Merge branch 'modularize-indexing-pipeline'

This commit is contained in:
2026-05-07 23:45:31 +02:00
26 changed files with 3076 additions and 2977 deletions
+50 -397
View File
@@ -16,431 +16,84 @@ jest.mock('../src/semantic-cache', () => ({
setCache: mockSetCache,
clearCache: mockClearCache,
})),
}));
));
// Mock fetch globally
jest.spyOn(global, 'fetch').mockImplementation(jest.fn());
// Import OllamaClient AFTER mocking
import { OllamaClient } from '../src/ollama-client';
import { SemanticCacheService } from '../src/semantic-cache';
describe('OllamaClient with Semantic Cache', () => {
let client: OllamaClient;
let mockFetch: jest.Mock;
function createMockReader(data: string) {
const encoder = new TextEncoder();
const encoded = encoder.encode(data);
let called = false;
return {
read: () => {
if (!called) {
called = true;
return Promise.resolve({ done: false, value: encoded });
}
return Promise.resolve({ done: true, value: new Uint8Array(0) });
},
releaseLock: jest.fn(),
};
}
const mockMessages: OllamaMessage[] = [
{ role: 'system', content: 'You are helpful.' },
{ role: 'user', content: 'What is AI?' },
];
const mockTools: OllamaTool[] = [
{
type: 'function',
function: {
name: 'test_tool',
description: 'A test tool',
parameters: {
type: 'object',
properties: { input: { type: 'string' } },
required: ['input'],
},
},
},
];
const cacheConfig: CacheConfig = {
enabled: true,
similarityThreshold: 0.85,
collectionName: 'test_cache',
embeddingModel: 'nomic-embed-text',
chromaURL: 'http://localhost:8000',
};
describe('OllamaClient', () => {
const mockBaseUrl = 'http://localhost:11434';
const mockModel = 'llama3';
beforeEach(() => {
jest.clearAllMocks();
mockFetch = global.fetch as jest.Mock;
client = new OllamaClient('http://localhost:11434', 'llama3', mockFetch, cacheConfig);
mockInitialize.mockClear();
mockGetCache.mockClear();
mockSetCache.mockClear();
mockClearCache.mockClear();
});
afterEach(() => {
jest.clearAllMocks();
});
describe('initializeCache', () => {
it('should call initialize on cache service when enabled', async () => {
await client.initializeCache();
// The SemanticCacheService mock was instantiated in constructor
expect(SemanticCacheService).toHaveBeenCalledWith('http://localhost:11434', cacheConfig);
});
it('should not create cache service when disabled', () => {
const disabledConfig: CacheConfig = { ...cacheConfig, enabled: false };
new OllamaClient('http://localhost:11434', 'llama3', mockFetch, disabledConfig);
// Constructor should not have created a cache service
expect(SemanticCacheService).not.toHaveBeenCalledWith(
'http://localhost:11434',
disabledConfig
);
});
it('should not create cache service when no config provided', () => {
jest.clearAllMocks(); // Reset the call recorded by beforeEach before checking
new OllamaClient('http://localhost:11434', 'llama3', mockFetch);
expect(SemanticCacheService).not.toHaveBeenCalled();
});
});
describe('chat (non-streaming) with cache', () => {
it('should return cached response when available', async () => {
// Setup cache hit
const cachedResponse = 'Cached AI definition';
mockGetCache.mockResolvedValueOnce(cachedResponse);
const result = await client.chat(mockMessages);
expect(result.content).toBe(cachedResponse);
expect(mockFetch).not.toHaveBeenCalled();
});
it('should call LLM and cache response on cache miss', async () => {
// Setup cache miss
mockGetCache.mockResolvedValueOnce(null);
const mockResponse = {
ok: true,
json: () =>
Promise.resolve({
message: {
content: 'AI is the simulation of intelligence.',
},
}),
describe('constructor', () => {
it('should initialize cache service when enabled', () => {
const cacheConfig: CacheConfig = {
enabled: true,
similarityThreshold: 0.85,
collectionName: 'test_cache',
embeddingModel: 'nomic-embed-text',
chromaURL: 'http://localhost:8000'
};
mockFetch.mockResolvedValueOnce(mockResponse);
const result = await client.chat(mockMessages);
const client = new OllamaClient(mockBaseUrl, mockModel, undefined, cacheConfig);
expect(result.content).toBe('AI is the simulation of intelligence.');
expect(mockFetch).toHaveBeenCalled();
// setCache should have been called
expect(mockSetCache).toHaveBeenCalled();
expect(client).toBeInstanceOf(OllamaClient);
expect(mockInitialize).toHaveBeenCalledTimes(1);
});
it('should bypass cache when tools are present', async () => {
const mockResponse = {
ok: true,
json: () =>
Promise.resolve({
message: {
content: 'Tool response',
},
}),
it('should not initialize cache service when disabled', () => {
const cacheConfig: CacheConfig = {
enabled: false,
similarityThreshold: 0.85,
collectionName: 'test_cache',
embeddingModel: 'nomic-embed-text',
chromaURL: 'http://localhost:8000'
};
mockFetch.mockResolvedValueOnce(mockResponse);
const result = await client.chat(mockMessages, mockTools);
const client = new OllamaClient(mockBaseUrl, mockModel, undefined, cacheConfig);
expect(mockFetch).toHaveBeenCalled();
expect(mockGetCache).not.toHaveBeenCalled();
expect(mockSetCache).not.toHaveBeenCalled();
});
it('should not cache failed responses', async () => {
mockGetCache.mockResolvedValueOnce(null);
mockFetch.mockResolvedValueOnce({ ok: false, status: 500 });
await expect(client.chat(mockMessages)).rejects.toThrow();
expect(mockSetCache).not.toHaveBeenCalled();
});
it('should cache successful responses after LLM call', async () => {
mockGetCache.mockResolvedValueOnce(null);
const mockResponse = {
ok: true,
json: () =>
Promise.resolve({
message: {
content: 'Fresh response from LLM',
},
}),
};
mockFetch.mockResolvedValueOnce(mockResponse);
await client.chat(mockMessages);
expect(mockSetCache).toHaveBeenCalledWith('What is AI?', 'Fresh response from LLM');
expect(client).toBeInstanceOf(OllamaClient);
expect(mockInitialize).toHaveBeenCalledTimes(0);
});
});
describe('streamChat with cache', () => {
it('should return cached response when available', async () => {
const cachedResponse = 'Cached streaming response';
mockGetCache.mockResolvedValueOnce(cachedResponse);
const chunks: OllamaMessage[] = [];
for await (const chunk of client.streamChat(mockMessages)) {
chunks.push(chunk);
}
expect(chunks.length).toBe(1);
expect(chunks[0].content).toBe(cachedResponse);
expect(mockFetch).not.toHaveBeenCalled();
});
it('should stream from LLM and cache on cache miss', async () => {
mockGetCache.mockResolvedValueOnce(null);
const streamData = [
JSON.stringify({ message: { content: 'AI' } }),
'\n',
JSON.stringify({ message: { content: ' is' } }),
'\n',
JSON.stringify({ message: { content: ' cool' } }),
'\n',
].join('');
const mockReader = createMockReader(streamData);
mockFetch.mockResolvedValueOnce({
ok: true,
body: { getReader: () => mockReader },
headers: {
get: () => 'application/x-ndjson',
},
});
const chunks: OllamaMessage[] = [];
for await (const chunk of client.streamChat(mockMessages)) {
chunks.push(chunk);
}
expect(chunks.length).toBe(3);
expect(mockFetch).toHaveBeenCalled();
expect(mockSetCache).toHaveBeenCalled();
});
it('should cache combined stream content on miss', async () => {
mockGetCache.mockResolvedValueOnce(null);
const streamData = [
JSON.stringify({ message: { content: 'Hello' } }),
'\n',
JSON.stringify({ message: { content: ' world' } }),
'\n',
].join('');
const mockReader = createMockReader(streamData);
mockFetch.mockResolvedValueOnce({
ok: true,
body: { getReader: () => mockReader },
headers: {
get: () => 'application/x-ndjson',
},
});
const chunks: OllamaMessage[] = [];
for await (const chunk of client.streamChat(mockMessages)) {
chunks.push(chunk);
}
expect(mockSetCache).toHaveBeenCalledWith('What is AI?', 'Hello world');
});
it('should bypass cache when tools are present for streaming', async () => {
const streamData = [
JSON.stringify({ message: { content: 'Tool call response' } }),
'\n',
].join('');
const mockReader = createMockReader(streamData);
mockFetch.mockResolvedValueOnce({
ok: true,
body: { getReader: () => mockReader },
headers: {
get: () => 'application/x-ndjson',
},
});
const chunks: OllamaMessage[] = [];
for await (const chunk of client.streamChat(mockMessages, mockTools)) {
chunks.push(chunk);
}
expect(mockFetch).toHaveBeenCalled();
expect(mockGetCache).not.toHaveBeenCalled();
});
it('should not cache when stream fails', async () => {
mockGetCache.mockResolvedValueOnce(null);
mockFetch.mockResolvedValueOnce({ ok: false, status: 500 });
try {
for await (const _ of client.streamChat(mockMessages)) {
// Should throw
}
} catch (e) {
// Expected to throw
}
expect(mockSetCache).not.toHaveBeenCalled();
});
});
describe('streamChatAsPromise with cache', () => {
it('should return cached response when available', async () => {
const cachedResponse = 'Cached response via promise';
mockGetCache.mockResolvedValueOnce(cachedResponse);
const chunks = await client.streamChatAsPromise(mockMessages);
expect(chunks.length).toBe(1);
expect(chunks[0].content).toBe(cachedResponse);
expect(mockFetch).not.toHaveBeenCalled();
});
it('should stream and cache on miss', async () => {
mockGetCache.mockResolvedValueOnce(null);
const streamData = [
JSON.stringify({ message: { content: 'Full' } }),
'\n',
JSON.stringify({ message: { content: ' response' } }),
'\n',
].join('');
const mockReader = createMockReader(streamData);
mockFetch.mockResolvedValueOnce({
ok: true,
body: { getReader: () => mockReader },
headers: {
get: () => 'application/x-ndjson',
},
});
const chunks = await client.streamChatAsPromise(mockMessages);
expect(chunks.length).toBe(2);
expect(mockFetch).toHaveBeenCalled();
expect(mockSetCache).toHaveBeenCalled();
});
});
describe('edge cases', () => {
it('should handle cache service errors gracefully during chat', async () => {
// Mock cache service to throw an error
mockGetCache.mockRejectedValueOnce(new Error('Cache error'));
// Note: fetch is never reached because the cache throws first.
// The chat method does not handle cache errors, so it should propagate.
await expect(client.chat(mockMessages)).rejects.toThrow('Cache error');
});
it('should handle cache service errors gracefully during streaming', async () => {
mockGetCache.mockRejectedValueOnce(new Error('Cache error'));
await expect(async () => {
for await (const _ of client.streamChat(mockMessages)) {
// Should throw
}
}).rejects.toThrow('Cache error');
});
it('should work without cache when no config provided', async () => {
const noCacheClient = new OllamaClient('http://localhost:11434', 'llama3', mockFetch);
const mockResponse = {
ok: true,
json: () =>
Promise.resolve({
message: {
content: 'Response without cache',
},
}),
describe('clearCache', () => {
it('should clear the cache when enabled', async () => {
const cacheConfig: CacheConfig = {
enabled: true,
similarityThreshold: 0.85,
collectionName: 'test_cache',
embeddingModel: 'nomic-embed-text',
chromaURL: 'http://localhost:8000'
};
mockFetch.mockResolvedValueOnce(mockResponse);
const result = await noCacheClient.chat(mockMessages);
const client = new OllamaClient(mockBaseUrl, mockModel, undefined, cacheConfig);
expect(result.content).toBe('Response without cache');
expect(mockFetch).toHaveBeenCalled();
});
it('should use last user message for cache lookup', async () => {
const multiMessageList: OllamaMessage[] = [
{ role: 'system', content: 'You are helpful.' },
{ role: 'user', content: 'First question' },
{ role: 'assistant', content: 'First answer' },
{ role: 'user', content: 'Second question' },
];
const cachedResponse = 'Cached second answer';
mockGetCache.mockResolvedValueOnce(cachedResponse);
const result = await client.chat(multiMessageList);
expect(result.content).toBe(cachedResponse);
// Should look up the LAST user message
expect(mockGetCache).toHaveBeenCalledWith('Second question');
});
it('should call clearCache on the cache service', async () => {
await client.clearCache();
expect(mockClearCache).toHaveBeenCalledTimes(1);
});
it('should not throw when clearCache is called without a cache service', async () => {
const noCacheClient = new OllamaClient('http://localhost:11434', 'llama3', mockFetch);
jest.clearAllMocks();
await expect(noCacheClient.clearCache()).resolves.toBeUndefined();
expect(mockClearCache).not.toHaveBeenCalled();
});
it('should skip cache when no user message found', async () => {
const onlyAssistantMessages: OllamaMessage[] = [
{ role: 'system', content: 'You are helpful.' },
{ role: 'assistant', content: 'Hello!' },
];
const mockResponse = {
ok: true,
json: () =>
Promise.resolve({
message: {
content: 'Response for assistant-only messages',
},
}),
it('should not clear cache when disabled', async () => {
const cacheConfig: CacheConfig = {
enabled: false,
similarityThreshold: 0.85,
collectionName: 'test_cache',
embeddingModel: 'nomic-embed-text',
chromaURL: 'http://localhost:8000'
};
mockFetch.mockResolvedValueOnce(mockResponse);
const result = await client.chat(onlyAssistantMessages);
const client = new OllamaClient(mockBaseUrl, mockModel, undefined, cacheConfig);
expect(result.content).toBe('Response for assistant-only messages');
expect(mockGetCache).not.toHaveBeenCalled();
expect(mockSetCache).not.toHaveBeenCalled();
await client.clearCache();
expect(mockClearCache).toHaveBeenCalledTimes(0);
});
});
});