feat(#59): implement wiki-link parser
- Created wiki-link-parser.ts utility for parsing [[links]] syntax - Supports multiple formats: [[Page Name]], [[Page|display]], [[slug]] - Returns parsed links with target, display text, and position info - Handles edge cases: nested brackets, escaped brackets, code blocks - Code block awareness: skips links in inline code, fenced blocks, and indented code - Comprehensive test suite with 43 passing tests (100% coverage) - Updated README.md with parser documentation Implements KNOW-007 (Issue #59) - Wiki-style linking foundation
This commit is contained in:
Jason Woltje
59
apps/api/src/ollama/dto/index.ts
Normal file
59
apps/api/src/ollama/dto/index.ts
Normal file
@@ -0,0 +1,59 @@
|
||||
/**
|
||||
* DTOs for Ollama module
|
||||
*/
|
||||
|
||||
export interface GenerateOptionsDto {
|
||||
temperature?: number;
|
||||
top_p?: number;
|
||||
max_tokens?: number;
|
||||
stop?: string[];
|
||||
stream?: boolean;
|
||||
}
|
||||
|
||||
export interface ChatMessage {
|
||||
role: 'system' | 'user' | 'assistant';
|
||||
content: string;
|
||||
}
|
||||
|
||||
export interface ChatOptionsDto {
|
||||
temperature?: number;
|
||||
top_p?: number;
|
||||
max_tokens?: number;
|
||||
stop?: string[];
|
||||
stream?: boolean;
|
||||
}
|
||||
|
||||
export interface GenerateResponseDto {
|
||||
response: string;
|
||||
model: string;
|
||||
done: boolean;
|
||||
}
|
||||
|
||||
export interface ChatResponseDto {
|
||||
message: ChatMessage;
|
||||
model: string;
|
||||
done: boolean;
|
||||
}
|
||||
|
||||
export interface EmbedResponseDto {
|
||||
embedding: number[];
|
||||
}
|
||||
|
||||
export interface OllamaModel {
|
||||
name: string;
|
||||
modified_at: string;
|
||||
size: number;
|
||||
digest: string;
|
||||
}
|
||||
|
||||
export interface ListModelsResponseDto {
|
||||
models: OllamaModel[];
|
||||
}
|
||||
|
||||
export interface HealthCheckResponseDto {
|
||||
status: 'healthy' | 'unhealthy';
|
||||
mode: 'local' | 'remote';
|
||||
endpoint: string;
|
||||
available: boolean;
|
||||
error?: string;
|
||||
}
|
||||
243
apps/api/src/ollama/ollama.controller.spec.ts
Normal file
243
apps/api/src/ollama/ollama.controller.spec.ts
Normal file
@@ -0,0 +1,243 @@
|
||||
import { describe, it, expect, beforeEach, vi } from "vitest";
|
||||
import { Test, TestingModule } from "@nestjs/testing";
|
||||
import { OllamaController } from "./ollama.controller";
|
||||
import { OllamaService } from "./ollama.service";
|
||||
import type { ChatMessage } from "./dto";
|
||||
|
||||
describe("OllamaController", () => {
|
||||
let controller: OllamaController;
|
||||
let service: OllamaService;
|
||||
|
||||
const mockOllamaService = {
|
||||
generate: vi.fn(),
|
||||
chat: vi.fn(),
|
||||
embed: vi.fn(),
|
||||
listModels: vi.fn(),
|
||||
healthCheck: vi.fn(),
|
||||
};
|
||||
|
||||
beforeEach(async () => {
|
||||
const module: TestingModule = await Test.createTestingModule({
|
||||
controllers: [OllamaController],
|
||||
providers: [
|
||||
{
|
||||
provide: OllamaService,
|
||||
useValue: mockOllamaService,
|
||||
},
|
||||
],
|
||||
}).compile();
|
||||
|
||||
controller = module.get<OllamaController>(OllamaController);
|
||||
service = module.get<OllamaService>(OllamaService);
|
||||
|
||||
vi.clearAllMocks();
|
||||
});
|
||||
|
||||
describe("generate", () => {
|
||||
it("should generate text from prompt", async () => {
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
response: "Generated text",
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockOllamaService.generate.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.generate({
|
||||
prompt: "Hello",
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.generate).toHaveBeenCalledWith(
|
||||
"Hello",
|
||||
undefined,
|
||||
undefined
|
||||
);
|
||||
});
|
||||
|
||||
it("should generate with options and custom model", async () => {
|
||||
const mockResponse = {
|
||||
model: "mistral",
|
||||
response: "Response",
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockOllamaService.generate.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.generate({
|
||||
prompt: "Test",
|
||||
model: "mistral",
|
||||
options: {
|
||||
temperature: 0.7,
|
||||
max_tokens: 100,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.generate).toHaveBeenCalledWith(
|
||||
"Test",
|
||||
{ temperature: 0.7, max_tokens: 100 },
|
||||
"mistral"
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe("chat", () => {
|
||||
it("should complete chat conversation", async () => {
|
||||
const messages: ChatMessage[] = [
|
||||
{ role: "user", content: "Hello!" },
|
||||
];
|
||||
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
message: {
|
||||
role: "assistant",
|
||||
content: "Hi there!",
|
||||
},
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockOllamaService.chat.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.chat({
|
||||
messages,
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.chat).toHaveBeenCalledWith(
|
||||
messages,
|
||||
undefined,
|
||||
undefined
|
||||
);
|
||||
});
|
||||
|
||||
it("should chat with options and custom model", async () => {
|
||||
const messages: ChatMessage[] = [
|
||||
{ role: "system", content: "You are helpful." },
|
||||
{ role: "user", content: "Hello!" },
|
||||
];
|
||||
|
||||
const mockResponse = {
|
||||
model: "mistral",
|
||||
message: {
|
||||
role: "assistant",
|
||||
content: "Hello!",
|
||||
},
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockOllamaService.chat.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.chat({
|
||||
messages,
|
||||
model: "mistral",
|
||||
options: {
|
||||
temperature: 0.5,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.chat).toHaveBeenCalledWith(
|
||||
messages,
|
||||
{ temperature: 0.5 },
|
||||
"mistral"
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe("embed", () => {
|
||||
it("should generate embeddings", async () => {
|
||||
const mockResponse = {
|
||||
embedding: [0.1, 0.2, 0.3],
|
||||
};
|
||||
|
||||
mockOllamaService.embed.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.embed({
|
||||
text: "Sample text",
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.embed).toHaveBeenCalledWith(
|
||||
"Sample text",
|
||||
undefined
|
||||
);
|
||||
});
|
||||
|
||||
it("should embed with custom model", async () => {
|
||||
const mockResponse = {
|
||||
embedding: [0.1, 0.2],
|
||||
};
|
||||
|
||||
mockOllamaService.embed.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.embed({
|
||||
text: "Test",
|
||||
model: "nomic-embed-text",
|
||||
});
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.embed).toHaveBeenCalledWith(
|
||||
"Test",
|
||||
"nomic-embed-text"
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe("listModels", () => {
|
||||
it("should list available models", async () => {
|
||||
const mockResponse = {
|
||||
models: [
|
||||
{
|
||||
name: "llama3.2:latest",
|
||||
modified_at: "2024-01-15T10:00:00Z",
|
||||
size: 4500000000,
|
||||
digest: "abc123",
|
||||
},
|
||||
],
|
||||
};
|
||||
|
||||
mockOllamaService.listModels.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.listModels();
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.listModels).toHaveBeenCalled();
|
||||
});
|
||||
});
|
||||
|
||||
describe("healthCheck", () => {
|
||||
it("should return health status", async () => {
|
||||
const mockResponse = {
|
||||
status: "healthy" as const,
|
||||
mode: "local" as const,
|
||||
endpoint: "http://localhost:11434",
|
||||
available: true,
|
||||
};
|
||||
|
||||
mockOllamaService.healthCheck.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.healthCheck();
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockOllamaService.healthCheck).toHaveBeenCalled();
|
||||
});
|
||||
|
||||
it("should return unhealthy status", async () => {
|
||||
const mockResponse = {
|
||||
status: "unhealthy" as const,
|
||||
mode: "local" as const,
|
||||
endpoint: "http://localhost:11434",
|
||||
available: false,
|
||||
error: "Connection refused",
|
||||
};
|
||||
|
||||
mockOllamaService.healthCheck.mockResolvedValue(mockResponse);
|
||||
|
||||
const result = await controller.healthCheck();
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(result.status).toBe("unhealthy");
|
||||
});
|
||||
});
|
||||
});
|
||||
92
apps/api/src/ollama/ollama.controller.ts
Normal file
92
apps/api/src/ollama/ollama.controller.ts
Normal file
@@ -0,0 +1,92 @@
|
||||
import { Controller, Post, Get, Body } from "@nestjs/common";
|
||||
import { OllamaService } from "./ollama.service";
|
||||
import type {
|
||||
GenerateOptionsDto,
|
||||
GenerateResponseDto,
|
||||
ChatMessage,
|
||||
ChatOptionsDto,
|
||||
ChatResponseDto,
|
||||
EmbedResponseDto,
|
||||
ListModelsResponseDto,
|
||||
HealthCheckResponseDto,
|
||||
} from "./dto";
|
||||
|
||||
/**
|
||||
* Request DTO for generate endpoint
|
||||
*/
|
||||
interface GenerateRequestDto {
|
||||
prompt: string;
|
||||
options?: GenerateOptionsDto;
|
||||
model?: string;
|
||||
}
|
||||
|
||||
/**
|
||||
* Request DTO for chat endpoint
|
||||
*/
|
||||
interface ChatRequestDto {
|
||||
messages: ChatMessage[];
|
||||
options?: ChatOptionsDto;
|
||||
model?: string;
|
||||
}
|
||||
|
||||
/**
|
||||
* Request DTO for embed endpoint
|
||||
*/
|
||||
interface EmbedRequestDto {
|
||||
text: string;
|
||||
model?: string;
|
||||
}
|
||||
|
||||
/**
|
||||
* Controller for Ollama API endpoints
|
||||
* Provides text generation, chat, embeddings, and model management
|
||||
*/
|
||||
@Controller("ollama")
|
||||
export class OllamaController {
|
||||
constructor(private readonly ollamaService: OllamaService) {}
|
||||
|
||||
/**
|
||||
* Generate text from a prompt
|
||||
* POST /ollama/generate
|
||||
*/
|
||||
@Post("generate")
|
||||
async generate(@Body() body: GenerateRequestDto): Promise<GenerateResponseDto> {
|
||||
return this.ollamaService.generate(body.prompt, body.options, body.model);
|
||||
}
|
||||
|
||||
/**
|
||||
* Complete a chat conversation
|
||||
* POST /ollama/chat
|
||||
*/
|
||||
@Post("chat")
|
||||
async chat(@Body() body: ChatRequestDto): Promise<ChatResponseDto> {
|
||||
return this.ollamaService.chat(body.messages, body.options, body.model);
|
||||
}
|
||||
|
||||
/**
|
||||
* Generate embeddings for text
|
||||
* POST /ollama/embed
|
||||
*/
|
||||
@Post("embed")
|
||||
async embed(@Body() body: EmbedRequestDto): Promise<EmbedResponseDto> {
|
||||
return this.ollamaService.embed(body.text, body.model);
|
||||
}
|
||||
|
||||
/**
|
||||
* List available models
|
||||
* GET /ollama/models
|
||||
*/
|
||||
@Get("models")
|
||||
async listModels(): Promise<ListModelsResponseDto> {
|
||||
return this.ollamaService.listModels();
|
||||
}
|
||||
|
||||
/**
|
||||
* Health check endpoint
|
||||
* GET /ollama/health
|
||||
*/
|
||||
@Get("health")
|
||||
async healthCheck(): Promise<HealthCheckResponseDto> {
|
||||
return this.ollamaService.healthCheck();
|
||||
}
|
||||
}
|
||||
37
apps/api/src/ollama/ollama.module.ts
Normal file
37
apps/api/src/ollama/ollama.module.ts
Normal file
@@ -0,0 +1,37 @@
|
||||
import { Module } from "@nestjs/common";
|
||||
import { OllamaController } from "./ollama.controller";
|
||||
import { OllamaService, OllamaConfig } from "./ollama.service";
|
||||
|
||||
/**
|
||||
* Factory function to create Ollama configuration from environment variables
|
||||
*/
|
||||
function createOllamaConfig(): OllamaConfig {
|
||||
const mode = (process.env.OLLAMA_MODE || "local") as "local" | "remote";
|
||||
const endpoint = process.env.OLLAMA_ENDPOINT || "http://localhost:11434";
|
||||
const model = process.env.OLLAMA_MODEL || "llama3.2";
|
||||
const timeout = parseInt(process.env.OLLAMA_TIMEOUT || "30000", 10);
|
||||
|
||||
return {
|
||||
mode,
|
||||
endpoint,
|
||||
model,
|
||||
timeout,
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* Module for Ollama integration
|
||||
* Provides AI capabilities via local or remote Ollama instances
|
||||
*/
|
||||
@Module({
|
||||
controllers: [OllamaController],
|
||||
providers: [
|
||||
{
|
||||
provide: "OLLAMA_CONFIG",
|
||||
useFactory: createOllamaConfig,
|
||||
},
|
||||
OllamaService,
|
||||
],
|
||||
exports: [OllamaService],
|
||||
})
|
||||
export class OllamaModule {}
|
||||
441
apps/api/src/ollama/ollama.service.spec.ts
Normal file
441
apps/api/src/ollama/ollama.service.spec.ts
Normal file
@@ -0,0 +1,441 @@
|
||||
import { describe, it, expect, beforeEach, vi } from "vitest";
|
||||
import { Test, TestingModule } from "@nestjs/testing";
|
||||
import { OllamaService } from "./ollama.service";
|
||||
import { HttpException, HttpStatus } from "@nestjs/common";
|
||||
import type {
|
||||
GenerateOptionsDto,
|
||||
ChatMessage,
|
||||
ChatOptionsDto,
|
||||
} from "./dto";
|
||||
|
||||
describe("OllamaService", () => {
|
||||
let service: OllamaService;
|
||||
let mockFetch: ReturnType<typeof vi.fn>;
|
||||
|
||||
const mockConfig = {
|
||||
mode: "local" as const,
|
||||
endpoint: "http://localhost:11434",
|
||||
model: "llama3.2",
|
||||
timeout: 30000,
|
||||
};
|
||||
|
||||
beforeEach(async () => {
|
||||
mockFetch = vi.fn();
|
||||
global.fetch = mockFetch;
|
||||
|
||||
const module: TestingModule = await Test.createTestingModule({
|
||||
providers: [
|
||||
OllamaService,
|
||||
{
|
||||
provide: "OLLAMA_CONFIG",
|
||||
useValue: mockConfig,
|
||||
},
|
||||
],
|
||||
}).compile();
|
||||
|
||||
service = module.get<OllamaService>(OllamaService);
|
||||
|
||||
vi.clearAllMocks();
|
||||
});
|
||||
|
||||
describe("generate", () => {
|
||||
it("should generate text from prompt", async () => {
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
response: "This is a generated response.",
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.generate("Hello, world!");
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/generate",
|
||||
expect.objectContaining({
|
||||
method: "POST",
|
||||
headers: { "Content-Type": "application/json" },
|
||||
body: JSON.stringify({
|
||||
model: "llama3.2",
|
||||
prompt: "Hello, world!",
|
||||
stream: false,
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should generate text with custom options", async () => {
|
||||
const options: GenerateOptionsDto = {
|
||||
temperature: 0.8,
|
||||
max_tokens: 100,
|
||||
stop: ["\n"],
|
||||
};
|
||||
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
response: "Custom response.",
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.generate("Hello", options);
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/generate",
|
||||
expect.objectContaining({
|
||||
body: JSON.stringify({
|
||||
model: "llama3.2",
|
||||
prompt: "Hello",
|
||||
stream: false,
|
||||
options: {
|
||||
temperature: 0.8,
|
||||
num_predict: 100,
|
||||
stop: ["\n"],
|
||||
},
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should use custom model when provided", async () => {
|
||||
const mockResponse = {
|
||||
model: "mistral",
|
||||
response: "Response from mistral.",
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.generate("Hello", {}, "mistral");
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
const callArgs = mockFetch.mock.calls[0];
|
||||
expect(callArgs[0]).toBe("http://localhost:11434/api/generate");
|
||||
const body = JSON.parse(callArgs[1].body as string);
|
||||
expect(body.model).toBe("mistral");
|
||||
expect(body.prompt).toBe("Hello");
|
||||
expect(body.stream).toBe(false);
|
||||
});
|
||||
|
||||
it("should throw HttpException on network error", async () => {
|
||||
mockFetch.mockRejectedValue(new Error("Network error"));
|
||||
|
||||
await expect(service.generate("Hello")).rejects.toThrow(HttpException);
|
||||
await expect(service.generate("Hello")).rejects.toThrow(
|
||||
"Failed to connect to Ollama"
|
||||
);
|
||||
});
|
||||
|
||||
it("should throw HttpException on non-ok response", async () => {
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: false,
|
||||
status: 500,
|
||||
statusText: "Internal Server Error",
|
||||
});
|
||||
|
||||
await expect(service.generate("Hello")).rejects.toThrow(HttpException);
|
||||
});
|
||||
|
||||
it("should handle timeout", async () => {
|
||||
// Mock AbortController to simulate timeout
|
||||
mockFetch.mockRejectedValue(new Error("The operation was aborted"));
|
||||
|
||||
// Create service with very short timeout
|
||||
const shortTimeoutModule = await Test.createTestingModule({
|
||||
providers: [
|
||||
OllamaService,
|
||||
{
|
||||
provide: "OLLAMA_CONFIG",
|
||||
useValue: { ...mockConfig, timeout: 1 },
|
||||
},
|
||||
],
|
||||
}).compile();
|
||||
|
||||
const shortTimeoutService =
|
||||
shortTimeoutModule.get<OllamaService>(OllamaService);
|
||||
|
||||
await expect(shortTimeoutService.generate("Hello")).rejects.toThrow(
|
||||
HttpException
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe("chat", () => {
|
||||
it("should complete chat with messages", async () => {
|
||||
const messages: ChatMessage[] = [
|
||||
{ role: "system", content: "You are helpful." },
|
||||
{ role: "user", content: "Hello!" },
|
||||
];
|
||||
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
message: {
|
||||
role: "assistant",
|
||||
content: "Hello! How can I help you?",
|
||||
},
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.chat(messages);
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/chat",
|
||||
expect.objectContaining({
|
||||
method: "POST",
|
||||
body: JSON.stringify({
|
||||
model: "llama3.2",
|
||||
messages,
|
||||
stream: false,
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should chat with custom options", async () => {
|
||||
const messages: ChatMessage[] = [
|
||||
{ role: "user", content: "Hello!" },
|
||||
];
|
||||
|
||||
const options: ChatOptionsDto = {
|
||||
temperature: 0.5,
|
||||
max_tokens: 50,
|
||||
};
|
||||
|
||||
const mockResponse = {
|
||||
model: "llama3.2",
|
||||
message: { role: "assistant", content: "Hi!" },
|
||||
done: true,
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
await service.chat(messages, options);
|
||||
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/chat",
|
||||
expect.objectContaining({
|
||||
body: JSON.stringify({
|
||||
model: "llama3.2",
|
||||
messages,
|
||||
stream: false,
|
||||
options: {
|
||||
temperature: 0.5,
|
||||
num_predict: 50,
|
||||
},
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should throw HttpException on chat error", async () => {
|
||||
mockFetch.mockRejectedValue(new Error("Connection refused"));
|
||||
|
||||
await expect(
|
||||
service.chat([{ role: "user", content: "Hello" }])
|
||||
).rejects.toThrow(HttpException);
|
||||
});
|
||||
});
|
||||
|
||||
describe("embed", () => {
|
||||
it("should generate embeddings for text", async () => {
|
||||
const mockResponse = {
|
||||
embedding: [0.1, 0.2, 0.3, 0.4],
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.embed("Hello world");
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/embeddings",
|
||||
expect.objectContaining({
|
||||
method: "POST",
|
||||
body: JSON.stringify({
|
||||
model: "llama3.2",
|
||||
prompt: "Hello world",
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should use custom model for embeddings", async () => {
|
||||
const mockResponse = {
|
||||
embedding: [0.1, 0.2],
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
await service.embed("Test", "nomic-embed-text");
|
||||
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/embeddings",
|
||||
expect.objectContaining({
|
||||
body: JSON.stringify({
|
||||
model: "nomic-embed-text",
|
||||
prompt: "Test",
|
||||
}),
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should throw HttpException on embed error", async () => {
|
||||
mockFetch.mockRejectedValue(new Error("Model not found"));
|
||||
|
||||
await expect(service.embed("Hello")).rejects.toThrow(HttpException);
|
||||
});
|
||||
});
|
||||
|
||||
describe("listModels", () => {
|
||||
it("should list available models", async () => {
|
||||
const mockResponse = {
|
||||
models: [
|
||||
{
|
||||
name: "llama3.2:latest",
|
||||
modified_at: "2024-01-15T10:00:00Z",
|
||||
size: 4500000000,
|
||||
digest: "abc123",
|
||||
},
|
||||
{
|
||||
name: "mistral:latest",
|
||||
modified_at: "2024-01-14T09:00:00Z",
|
||||
size: 4200000000,
|
||||
digest: "def456",
|
||||
},
|
||||
],
|
||||
};
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => mockResponse,
|
||||
});
|
||||
|
||||
const result = await service.listModels();
|
||||
|
||||
expect(result).toEqual(mockResponse);
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://localhost:11434/api/tags",
|
||||
expect.objectContaining({
|
||||
method: "GET",
|
||||
})
|
||||
);
|
||||
});
|
||||
|
||||
it("should throw HttpException when listing fails", async () => {
|
||||
mockFetch.mockRejectedValue(new Error("Server error"));
|
||||
|
||||
await expect(service.listModels()).rejects.toThrow(HttpException);
|
||||
});
|
||||
});
|
||||
|
||||
describe("healthCheck", () => {
|
||||
it("should return healthy status when Ollama is available", async () => {
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => ({ status: "ok" }),
|
||||
});
|
||||
|
||||
const result = await service.healthCheck();
|
||||
|
||||
expect(result).toEqual({
|
||||
status: "healthy",
|
||||
mode: "local",
|
||||
endpoint: "http://localhost:11434",
|
||||
available: true,
|
||||
});
|
||||
});
|
||||
|
||||
it("should return unhealthy status when Ollama is unavailable", async () => {
|
||||
mockFetch.mockRejectedValue(new Error("Connection refused"));
|
||||
|
||||
const result = await service.healthCheck();
|
||||
|
||||
expect(result).toEqual({
|
||||
status: "unhealthy",
|
||||
mode: "local",
|
||||
endpoint: "http://localhost:11434",
|
||||
available: false,
|
||||
error: "Connection refused",
|
||||
});
|
||||
});
|
||||
|
||||
it("should handle non-ok response in health check", async () => {
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: false,
|
||||
status: 503,
|
||||
statusText: "Service Unavailable",
|
||||
});
|
||||
|
||||
const result = await service.healthCheck();
|
||||
|
||||
expect(result.status).toBe("unhealthy");
|
||||
expect(result.available).toBe(false);
|
||||
});
|
||||
});
|
||||
|
||||
describe("configuration", () => {
|
||||
it("should use remote mode configuration", async () => {
|
||||
const remoteConfig = {
|
||||
mode: "remote" as const,
|
||||
endpoint: "http://remote-server:11434",
|
||||
model: "mistral",
|
||||
timeout: 60000,
|
||||
};
|
||||
|
||||
const remoteModule = await Test.createTestingModule({
|
||||
providers: [
|
||||
OllamaService,
|
||||
{
|
||||
provide: "OLLAMA_CONFIG",
|
||||
useValue: remoteConfig,
|
||||
},
|
||||
],
|
||||
}).compile();
|
||||
|
||||
const remoteService = remoteModule.get<OllamaService>(OllamaService);
|
||||
|
||||
mockFetch.mockResolvedValue({
|
||||
ok: true,
|
||||
json: async () => ({
|
||||
model: "mistral",
|
||||
response: "Remote response",
|
||||
done: true,
|
||||
}),
|
||||
});
|
||||
|
||||
await remoteService.generate("Test");
|
||||
|
||||
expect(mockFetch).toHaveBeenCalledWith(
|
||||
"http://remote-server:11434/api/generate",
|
||||
expect.any(Object)
|
||||
);
|
||||
});
|
||||
});
|
||||
});
|
||||
344
apps/api/src/ollama/ollama.service.ts
Normal file
344
apps/api/src/ollama/ollama.service.ts
Normal file
@@ -0,0 +1,344 @@
|
||||
import { Injectable, Inject, HttpException, HttpStatus } from "@nestjs/common";
|
||||
import type {
|
||||
GenerateOptionsDto,
|
||||
GenerateResponseDto,
|
||||
ChatMessage,
|
||||
ChatOptionsDto,
|
||||
ChatResponseDto,
|
||||
EmbedResponseDto,
|
||||
ListModelsResponseDto,
|
||||
HealthCheckResponseDto,
|
||||
} from "./dto";
|
||||
|
||||
/**
|
||||
* Configuration for Ollama service
|
||||
*/
|
||||
export interface OllamaConfig {
|
||||
mode: "local" | "remote";
|
||||
endpoint: string;
|
||||
model: string;
|
||||
timeout: number;
|
||||
}
|
||||
|
||||
/**
|
||||
* Service for interacting with Ollama API
|
||||
* Supports both local and remote Ollama instances
|
||||
*/
|
||||
@Injectable()
|
||||
export class OllamaService {
|
||||
constructor(
|
||||
@Inject("OLLAMA_CONFIG")
|
||||
private readonly config: OllamaConfig
|
||||
) {}
|
||||
|
||||
/**
|
||||
* Generate text from a prompt
|
||||
* @param prompt - The text prompt to generate from
|
||||
* @param options - Generation options (temperature, max_tokens, etc.)
|
||||
* @param model - Optional model override (defaults to config model)
|
||||
* @returns Generated text response
|
||||
*/
|
||||
async generate(
|
||||
prompt: string,
|
||||
options?: GenerateOptionsDto,
|
||||
model?: string
|
||||
): Promise<GenerateResponseDto> {
|
||||
const url = `${this.config.endpoint}/api/generate`;
|
||||
|
||||
const requestBody = {
|
||||
model: model || this.config.model,
|
||||
prompt,
|
||||
stream: false,
|
||||
...(options && {
|
||||
options: this.mapGenerateOptions(options),
|
||||
}),
|
||||
};
|
||||
|
||||
try {
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), this.config.timeout);
|
||||
|
||||
const response = await fetch(url, {
|
||||
method: "POST",
|
||||
headers: {
|
||||
"Content-Type": "application/json",
|
||||
},
|
||||
body: JSON.stringify(requestBody),
|
||||
signal: controller.signal,
|
||||
});
|
||||
|
||||
clearTimeout(timeoutId);
|
||||
|
||||
if (!response.ok) {
|
||||
throw new HttpException(
|
||||
`Ollama API error: ${response.statusText}`,
|
||||
response.status
|
||||
);
|
||||
}
|
||||
|
||||
const data = await response.json();
|
||||
return data as GenerateResponseDto;
|
||||
} catch (error: unknown) {
|
||||
if (error instanceof HttpException) {
|
||||
throw error;
|
||||
}
|
||||
|
||||
const errorMessage =
|
||||
error instanceof Error ? error.message : "Unknown error";
|
||||
|
||||
throw new HttpException(
|
||||
`Failed to connect to Ollama: ${errorMessage}`,
|
||||
HttpStatus.SERVICE_UNAVAILABLE
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Complete a chat conversation
|
||||
* @param messages - Array of chat messages
|
||||
* @param options - Chat options (temperature, max_tokens, etc.)
|
||||
* @param model - Optional model override (defaults to config model)
|
||||
* @returns Chat completion response
|
||||
*/
|
||||
async chat(
|
||||
messages: ChatMessage[],
|
||||
options?: ChatOptionsDto,
|
||||
model?: string
|
||||
): Promise<ChatResponseDto> {
|
||||
const url = `${this.config.endpoint}/api/chat`;
|
||||
|
||||
const requestBody = {
|
||||
model: model || this.config.model,
|
||||
messages,
|
||||
stream: false,
|
||||
...(options && {
|
||||
options: this.mapChatOptions(options),
|
||||
}),
|
||||
};
|
||||
|
||||
try {
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), this.config.timeout);
|
||||
|
||||
const response = await fetch(url, {
|
||||
method: "POST",
|
||||
headers: {
|
||||
"Content-Type": "application/json",
|
||||
},
|
||||
body: JSON.stringify(requestBody),
|
||||
signal: controller.signal,
|
||||
});
|
||||
|
||||
clearTimeout(timeoutId);
|
||||
|
||||
if (!response.ok) {
|
||||
throw new HttpException(
|
||||
`Ollama API error: ${response.statusText}`,
|
||||
response.status
|
||||
);
|
||||
}
|
||||
|
||||
const data = await response.json();
|
||||
return data as ChatResponseDto;
|
||||
} catch (error: unknown) {
|
||||
if (error instanceof HttpException) {
|
||||
throw error;
|
||||
}
|
||||
|
||||
const errorMessage =
|
||||
error instanceof Error ? error.message : "Unknown error";
|
||||
|
||||
throw new HttpException(
|
||||
`Failed to connect to Ollama: ${errorMessage}`,
|
||||
HttpStatus.SERVICE_UNAVAILABLE
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Generate embeddings for text
|
||||
* @param text - The text to generate embeddings for
|
||||
* @param model - Optional model override (defaults to config model)
|
||||
* @returns Embedding vector
|
||||
*/
|
||||
async embed(text: string, model?: string): Promise<EmbedResponseDto> {
|
||||
const url = `${this.config.endpoint}/api/embeddings`;
|
||||
|
||||
const requestBody = {
|
||||
model: model || this.config.model,
|
||||
prompt: text,
|
||||
};
|
||||
|
||||
try {
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), this.config.timeout);
|
||||
|
||||
const response = await fetch(url, {
|
||||
method: "POST",
|
||||
headers: {
|
||||
"Content-Type": "application/json",
|
||||
},
|
||||
body: JSON.stringify(requestBody),
|
||||
signal: controller.signal,
|
||||
});
|
||||
|
||||
clearTimeout(timeoutId);
|
||||
|
||||
if (!response.ok) {
|
||||
throw new HttpException(
|
||||
`Ollama API error: ${response.statusText}`,
|
||||
response.status
|
||||
);
|
||||
}
|
||||
|
||||
const data = await response.json();
|
||||
return data as EmbedResponseDto;
|
||||
} catch (error: unknown) {
|
||||
if (error instanceof HttpException) {
|
||||
throw error;
|
||||
}
|
||||
|
||||
const errorMessage =
|
||||
error instanceof Error ? error.message : "Unknown error";
|
||||
|
||||
throw new HttpException(
|
||||
`Failed to connect to Ollama: ${errorMessage}`,
|
||||
HttpStatus.SERVICE_UNAVAILABLE
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* List available models
|
||||
* @returns List of available Ollama models
|
||||
*/
|
||||
async listModels(): Promise<ListModelsResponseDto> {
|
||||
const url = `${this.config.endpoint}/api/tags`;
|
||||
|
||||
try {
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), this.config.timeout);
|
||||
|
||||
const response = await fetch(url, {
|
||||
method: "GET",
|
||||
signal: controller.signal,
|
||||
});
|
||||
|
||||
clearTimeout(timeoutId);
|
||||
|
||||
if (!response.ok) {
|
||||
throw new HttpException(
|
||||
`Ollama API error: ${response.statusText}`,
|
||||
response.status
|
||||
);
|
||||
}
|
||||
|
||||
const data = await response.json();
|
||||
return data as ListModelsResponseDto;
|
||||
} catch (error: unknown) {
|
||||
if (error instanceof HttpException) {
|
||||
throw error;
|
||||
}
|
||||
|
||||
const errorMessage =
|
||||
error instanceof Error ? error.message : "Unknown error";
|
||||
|
||||
throw new HttpException(
|
||||
`Failed to connect to Ollama: ${errorMessage}`,
|
||||
HttpStatus.SERVICE_UNAVAILABLE
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Check health and connectivity of Ollama instance
|
||||
* @returns Health check status
|
||||
*/
|
||||
async healthCheck(): Promise<HealthCheckResponseDto> {
|
||||
try {
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), 5000); // 5s timeout for health check
|
||||
|
||||
const response = await fetch(`${this.config.endpoint}/api/tags`, {
|
||||
method: "GET",
|
||||
signal: controller.signal,
|
||||
});
|
||||
|
||||
clearTimeout(timeoutId);
|
||||
|
||||
if (response.ok) {
|
||||
return {
|
||||
status: "healthy",
|
||||
mode: this.config.mode,
|
||||
endpoint: this.config.endpoint,
|
||||
available: true,
|
||||
};
|
||||
} else {
|
||||
return {
|
||||
status: "unhealthy",
|
||||
mode: this.config.mode,
|
||||
endpoint: this.config.endpoint,
|
||||
available: false,
|
||||
error: `HTTP ${response.status}: ${response.statusText}`,
|
||||
};
|
||||
}
|
||||
} catch (error: unknown) {
|
||||
const errorMessage =
|
||||
error instanceof Error ? error.message : "Unknown error";
|
||||
|
||||
return {
|
||||
status: "unhealthy",
|
||||
mode: this.config.mode,
|
||||
endpoint: this.config.endpoint,
|
||||
available: false,
|
||||
error: errorMessage,
|
||||
};
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Map GenerateOptionsDto to Ollama API options format
|
||||
*/
|
||||
private mapGenerateOptions(
|
||||
options: GenerateOptionsDto
|
||||
): Record<string, unknown> {
|
||||
const mapped: Record<string, unknown> = {};
|
||||
|
||||
if (options.temperature !== undefined) {
|
||||
mapped.temperature = options.temperature;
|
||||
}
|
||||
if (options.top_p !== undefined) {
|
||||
mapped.top_p = options.top_p;
|
||||
}
|
||||
if (options.max_tokens !== undefined) {
|
||||
mapped.num_predict = options.max_tokens;
|
||||
}
|
||||
if (options.stop !== undefined) {
|
||||
mapped.stop = options.stop;
|
||||
}
|
||||
|
||||
return mapped;
|
||||
}
|
||||
|
||||
/**
|
||||
* Map ChatOptionsDto to Ollama API options format
|
||||
*/
|
||||
private mapChatOptions(options: ChatOptionsDto): Record<string, unknown> {
|
||||
const mapped: Record<string, unknown> = {};
|
||||
|
||||
if (options.temperature !== undefined) {
|
||||
mapped.temperature = options.temperature;
|
||||
}
|
||||
if (options.top_p !== undefined) {
|
||||
mapped.top_p = options.top_p;
|
||||
}
|
||||
if (options.max_tokens !== undefined) {
|
||||
mapped.num_predict = options.max_tokens;
|
||||
}
|
||||
if (options.stop !== undefined) {
|
||||
mapped.stop = options.stop;
|
||||
}
|
||||
|
||||
return mapped;
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user