Qoder CN Series:Configure filters

• Processing method: Configure filters using regular expressions. Three modes are supported.

  No action on match	Take no action when a match occurs.
  Block on match	Block the request and stop the model call when a match occurs.
  Replace on match	Replace content according to the configuration when a match occurs.

• Notification: Enable notifications to send alerts to your preferred messaging platform using webhooks.

• Execution order: Filters run in the order you configure them.

• Maximum number of regular expressions: You can add up to 10 expressions.

• Regular expression standard: Expressions follow the ECMAScript standard. Common flags such as i (case-insensitive), g (global), and s (DOTALL) are supported.

• Example configurations:

  Rule name	Regular expression	Replacement	Original text	Replaced text
  ID number	(?<pre>.*)(\d{15})((\d{2})([0-9Xx]))(?<post>.*)	$<pre>***$<post>	ID number: 330204197709022312.	ID number: ***.
  Email address	\w+([-+.]\w+)*@\w+([-.]\w+)*\.\w+([-.]\w+)*	***	My email is lin***@aliyunmail.com	My email is ***
  Password	(.*password=)([\w\d]+)(.*)	$1***$3	{password=1213213}	{password=***}

Step 1: Develop the script

Write scripts in TypeScript using the sample code as a reference.

1. Download the template repository: Click the repository URL: lingma-extension-template. This template includes scaffolding for script development. Read the README.md file and code examples carefully.

2. Implement the pre-processing interface: Implement the RequestPreHandler interface. For API details, see Custom Script API. Here is an example implementation in SensitiveContentFilter.ts:

   /**
    * Sensitive content filter. Pre-processes data sent to the model.
    */
   export const sensitiveContentFilter: RequestPreHandler = {
     handle: async (request: RawRequest, SDKTool: Qoder CNSDKTool) => {
       const dataMap = PayloadUtil.getPayloadData(request.payload);
       for (const [key, value] of dataMap.entries()) {
         if (value.includes('password')) {
           return ResultUtil.buildBlockResult('Content contains password');
         }
       }
       // Handle different actions differently
       switch (request.action) {
         case ActionEnum.COMPLETION:
           // do something
           break;
         case ActionEnum.CODE_PROBLEM_SOLVE:
           // do something
           break;
         default:
           return ResultUtil.buildNoOpsResult();
       }
       return ResultUtil.buildNoOpsResult();
     },
   };

3. Run and debug the script. Test it by running the main function. Follow these steps:

   Step 1

   Edit the src/index.ts file. Modify the main function to test your script. Example: async function main() { const value1 = ['password=123', 'abc']; const value2 = 'hello world'; const dataMap = new Map<PayloadDataKeyEnum, PayloadDataValueType>(); dataMap.set(PayloadDataKeyEnum.SELECTED_CODE, value1); dataMap.set(PayloadDataKeyEnum.USER_INPUT, value2); const mockRequest: RawRequest = { action: ActionEnum.CODE_GENERATE_COMMENT, payload: { associatedContexts: [], data: dataMap, }, requestId: '123', }; const response = await sensitiveContentFilter.handle(mockRequest, SDKTool); console.log(response); }

   Step 2

   In VS Code, open the script file and set breakpoints. Then, in the Debug view, select Launch Program and click Run. The sample code file index.ts contains the following content. import { sensitiveContentFilter } from './demo/SensitiveContentFilter'; import { ActionEnum, PayloadDataKeyEnum, PayloadDataValueType, RawRequest } from './sdk/RequestPreHxxx'; const SDKTool = { idePlatform: 'vscode', ideVersion: '1.0', pluginVersion: '1.0', user: { name: 'tester', organizationId: '123', organizationName: 'test_org', uid: 'org_123', }, };

Step 2: Compile and Build

Compile TypeScript to JavaScript. For example, compile SensitiveContentFilter.ts into SensitiveContentFilter.js:

1. Open the configuration file src/build.js. Update the entryPoints and outfile parameters. Set entryPoints to the path of your TypeScript file. Set outfile to the output path for the compiled JavaScript file.

2. Run the command node build.js from the root directory of your repository. After successful execution, the JavaScript file appears at the path specified in outfile.

Step 3: Test locally

Test the script locally before uploading to verify it integrates with the Qoder CN IDE extension and filters correctly.

1. Copy the compiled JavaScript file to the /extension/local/script/ directory under the local storage path of Qoder CN.

2. Edit the config.json file. This file is located in the /extension/local/ directory under the local storage path of Qoder CN. Open config.json and find contentHandlerScripts. Add your script's configuration. If contentHandlerScripts does not exist, create a new array. Example:

   {
     "contentHandlerScripts": [
       {
         "identifier": "SensitiveContentFilter", 
         "name": "Sensitive Content Filter",
         "version": "1.0.0",
         "scriptPath": "~/.lingma/extension/local/script/SensitiveContentFilter.js",
         "state": "enabled",
         "bizType": "completion"
       }
     ]
   }

   Parameter descriptions:

   Parameter	Description
   identifier	Script ID. Must be unique.
   name	Script name.
   version	Script version. Increment after each change, or the updated script will not take effect.
   scriptPath	Script file path. Note: The script must be placed in the /extension/local/script/ directory under the local storage path. The JavaScript file name (for example, SensitiveContentFilter.js) must match the identifier value.
   state	Script state. enabled enables the script. disabled disables it.
   bizType	Business scenario. completion applies to inline code generation. chat applies to AI chat.

Step 4: Upload the script

After local testing, upload the script:

1. Go to the Qoder CN console. Choose Policy Management. Select the scenario where you want to enable the security filter.

2. Select Custom Script as the filter option.

3. Upload the compiled JavaScript file.

4. Click Save Configuration. The configuration takes effect on the extension within about 5 minutes.

Custom Script API

Custom scripts support three processing methods:

• Block: Stop further processing. No inference call is made to the LLM. The current request ends.

• Filter: Modify the data sent (for example, obfuscate, delete, or replace content), then continue processing.

• No operation: Return the data unchanged and continue processing.

/**
 * Pre-processing interface for Qoder CN
 */
export interface RequestPreHandler {
  // Process the request
  handle: (request: RawRequest, SDKTool: Qoder CNSDKTool) => Promise<HandlerResponse>;
}

/**
 * Request object. Includes the action and raw data sent to the LLM.
 */
export interface RawRequest {
  // Unique identifier for the request. Used for tracing.
  action: ActionEnum;
  // Enum for the action that triggered the request.
  payload: ContentPayload;
  // Payload containing raw data.
  requestId: string;
}

// Value type for ContentPayload.data
export type PayloadDataValueType = string | number | string[];
/**
 * Payload containing raw data sent to the LLM.
 */
export class ContentPayload {
  // Collection of data to process. Keys are defined in ContextValueKeyEnum.
  data: Map<PayloadDataKeyEnum, PayloadDataValueType>;
  // Context associated with the processing.
  associatedContexts: ContextItem[];

  constructor() {
    this.data = new Map<PayloadDataKeyEnum, PayloadDataValueType>();
    this.associatedContexts = [];
  }
}

/**
 * Keys for ContentPayload.data
 */
export enum PayloadDataKeyEnum {
  // Selected code snippet
  SELECTED_CODE ='lingma:code',
  // User input text
  USER_INPUT = 'lingma:text',
  // Error messages
  ERROR_MESSAGES = 'lingma:error_messages',
  // Terminal log output
  TERMINAL_CONTENT = 'lingma:terminal_content',
  // Code before cursor for code completion
  PREFIX_CODE = 'lingma:code_prefix',
  // Code after cursor for code completion
  SUFFIX_CODE = 'lingma:code_suffix',
  // Similar code snippets
  SIMILAR_CODE = 'lingma:similar_code',
}

/**
 * Enum for request actions
 */
export enum ActionEnum {
    // Unit test
    GENERATE_TESTCASE           = 'GENERATE_TESTCASE',
    // Generate comments
    CODE_GENERATE_COMMENT       = 'CODE_GENERATE_COMMENT',
    // Explain code
    EXPLAIN_CODE                = 'EXPLAIN_CODE',
    // Optimize code
    OPTIMIZE_CODE               = 'OPTIMIZE_CODE',
    // Free input (direct text entry in the chat box)
    FREE_INPUT                  = 'FREE_INPUT',
    // Quick fix for code issues
    CODE_PROBLEM_SOLVE          = 'CODE_PROBLEM_SOLVE',
    // Generate shell commands
    TERMINAL_COMMAND_GENERATION = 'TERMINAL_COMMAND_GENERATION',
    // Fix terminal errors
    TERMINAL_EXPLAIN_FIX        = 'TERMINAL_EXPLAIN_FIX',
    // Code completion
    COMPLETION                  = 'COMPLETION',
}

/**
 * Pre-processing result
 */
export class HandlerResponse {
  // Processing policy. Controls subsequent logic.
  handlePolicy: HandlePolicy;
  // Reason description
  reason?: string;
  // Required when handlePolicy=FILTER. Contains filtered data. Must match ContentRequest.payload structure.
  payload?: ContentPayload;
  constructor() {
    // Default values
    // eslint-disable-next-line @typescript-eslint/no-use-before-define
    this.handlePolicy = HandlePolicy.NO_OPS;
    this.reason = '';
    this.payload = new ContentPayload();
  }
}

/**
 * Processing policy enum
 */
export enum HandlePolicy {
  // Block the request
  BLOCK = 'BLOCK',
  // Filter the request and modify the payload
  FILTER = 'FILTER',
  // Take no action
  NO_OPS = 'NO_OPS',
}

• Processing method: Configure filters using regular expressions. Only one mode is supported:

  No action on match

  Take no action when a match occurs.

• Notification: Enable notifications to send alerts to your preferred messaging platform using webhooks.

• Execution order: Filters run in the order you configure them.

• Maximum number of regular expressions: You can add up to 10 expressions.

• Regular expression standard: Expressions follow the ECMAScript standard. Common flags such as i (case-insensitive), g (global), and s (DOTALL) are supported.

• Example configurations:

  Rule name	Regular expression	Original text
  ID number	(?<pre>.*)(\d{15})((\d{2})([0-9Xx]))(?<post>.*)	ID number: 330204197709022312.
  Email address	\w+([-+.]\w+)*@\w+([-.]\w+)*\.\w+([-.]\w+)*	My email is lin***@aliyunmail.com
  Password	(.*password=)([\w\d]+)(.*)	{password=1213213}

Step 1: Develop the script

Write scripts in TypeScript using the sample code as a reference.

1. Download the template repository: Click the repository URL: lingma-extension-template. This template includes scaffolding for script development. Read the README.md file and code examples carefully.

2. Implement the post-processing interface: Implement the RequestPostHandler interface. For the API, refer to Custom Script API. The following is an implementation example from the code snippet LLMChatAuditHandler.ts, which audits AIChat operations of Qoder CN and reports the audit content to Alibaba Cloud SLS.

   import {ResultUtil} from '../common/HandlerRespUtil';
   import { JsonUtil } from '../common/JsonUtil';
   import {PayloadUtil} from '../common/PayloadUtil';
   import {Config} from '../sdk/ConfigManager';
   import {Qoder CNSDKTool} from '../sdk/Qoder CNSDKTool';
   import axios from "axios";
   import moment from "moment";
   import os from "os";
   import {
     ActionEnum, AIResponse,
     HandlePolicy,
     RawRequest,
     RequestPostHandler,
     RequestPreHandler
   } from '../sdk/RequestHandleSDK';
   
   /**
    * Custom post-filter script. Sends request content to a remote server for processing (such as code scanning or content auditing).
    */
   export const llmResultAuditHandler: RequestPostHandler = {
     handle: async (request: RawRequest, response: AIResponse,SDKTool: Qoder CNSDKTool) => {
        // Operator name
        let userName = SDKTool.user.name;
        // Operator ID
        let userId = SDKTool.user.uid;
        // IDE
        let ide = SDKTool.idePlatform;
        // IDE version
        let ideVersion = SDKTool.ideVersion;     
        // Operation time
        let operationTime = moment().format("YYYY-MM-DD HH:mm:ss");
        // Operation IP
        let opeartionIp = getIpAddress();
        // Business scenario (completion or chat)
        let bizType = "chat";
        // Request ID
        let requestId = request.requestId;
        // Action
        let action = request.action;
        // Operation content (select fields based on your audit needs. Avoid large payloads over 16 KB.)
        let inferredResult = response.inferredResult.text;
       
        // Report to SLS
        // SLS project name
        let slsProject = "xxx";
        // SLS LogStore name
        let slsLogStore = "xxx";
        // SLS endpoint
        let endPoint = "cn-hangzhou.log.aliyuncs.com";
        let slsWebTrackingUrl = `http://${slsProject}.${endPoint}/logstores/${slsLogStore}/track?APIVersion=0.6.0&request_id=${requestId}&action=${action}&biz_type=${bizType}&user_name=${userName}&user_id=${userId}&ide=${ide}&ide_version=${ideVersion}&operation_time=${operationTime}&opeartion_ip=${opeartionIp}&inferredResult=${inferredResult}`;
        axios.get(slsWebTrackingUrl).catch((error) => {
          console.error(error);
        });
   
        // Return filter result
       return ResultUtil.buildPostHandlerResponse(HandlePolicy.NO_OPS, response.inferredResult,'No action required');
     },
   };
   
   /**
    * Add the custom script filter to the configuration
    * @param config Config object provided by Qoder CNExtensionSDK
    */
   export function modifyConfig(config: Config) {
     config.postContentHandlers.push(llmResultAuditHandler);
     return config;
   }
   
   function getIpAddress() {
     const interfaces = os.networkInterfaces();
     for (let devName in interfaces) {
       let iface = interfaces[devName];
   
       for (let i = 0; i < iface.length; i++) {
         let alias = iface[i];
         if (
           alias.family === "IPv4" &&
           alias.address !== "127.0.0.1" &&
           !alias.internal
         )
           return alias.address;
       }
     }
   
     return "No IP address found";
   }

3. Run and debug the script. Test it by running the main function. Follow these steps:

   Step 1

   Edit the src/index.ts file. Modify the main function to test your script. Example: async function main() { const value2 = 'hello world'; const dataMap = new Map<PayloadDataKeyEnum, PayloadDataValueType>(); dataMap.set(PayloadDataKeyEnum.USER_INPUT, value2); const request: RawRequest = { action: ActionEnum.CODE_GENERATE_COMMENT, payload: { associatedContexts: [], data: dataMap, }, requestId: 'test-request-id', }; const aiResponse: AIResponse = { inferredResult: { text: 'reply hello world', }, }; const response = await llmResultAuditHandler.handle(request, aiResponse, SDKTool); console.log(response); }

   Step 2

   In VS Code, open the script file and set breakpoints. Then, in the Debug view, select Launch Program and click Run. The sample code file index.ts contains the following content. import { sensitiveContentFilter } from './demo/SensitiveContentFilter'; import { ActionEnum, PayloadDataKeyEnum, PayloadDataValueType, RawRequest } from './sdk/RequestPreHxxx'; const SDKTool = { idePlatform: 'vscode', ideVersion: '1.0', pluginVersion: '1.0', user: { name: 'tester', organizationId: '123', organizationName: 'test_org', uid: 'org_123', }, };

Step 2: Compile and Build

Compile TypeScript to JavaScript. For example, compile LLMChatAuditHandler.ts into LLMChatAuditHandler.js:

1. Open the configuration file src/build.js. Update the entryPoints and outfile parameters. Set entryPoints to the path of your TypeScript file. Set outfile to the output path for the compiled JavaScript file.

2. Run the command node build.js from the root directory of your repository. After successful execution, the JavaScript file appears at the path specified in outfile.

Step 3: Test locally

Test the script locally before uploading to verify it integrates with the Qoder CN IDE extension and filters correctly.

1. Copy the compiled JavaScript file to the /extension/local/script/ directory under the local storage path of Qoder CN.

2. Edit the config.json file. This file is located in the /extension/local/ directory under the local storage path of Qoder CN. Open config.json and find contentHandlerScripts. Add your script's configuration. If contentHandlerScripts does not exist, create a new array. Example:

   {
     "contentHandlerScripts": [
       {
         "identifier": "LLMChatAuditHandler", 
         "name": "AI Chat Audit",
         "version": "1.0.0",
         "scriptPath": "~/.lingma/extension/local/script/LLMChatAuditHandler.js",
         "state": "enabled",
         "stage":"post",
         "bizType": "completion"
       }
     ]
   }

   Parameter descriptions:

   Parameter	Description
   identifier	Script ID. Must be unique.
   name	Script name.
   version	Script version. Increment after each change, or the updated script will not take effect.
   scriptPath	Script file path. Note: The script must be placed in the /extension/local/script/ directory under the local storage path. The JavaScript file name (for example, LLMChatAuditHandler.js) must match the identifier value.
   state	Script state. enabled enables the script. disabled disables it.
   stage	Processing stage. post means post-filter. pre means pre-filter. Default is pre.
   bizType	Business scenario. completion applies to inline code generation. chat applies to AI chat.

Step 4: Upload the script

After local testing, upload the script:

1. Go to the Qoder CN console. Choose Policy Management. Select the scenario where you want to enable the security filter.

2. Select Custom Script as the filter option.

3. Upload the compiled JavaScript file.

4. Click Save Configuration. The configuration takes effect on the extension within about 5 minutes.

Custom Script API

Custom scripts support only one processing method:

• No operation: Return the data unchanged and continue processing.

/**
 * Post-processing interface for Qoder CN
 * @param request User request
 * @param response Inference result from the LLM
 * @param SDKTool SDK utility class. Provides IDE and plugin information.
 * @returns Result after post-processing
 */
export interface RequestPostHandler {
  // Post-processing method
  handle: (request: RawRequest, response: AIResponse, SDKTool: Qoder CNSDKTool) => Promise<PostHandlerResponse>;
}

/**
 * Request object. Includes the action and raw data sent to the LLM.
 */
export interface RawRequest {
  // Unique identifier for the request. Used for tracing.
  action: ActionEnum;
  // Enum for the action that triggered the request.
  payload: ContentPayload;
  // Payload containing raw data.
  requestId: string;
}

/**
 * Inference result from the model
 */
export class InferredResult {
  // Text generated by the LLM
  text: string;
  constructor() {
    this.text = '';
  }
}

// Value type for ContentPayload.data
export type PayloadDataValueType = string | number | string[];
/**
 * Payload containing raw data sent to the LLM.
 */
export class ContentPayload {
  // Collection of data to process. Keys are defined in ContextValueKeyEnum.
  data: Map<PayloadDataKeyEnum, PayloadDataValueType>;
  // Context associated with the processing.
  associatedContexts: ContextItem[];

  constructor() {
    this.data = new Map<PayloadDataKeyEnum, PayloadDataValueType>();
    this.associatedContexts = [];
  }
}

/**
 * Keys for ContentPayload.data
 */
export enum PayloadDataKeyEnum {
  // Selected code snippet
  SELECTED_CODE ='lingma:code',
  // User input text
  USER_INPUT = 'lingma:text',
  // Error messages
  ERROR_MESSAGES = 'lingma:error_messages',
  // Terminal log output
  TERMINAL_CONTENT = 'lingma:terminal_content',
  // Code before cursor for code completion
  PREFIX_CODE = 'lingma:code_prefix',
  // Code after cursor for code completion
  SUFFIX_CODE = 'lingma:code_suffix',
  // Similar code snippets
  SIMILAR_CODE = 'lingma:similar_code',
  // File path for code completion
  FILE_PATH = 'lingma:file_path',
}

/**
 * Enum for request actions
 */
export enum ActionEnum {
    // Unit test
    GENERATE_TESTCASE           = 'GENERATE_TESTCASE',
    // Generate comments
    CODE_GENERATE_COMMENT       = 'CODE_GENERATE_COMMENT',
    // Explain code
    EXPLAIN_CODE                = 'EXPLAIN_CODE',
    // Optimize code
    OPTIMIZE_CODE               = 'OPTIMIZE_CODE',
    // Free input (direct text entry in the chat box)
    FREE_INPUT                  = 'FREE_INPUT',
    // Quick fix for code issues
    CODE_PROBLEM_SOLVE          = 'CODE_PROBLEM_SOLVE',
    // Generate shell commands
    TERMINAL_COMMAND_GENERATION = 'TERMINAL_COMMAND_GENERATION',
    // Fix terminal errors
    TERMINAL_EXPLAIN_FIX        = 'TERMINAL_EXPLAIN_FIX',
    // Code completion
    COMPLETION                  = 'COMPLETION',
}

/**
 * Post-processing result
 */
export class PostHandlerResponse {
  // Processing policy. Controls subsequent logic.
  handlePolicy: HandlePolicy;
  // Reason description
  reason?: string;
  // Processed inference result
  processedResult: InferredResult;
  constructor() {
    // Default values
    this.handlePolicy = HandlePolicy.NO_OPS;
    this.reason = '';
    this.processedResult = new InferredResult();
  }
}

/**
 * Encapsulates the LLM response
 */
export class AIResponse {
  // Inference result
  inferredResult: InferredResult;
  constructor() {
    this.inferredResult = new InferredResult();
  }
}

/**
 * Inference result from the model
 */
export class InferredResult {
  // Text generated by the LLM
  text: string;
  constructor() {
    this.text = '';
  }
}

/**
 * Processing policy enum (post-filters support only NO_OPS)
 */
export enum HandlePolicy {
  // Block the request
  BLOCK = 'BLOCK',
  // Filter the request and modify the payload
  FILTER = 'FILTER',
  // Take no action
  NO_OPS = 'NO_OPS',
}