to: basically done embedding context provider logic, but with a little bug

Author: Sma1lboy

backend/package.json

@@ -40,22 +40,26 @@
     "@types/bcrypt": "^5.0.2",
     "@types/fs-extra": "^11.0.4",
     "@types/normalize-path": "^3.0.2",
+    "@types/pacote": "^11.1.8",
     "@types/toposort": "^2.0.7",
-    "toposort": "^2.0.2",
     "axios": "^1.7.7",
     "bcrypt": "^5.1.1",
     "class-validator": "^0.14.1",
+    "fastembed": "^1.14.1",
     "fs-extra": "^11.2.0",
     "graphql": "^16.9.0",
     "graphql-subscriptions": "^2.0.0",
     "graphql-ws": "^5.16.0",
     "lodash": "^4.17.21",
     "markdown-to-txt": "^2.0.1",
     "normalize-path": "^3.0.0",
+    "pacote": "^21.0.0",
     "reflect-metadata": "^0.2.2",
     "rxjs": "^7.8.1",
     "sqlite3": "^5.1.7",
     "subscriptions-transport-ws": "^0.11.0",
+    "tar": "^7.4.3",
+    "toposort": "^2.0.2",
     "typeorm": "^0.3.20",
     "uuid": "^10.0.0"
   },

backend/src/build-system/__tests__/dependencies-embedding-handler.integration.spec.ts

@@ -0,0 +1,87 @@
+import { Logger } from '@nestjs/common';
+import DependenciesEmbeddingHandler from '../dependencies-embedding-handler';
+
+// Initialize a global logger instance
+const logger = new Logger('dependencies embed tester');
+
+// Only run integration tests if INTEGRATION_TEST environment variable is set to '1'
+const isIntegrationTest = process.env.INTEGRATION_TEST === '1';
+
+if (!isIntegrationTest) {
+  logger.log(
+    'Integration tests are skipped. Set INTEGRATION_TEST=1 to run them.',
+  );
+} else {
+  describe('DependenciesEmbeddingHandler Integration Tests', () => {
+    let handler: DependenciesEmbeddingHandler;
+
+    // Increase the default timeout for integration tests
+    jest.setTimeout(300000); // 5 minutes
+
+    beforeAll(async () => {
+      logger.log(
+        'Initializing DependenciesEmbeddingHandler for integration tests...',
+      );
+      handler = new DependenciesEmbeddingHandler();
+      // Wait for the handler to initialize
+      await new Promise((resolve) => setTimeout(resolve, 5000)); // Wait 5 seconds
+      logger.log('Initialization complete.');
+    });
+
+    afterAll(() => {
+      logger.log('Integration tests completed.');
+    });
+
+    /**
+     * Integration Test Case: Add Real Packages and Perform a Search
+     *
+     * Purpose:
+     * - To verify that DependenciesEmbeddingHandler can handle real packages by fetching their type definitions,
+     *   generating embeddings, and storing them correctly.
+     * - To ensure that the search functionality can retrieve relevant packages based on a real query.
+     *
+     * Steps:
+     * 1. Add multiple real npm packages using the addPackage method.
+     * 2. Perform a search with a query related to one of the added packages.
+     * 3. Validate that the search results include the relevant package(s) and are correctly ranked.
+     */
+    test('should add real packages and perform a relevant search', async () => {
+      // Define real packages to add
+      const packagesToAdd = [
+        { name: 'lodash', version: '4.17.21' },
+        // { name: 'express', version: '4.18.2' },
+        // { name: 'react', version: '18.2.0' },
+        // { name: 'typescript', version: '4.9.5' },
+      ];
+
+      logger.log('Adding real packages...');
+
+      // Add all packages concurrently
+      await handler.addPackages(packagesToAdd);
+
+      logger.log('Packages added successfully.');
+
+      // Define a search query related to one of the packages, e.g., React
+      const searchQuery = 'React component lifecycle methods';
+
+      logger.log('Executing search with query:', searchQuery);
+
+      // Perform the search
+      const results = await handler.searchContext(searchQuery);
+
+      logger.log('Search results received.');
+
+      // Validate that results are returned
+      expect(results.length).toBeGreaterThan(0);
+
+      // Check that at least one of the top results is related to 'react'
+      const topResult = results[0];
+      expect(topResult.name).toBe('react');
+      expect(topResult.version).toBe('18.2.0');
+
+      logger.log('Top search result:', topResult);
+
+      // Optionally, you can print more details or perform additional assertions
+    });
+  });
+}

backend/src/build-system/__tests__/dependencies-embedding-handler.spec.ts

@@ -0,0 +1,177 @@
+import axios from 'axios';
+import DependenciesEmbeddingHandler from '../dependencies-embedding-handler';
+import { Logger } from '@nestjs/common';
+
+// Initialize a global logger instance
+const logger = new Logger('dependencies embed tester');
+
+// Mock axios to control HTTP requests during tests
+jest.mock('axios');
+const mockedAxios = axios as jest.Mocked<typeof axios>;
+
+// Mock fastembed to control embedding behavior during tests
+jest.mock('fastembed', () => ({
+  EmbeddingModel: {
+    BGEBaseEN: 'BGEBaseEN',
+  },
+  FlagEmbedding: {
+    init: jest.fn().mockResolvedValue({
+      passageEmbed: jest.fn(async function* (
+        types: string[],
+        batchSize: number,
+      ) {
+        for (const type of types) {
+          // Yield simulated embedding data as Float32Array
+          yield [new Float32Array([1, 2, 3])];
+        }
+      }),
+      queryEmbed: jest.fn(async (query: string) => [1, 2, 3]),
+    }),
+  },
+}));
+
+describe('DependenciesEmbeddingHandler', () => {
+  let handler: DependenciesEmbeddingHandler;
+
+  beforeEach(() => {
+    // Initialize a new instance of DependenciesEmbeddingHandler before each test
+    handler = new DependenciesEmbeddingHandler();
+    // Clear all mock calls and instances before each test
+    jest.clearAllMocks();
+  });
+
+  /**
+   * Test Case: Successfully add a package with built-in type definitions
+   *
+   * Purpose:
+   * - To verify that DependenciesEmbeddingHandler can correctly add a package that includes built-in type definitions.
+   * - To ensure that the handler retrieves the package's package.json, extracts the type definitions, and generates embeddings.
+   *
+   * Steps:
+   * 1. Mock axios.get to return a package.json containing the 'types' field.
+   * 2. Mock axios.get to return the content of the type definitions file.
+   * 3. Call the addPackage method to add the package.
+   * 4. Verify that the package information is correctly stored, including the generated embedding.
+   */
+  test('should successfully add a package with built-in types', async () => {
+    // Mock the response for fetching package.json, including the 'types' field
+    mockedAxios.get.mockImplementationOnce(() =>
+      Promise.resolve({
+        data: {
+          name: 'test-package',
+          version: '1.0.0',
+          types: 'dist/index.d.ts',
+        },
+      }),
+    );
+
+    // Mock the response for fetching the type definitions file
+    mockedAxios.get.mockImplementationOnce(() =>
+      Promise.resolve({
+        data: `
+          interface TestInterface {
+            prop1: string;
+            prop2: number;
+          }
+          
+          type TestType = {
+            field1: string;
+            field2: boolean;
+          };
+        `,
+      }),
+    );
+
+    // Add the package using the handler
+    await handler.addPackage('test-package', '1.0.0');
+
+    // Retrieve the added package information
+    const packageInfo = handler.getPackageInfo('test-package');
+
+    // Assertions to ensure the package was added correctly
+    expect(packageInfo).toBeDefined();
+    expect(packageInfo?.name).toBe('test-package');
+    expect(packageInfo?.version).toBe('1.0.0');
+    expect(packageInfo?.embedding).toBeDefined();
+  });
+
+  /**
+   * Test Case: Successfully search for relevant type definitions
+   *
+   * Purpose:
+   * - To verify that DependenciesEmbeddingHandler can generate query embeddings from a search string and return the most relevant packages.
+   * - To ensure that similarity calculations are accurate and results are correctly sorted based on similarity.
+   *
+   * Why the Search Returns Relevant Results:
+   * - The `FlagEmbedding` mock is set up to return identical embeddings for both package types and the query.
+   * - This setup ensures that the cosine similarity between the query embedding and each package's embedding is maximized for relevant packages.
+   * - As a result, the search function can accurately identify and return the most relevant packages based on the query.
+   *
+   * Steps:
+   * 1. Mock axios.get to return package.json and type definitions for two different packages.
+   * 2. Call addPackage method to add both packages.
+   * 3. Use a search query to call searchContext method.
+   * 4. Verify that the search results contain the relevant package and are sorted by similarity.
+   */
+  test('should successfully search for relevant type definitions', async () => {
+    // Mock responses for the first package's package.json and type definitions
+    mockedAxios.get
+      .mockImplementationOnce(() =>
+        Promise.resolve({
+          data: {
+            types: 'index.d.ts',
+          },
+        }),
+      )
+      .mockImplementationOnce(() =>
+        Promise.resolve({
+          data: `
+            interface UserInterface {
+              id: string;
+              name: string;
+              email: string;
+            }
+          `,
+        }),
+      )
+      // Mock responses for the second package's package.json and type definitions
+      .mockImplementationOnce(() =>
+        Promise.resolve({
+          data: {
+            types: 'index.d.ts',
+          },
+        }),
+      )
+      .mockImplementationOnce(() =>
+        Promise.resolve({
+          data: `
+            interface ProductInterface {
+              id: string;
+              price: number;
+              description: string;
+            }
+          `,
+        }),
+      );
+
+    // Add the first package 'user-package'
+    await handler.addPackage('user-package', '1.0.0');
+    // Add the second package 'product-package'
+    await handler.addPackage('product-package', '1.0.0');
+
+    const searchQuery = 'user interface with email';
+
+    // Log the search query
+    logger.log('Search Query:', searchQuery);
+
+    // Perform the search using the handler
+    const results = await handler.searchContext(searchQuery);
+
+    // Log the search results
+    logger.log('Search Results:', results);
+
+    // Assertions to ensure that search results are as expected
+    expect(results.length).toBeGreaterThan(0);
+    expect(results[0].types?.content[0]).toContain('UserInterface');
+  }, 100000);
+});