ThoughtMCP

/** * Tests for Hybrid Embedding Model * * Verifies the hybrid embedding model correctly switches between * real Ollama and mock embeddings based on availability. */ import { beforeEach, describe, expect, it } from "vitest"; import { createHybridEmbeddingModel, getUnitTestModel } from "../../utils/hybrid-embedding-model"; import { MockOllamaEmbeddingModel } from "../../utils/mock-embeddings"; import { clearOllamaStatusCache } from "../../utils/ollama-detector"; describe("Hybrid Embedding Model", () => { beforeEach(() => { // Clear cached Ollama status between tests clearOllamaStatusCache(); // Reset environment variables delete process.env.USE_MOCK_EMBEDDINGS; delete process.env.USE_REAL_OLLAMA; }); describe("getUnitTestModel", () => { it("should return a MockOllamaEmbeddingModel", () => { const model = getUnitTestModel(); expect(model).toBeInstanceOf(MockOllamaEmbeddingModel); expect(model.getDimension()).toBe(768); expect(model.getModelName()).toBe("nomic-embed-text"); }); it("should respect custom configuration", () => { const model = getUnitTestModel({ modelName: "custom-model", dimension: 1024, }); expect(model.getDimension()).toBe(1024); expect(model.getModelName()).toBe("custom-model"); }); it("should generate deterministic embeddings", async () => { const model = getUnitTestModel(); const embedding1 = await model.generate("test text"); const embedding2 = await model.generate("test text"); expect(embedding1).toEqual(embedding2); expect(embedding1.length).toBe(768); }); it("should generate different embeddings for different text", async () => { const model = getUnitTestModel(); const embedding1 = await model.generate("first text"); const embedding2 = await model.generate("second text"); expect(embedding1).not.toEqual(embedding2); }); }); describe("createHybridEmbeddingModel", () => { it("should return mock mode when USE_MOCK_EMBEDDINGS is true", async () => { process.env.USE_MOCK_EMBEDDINGS = "true"; const result = await createHybridEmbeddingModel(); expect(result.mode).toBe("mock"); expect(result.reason).toBe("USE_MOCK_EMBEDDINGS=true"); }); it("should return real mode when USE_REAL_OLLAMA is true", async () => { process.env.USE_REAL_OLLAMA = "true"; const result = await createHybridEmbeddingModel(); expect(result.mode).toBe("real"); expect(result.reason).toBe("USE_REAL_OLLAMA=true"); }); it("should return mock mode when forceMode is mock", async () => { const result = await createHybridEmbeddingModel({ forceMode: "mock" }); expect(result.mode).toBe("mock"); expect(result.reason).toBe("Forced mock mode"); }); it("should return real mode when forceMode is real", async () => { const result = await createHybridEmbeddingModel({ forceMode: "real" }); expect(result.mode).toBe("real"); expect(result.reason).toBe("Forced real mode"); }); it("should respect custom configuration in mock mode", async () => { const result = await createHybridEmbeddingModel({ forceMode: "mock", modelName: "custom-model", dimension: 512, }); expect(result.model.getDimension()).toBe(512); expect(result.model.getModelName()).toBe("custom-model"); }); }); describe("MockOllamaEmbeddingModel with cache", () => { it("should track cache statistics", async () => { const model = getUnitTestModel(); // Generate some embeddings await model.generate("text 1"); await model.generate("text 2"); await model.generate("text 1"); // Should hit local cache const stats = model.getCacheStats(); // All should be misses from the pre-computed cache (no cache file exists yet) // but the local cache should have 2 entries expect(model.getCacheSize()).toBe(2); expect(stats.misses).toBeGreaterThanOrEqual(2); }); it("should clear cache and statistics", async () => { const model = getUnitTestModel(); await model.generate("test"); expect(model.getCacheSize()).toBe(1); model.clearCache(); expect(model.getCacheSize()).toBe(0); const stats = model.getCacheStats(); expect(stats.hits).toBe(0); expect(stats.misses).toBe(0); }); it("should validate input like real OllamaEmbeddingModel", async () => { const model = getUnitTestModel(); await expect(model.generate("")).rejects.toThrow(); await expect(model.generate(" ")).rejects.toThrow(); await expect(model.generate(null as unknown as string)).rejects.toThrow(); }); it("should return normalized unit vectors", async () => { const model = getUnitTestModel(); const embedding = await model.generate("test text"); // Calculate magnitude const magnitude = Math.sqrt(embedding.reduce((sum, val) => sum + val * val, 0)); // Should be approximately 1 (unit vector) expect(magnitude).toBeCloseTo(1, 5); }); }); });