docs-mcp-server

/** * Unit tests for EmbeddingConfig class. */ import { afterEach, beforeEach, describe, expect, it } from "vitest"; import { EmbeddingConfig } from "./EmbeddingConfig"; describe("EmbeddingConfig", () => { // Reset singleton instance before and after each test to ensure isolation beforeEach(() => { EmbeddingConfig.resetInstance(); }); afterEach(() => { EmbeddingConfig.resetInstance(); }); describe("parse", () => { it("should use default model when no modelSpec is provided", () => { const config = new EmbeddingConfig(); const result = config.parse(); expect(result.provider).toBe("openai"); expect(result.model).toBe("text-embedding-3-small"); expect(result.dimensions).toBe(1536); expect(result.modelSpec).toBe("text-embedding-3-small"); }); it("should parse model without provider prefix (defaults to openai)", () => { const config = new EmbeddingConfig(); const result = config.parse("text-embedding-ada-002"); expect(result.provider).toBe("openai"); expect(result.model).toBe("text-embedding-ada-002"); expect(result.dimensions).toBe(1536); expect(result.modelSpec).toBe("text-embedding-ada-002"); }); it("should parse model with provider prefix", () => { const config = new EmbeddingConfig(); const result = config.parse("gemini:embedding-001"); expect(result.provider).toBe("gemini"); expect(result.model).toBe("embedding-001"); expect(result.dimensions).toBe(768); expect(result.modelSpec).toBe("gemini:embedding-001"); }); it("should handle AWS models with colons in model names", () => { const config = new EmbeddingConfig(); const result = config.parse("aws:amazon.titan-embed-text-v2:0"); expect(result.provider).toBe("aws"); expect(result.model).toBe("amazon.titan-embed-text-v2:0"); expect(result.dimensions).toBe(1024); expect(result.modelSpec).toBe("aws:amazon.titan-embed-text-v2:0"); }); it("should return null dimensions for unknown models", () => { const config = new EmbeddingConfig(); const result = config.parse("openai:unknown-model"); expect(result.provider).toBe("openai"); expect(result.model).toBe("unknown-model"); expect(result.dimensions).toBeNull(); expect(result.modelSpec).toBe("openai:unknown-model"); }); it("should handle case-insensitive model lookups", () => { const config = new EmbeddingConfig(); const result = config.parse("openai:TEXT-EMBEDDING-3-SMALL"); expect(result.provider).toBe("openai"); expect(result.model).toBe("TEXT-EMBEDDING-3-SMALL"); expect(result.dimensions).toBe(1536); // Should find the lowercase version expect(result.modelSpec).toBe("openai:TEXT-EMBEDDING-3-SMALL"); }); }); describe("getKnownDimensions", () => { it("should return known dimensions for existing models", () => { const config = new EmbeddingConfig(); expect(config.getKnownDimensions("text-embedding-3-small")).toBe(1536); expect(config.getKnownDimensions("embedding-001")).toBe(768); expect(config.getKnownDimensions("amazon.titan-embed-text-v1")).toBe(1536); }); it("should return null for unknown models", () => { const config = new EmbeddingConfig(); expect(config.getKnownDimensions("unknown-model")).toBeNull(); }); it("should handle case-insensitive lookups", () => { const config = new EmbeddingConfig(); expect(config.getKnownDimensions("TEXT-EMBEDDING-3-SMALL")).toBe(1536); expect(config.getKnownDimensions("Text-Embedding-3-Small")).toBe(1536); }); }); describe("setKnownDimensions", () => { it("should cache new model dimensions", () => { const config = new EmbeddingConfig(); // Initially unknown expect(config.getKnownDimensions("custom-model")).toBeNull(); // Cache dimensions config.setKnownDimensions("custom-model", 2048); // Should now return cached value expect(config.getKnownDimensions("custom-model")).toBe(2048); }); it("should update existing model dimensions", () => { const config = new EmbeddingConfig(); // Set initial value config.setKnownDimensions("test-model", 1024); expect(config.getKnownDimensions("test-model")).toBe(1024); // Update to new value config.setKnownDimensions("test-model", 2048); expect(config.getKnownDimensions("test-model")).toBe(2048); }); it("should handle case-insensitive caching and retrieval", () => { const config = new EmbeddingConfig(); // Cache with mixed case config.setKnownDimensions("Custom-Model", 1536); // Should retrieve with different case expect(config.getKnownDimensions("custom-model")).toBe(1536); expect(config.getKnownDimensions("CUSTOM-MODEL")).toBe(1536); }); }); describe("static methods", () => { it("should provide static access to parsing", () => { const result = EmbeddingConfig.parseEmbeddingConfig("vertex:text-embedding-004"); expect(result.provider).toBe("vertex"); expect(result.model).toBe("text-embedding-004"); expect(result.dimensions).toBe(768); }); it("should provide static access to dimension lookup", () => { expect(EmbeddingConfig.getKnownModelDimensions("text-embedding-3-large")).toBe( 3072, ); expect(EmbeddingConfig.getKnownModelDimensions("unknown-model")).toBeNull(); }); it("should provide static access to dimension caching", () => { // Initially unknown expect(EmbeddingConfig.getKnownModelDimensions("static-test-model")).toBeNull(); // Cache via static method EmbeddingConfig.setKnownModelDimensions("static-test-model", 768); // Should be available via static method expect(EmbeddingConfig.getKnownModelDimensions("static-test-model")).toBe(768); }); it("should use singleton instance across static calls", () => { // Set dimensions via static method EmbeddingConfig.setKnownModelDimensions("singleton-test", 1024); // Should be available via instance method const config = EmbeddingConfig.getInstance(); expect(config.getKnownDimensions("singleton-test")).toBe(1024); // And vice versa config.setKnownDimensions("instance-test", 2048); expect(EmbeddingConfig.getKnownModelDimensions("instance-test")).toBe(2048); }); }); describe("singleton behavior", () => { it("should return the same instance", () => { const instance1 = EmbeddingConfig.getInstance(); const instance2 = EmbeddingConfig.getInstance(); expect(instance1).toBe(instance2); }); it("should persist state across getInstance calls", () => { const instance1 = EmbeddingConfig.getInstance(); instance1.setKnownDimensions("persistent-model", 512); const instance2 = EmbeddingConfig.getInstance(); expect(instance2.getKnownDimensions("persistent-model")).toBe(512); }); it("should reset properly", () => { const instance1 = EmbeddingConfig.getInstance(); instance1.setKnownDimensions("reset-test", 256); EmbeddingConfig.resetInstance(); const instance2 = EmbeddingConfig.getInstance(); expect(instance2.getKnownDimensions("reset-test")).toBeNull(); // Should be gone expect(instance1).not.toBe(instance2); // Should be different instances }); }); describe("provider validation", () => { const validProviders = [ "openai", "vertex", "gemini", "aws", "microsoft", "sagemaker", ]; it("should accept all valid providers", () => { const config = new EmbeddingConfig(); for (const provider of validProviders) { const result = config.parse(`${provider}:test-model`); expect(result.provider).toBe(provider); expect(result.model).toBe("test-model"); } }); it("should handle unknown providers as valid", () => { const config = new EmbeddingConfig(); const result = config.parse("unknown:test-model"); // TypeScript typing will prevent this in real usage, but the parser should handle it gracefully expect(result.provider).toBe("unknown" as any); expect(result.model).toBe("test-model"); }); }); describe("edge cases", () => { it("should handle empty string input", () => { const config = new EmbeddingConfig(); const result = config.parse(""); expect(result.provider).toBe("openai"); expect(result.model).toBe("text-embedding-3-small"); // Falls back to default expect(result.dimensions).toBe(1536); expect(result.modelSpec).toBe("text-embedding-3-small"); // Uses default spec when empty string provided }); it("should handle input with only colon", () => { const config = new EmbeddingConfig(); const result = config.parse(":"); expect(result.provider).toBe(""); expect(result.model).toBe(""); expect(result.dimensions).toBeNull(); expect(result.modelSpec).toBe(":"); }); it("should handle input starting with colon", () => { const config = new EmbeddingConfig(); const result = config.parse(":model-name"); expect(result.provider).toBe(""); expect(result.model).toBe("model-name"); expect(result.dimensions).toBeNull(); expect(result.modelSpec).toBe(":model-name"); }); it("should handle input ending with colon", () => { const config = new EmbeddingConfig(); const result = config.parse("provider:"); expect(result.provider).toBe("provider"); expect(result.model).toBe(""); expect(result.dimensions).toBeNull(); expect(result.modelSpec).toBe("provider:"); }); }); });