GetHelp Script include - Pull6

Server-Side Components/Script Includes/GetServiceDeskAgentHelpAIUtil/README.md

@@ -0,0 +1,68 @@
+GetServiceDeskAgentHelpAIUtil
+Overview
+This is a client-callable Script Include designed for the ServiceNow platform that integrates with an external Databricks AI endpoint. Its purpose is to assist Service Desk agents by providing AI-generated responses to user queries, which can be used to populate incident information or assist in troubleshooting.
+The script acts as a server-side proxy, handling the client-side request, invoking a Flow Designer Action to communicate with the Databricks service, and returning the AI's response to the client.
+Features
+AI-Powered Responses: Sends user search queries to a Databricks-powered Generative AI model to get intelligent answers.
+Client-Callable: Can be invoked from client-side scripts (e.g., a UI Action or Client Script) using GlideAjax.
+Flow Designer Integration: Uses a Flow Designer Action to execute the REST call to the Databricks endpoint, centralizing the external API logic.
+Structured Output: Returns a JSON object containing the AI's response, model metadata, and a trace ID for debugging.
+Prerequisites
+ServiceNow Configuration
+Flow Designer Action: A Flow Designer Action named global.genai_action must be created and configured to handle the REST call to the Databricks AI endpoint. This action must have:
+An input named search_query (String).
+An output named model_output (String).
+Databricks Connection: The Flow Designer Action must be correctly configured with the necessary credentials to connect to the external Databricks API.
+System Property: A system property named user.prompt is referenced in the script. It should be created and configured with the required prompt text for the AI.
+How to use
+1. Calling from a Client Script
+You can use GlideAjax to call the getSearchResults function from a client-side script, such as a UI Action or a Catalog Client Script.
+javascript
+// Example client-side script using GlideAjax
+var ga = new GlideAjax('GetServiceDeskAgentHelpAIUtil');
+ga.addParam('sysparm_name', 'getSearchResults');
+ga.addParam('sysparm_search_key', g_form.getValue('short_description')); // Pass the user's input
+ga.getXML(getResponse);
+
+function getResponse(response) {
+    var answer = response.responseXML.documentElement.getAttribute("answer");
+    if (answer) {
+        var result = JSON.parse(answer);
+        g_form.setValue('comments', result.modelResponse); // Set the AI response in a field
+    }
+}
+Use code with caution.
+
+2. Using from a Server Script
+The functions can also be called directly from other server-side scripts (e.g., Business Rules).
+javascript
+// Example server-side script
+var searchKey = 'What are the steps to reset my password?';
+var aiUtil = new GetServiceDeskAgentHelpAIUtil();
+var response = aiUtil.getSearchResults(searchKey);
+
+gs.info(response);
+Use code with caution.
+
+Script details
+getSearchResults()
+This is the main function that coordinates the process.
+Retrieves the search term from the client parameters.
+Calls the getDataBricksModelResponse() function to get the AI-generated answer.
+Constructs a JSON object with the AI's response and model information.
+Returns the JSON object as a string.
+getDataBricksModelResponse(search)
+This function handles the integration with the Databricks AI.
+Takes the search query as a parameter.
+Executes the global.genai_action Flow Designer Action.
+Parses the model_output from the Flow Action's outputs.
+Extracts the AI's message content and a trace ID for debugging.
+Returns a stringified JSON object containing the AI's response, date, and trace ID.
+Includes a try/catch block to handle and log potential errors during the integration process.
+Dependencies
+Flow Designer Action: global.genai_action
+System Property: user.prompt
+Troubleshooting
+Check the Flow Execution: If the AI response is not received, check the Flow Designer execution logs to ensure global.genai_action is running successfully and the REST call to Databricks is returning a valid response.
+Review System Logs: Examine the System Logs (gs.info and gs.error messages) for debugging information related to the script's execution or potential errors from the Databricks API.
+Verify Databricks Credentials: Ensure that the credentials and configuration within the Flow Designer action for connecting to Databricks are correct.

Server-Side Components/Script Includes/GetServiceDeskAgentHelpAIUtil/script.js

@@ -0,0 +1,110 @@
+var GetServiceDeskAgentHelpAIUtil = Class.create();
+GetServiceDeskAgentHelpAIUtil.prototype = Object.extendsObject(global.AbstractAjaxProcessor, {
+
+  /**
+     * Main function that processes a search term from a client-side call,
+     * sends it to a Databricks-powered AI, and returns the response.
+     * This function is intended to assist Service Desk agents.
+     *
+     * @returns {string} A JSON string containing the AI's response, model details, and metrics.
+     */
+    getSearchResults: function() {
+        // Defines the use case for logging and metric purposes.
+        var usecase = "ServiceDesk Helper";
+        // Gets the current user's Sys ID, though it is not used in the current implementation.
+        var user = gs.getUserID();
+        // Retrieves the search term passed from the client-side script.
+        var searchText = this.getParameter("sysparm_search_key");
+        // Replaces double quotes with single quotes in the search text to prevent JSON parsing issues.
+        searchText = searchText.replaceAll('"', "'");
+
+        var searchObj = {
+            "searchValue": searchText.toString()
+        };
+
+        // Extracts the raw search value from the object.
+        var search = searchObj["searchValue"];
+
+        // This object is structured to create a prompt for another potential AI endpoint (possibly for a brief statement),
+        // but it is currently not used.
+        var brief_statement_payload = {
+            "messages": [{
+                    "role": "system",
+                    "content": "You are an Expert ServiceNow bot that helps the users to create an incident"
+                },
+                {
+                    "role": "user",
+                    "content": gs.getProperty('user.prompt') + search
+                }
+            ]
+        };
+
+        var databricks_model_response = {};
+        // Calls the internal method to get the response from the Databricks model.
+        var response = this.getDataBricksModelResponse(search);
+        // UNCOMMENT THIS WHEN WE HAVE A PROPER SOLUTION FOR BRIEF RESPONSE GENERATION
+        // var brief_response = this.getBriefResponse(brief_statement_payload);
+        // The brief response is hardcoded to an empty JSON object
+        // Assigns the model response to the output object.
+        databricks_model_response.modelResponse = response;
+        // Assigns a hardcoded model ID.
+        databricks_model_response.model_id = "Databricks Runbook";
+
+        // Converts the final response object to a JSON string for client-side processing.
+        databricks_model_response = JSON.stringify(databricks_model_response);
+        // Logs the final JSON string for debugging purposes.
+        gs.info("Service Desk Helper Results: Testing value of the final databricks response being sent: " + databricks_model_response);
+
+        // Returns the JSON string to the calling client script.
+        return databricks_model_response;
+    },
+
+    /**
+     * This function calls the Databricks endpoint via a Flow Designer action
+     * to generate an answer for the user's query.
+     *
+     * @param {string} search - The user's search query.
+     * @returns {string} A JSON string containing the AI's response, the current date, and a trace ID.
+     */
+    getDataBricksModelResponse: function(search) {
+        try {
+            var inputs = {};
+            // Maps the search query to the input expected by the flow action.
+            inputs['search_query'] = search;
+
+            // Executes the specified Flow Designer action with the provided inputs.
+            // The action is run in the foreground, meaning the script will wait for a response.
+            var result = sn_fd.FlowAPI.getRunner().action('global.genai_action').inForeground().withInputs(inputs).run();
+            // Retrieves the outputs from the completed flow action.
+            var outputs = result.getOutputs();
+
+            // Extracts the model output from the flow action outputs.
+            var model_output = outputs['model_output'];
+            // Attempts to parse and extract vector response data, though the variable is not used after this line.
+            var databricks_vector_response = JSON.parse(model_output).databricks_output.trace.data.spans;
+
+            // Logs the raw response from the Databricks model for debugging.
+            gs.info("Helper Results: Databricks flow action response: " + JSON.stringify(model_output));
+
+            var current_date = new GlideDateTime();
+            var output = {};
+            // Parses the model output to extract the AI's content.
+            output.response = JSON.parse(model_output).choices[0].message.content;
+            // Adds the current date to the output object.
+            output.date = current_date.getDisplayValue();
+            // Logs the trace ID for tracking purposes.
+            gs.info("Helper Results: GEN AI flow action TraceID value: " + JSON.parse(model_output).id);
+            // Adds the trace ID to the output object.
+            output.traceID = JSON.parse(model_output).id;
+            // Returns the constructed output object as a JSON string.
+            return JSON.stringify(output);
+
+        } catch (ex) {
+            // Catches any exceptions during the flow execution and logs an error.
+            var message = ex.getMessage();
+            gs.error(message);
+        }
+    },
+
+  type: 'GetServiceDeskAgentHelpAIUtil'
+});