model;
receiving, in response to the prompt, the relational database query using the specified query language;
presenting the relational database query to an operator of the 5G telecommunications network;
receiving a user selection, selecting the relational database query;
in response to receiving the user selection, querying the relational database using the relational database query; and
receiving a response from the relational database and triggering an action related to the natural language query using the received response.
18 . The method of claim 17 , wherein computing the metadata request comprises computing a request for metadata where personally identifiable information is requested to be masked.
19 . The method of claim 18 , wherein receiving the metadata comprises receiving metadata with masking, whereby personally identifiable information is masked such that, the response from the relational database comprises tokens replacing personally identifiable information.
20 . The method of claim 19 , wherein the masking is carried out by the language model.