refactor(models): replace provider registry with protocol-based model connections

.github/workflows/verify.yml

@@ -12,8 +12,9 @@ name: Verify
 #
 #   better_sqlite3.node, vec0.so/.dylib/.dll, skia.*.node
 #
-# `lint` is deliberately not part of this gate: the repository currently has 6
-# pre-existing renderer react-hooks errors, unrelated to build correctness.
+# `lint` is deliberately not part of this gate: it reports no errors but still
+# carries a large pre-existing `@typescript-eslint/no-explicit-any` baseline,
+# unrelated to build correctness.
 
 on:
   pull_request:
@@ -51,6 +52,9 @@ jobs:
       - name: Typecheck
         run: npm run typecheck
 
+      - name: Unit tests
+        run: npm test
+
       # Runs `npm run build` first, then electron-builder --dir, so the bundle and
       # the package are both covered by this one step.
       - name: Build and package (unpacked)

CONTRIBUTING.md

@@ -42,7 +42,8 @@ npm run dev
 | Command                  | What it does                                                  |
 | ------------------------ | ------------------------------------------------------------- |
 | `npm run dev`            | Start the app in development with HMR.                        |
-| `npm run typecheck`      | Typecheck the main/preload and renderer projects.             |
+| `npm run typecheck`      | Typecheck the main/preload, renderer and test projects.       |
+| `npm test`               | Run the Node test suite (`test/**/*.test.ts`).                |
 | `npm run lint`           | ESLint (see the note on the current baseline below).          |
 | `npm run format`         | Prettier over the whole repository.                           |
 | `npm run build`          | Typecheck, then bundle with electron-vite.                    |
@@ -55,6 +56,7 @@ npm run dev
 
 ```bash
 npm run typecheck   # required — this is a CI gate
+npm test            # required — this is a CI gate
 npm run build
 ```
 
@@ -75,21 +77,22 @@ against the real `app.asar`, so it is the step that actually catches these.
 
 Two caveats:
 
-- **`lint` is not a gate.** `main` currently reports 6 pre-existing
-  `react-hooks` errors and 156 warnings. Don't fix them in an unrelated PR, and
-  don't add new ones.
+- **`lint` is not a gate.** `npm run lint` has no errors, but `main` still
+  reports ~118 pre-existing `@typescript-eslint/no-explicit-any` warnings.
+  Don't add new ones, and don't mass-refactor them in an unrelated PR.
 - **Don't run `npm run format` on the whole repository.** It rewrites every
-  file, and one pre-existing file (`src/main/services/FileParserService.ts`) is
-  not prettier-clean, so a repo-wide format pulls unrelated changes into your
-  diff. Format only what you touched:
+  file, and a repo-wide format pulls unrelated changes into your diff. Format
+  only what you touched:
 
   ```bash
   npx prettier --write <files>
   ```
 
-There is **no automated test suite yet**. `test/fixtures/` contains sample
-documents for manual parser checks. Describe your manual verification in the PR
-instead — which platform, which workflow, what you observed.
+Tests live in `test/**/*.test.ts` and run on Node's built-in test runner via
+`npm test` (no test framework dependency). `test/fixtures/` additionally
+contains sample documents for manual parser checks. For anything not covered by
+a test, describe your manual verification in the PR instead — which platform,
+which workflow, what you observed.
 
 ## Conventions
 
@@ -109,7 +112,7 @@ welcome when it is meaningful: `fix(mac): ...`.
 src/main/       Electron main process
   db/           Drizzle schema, migrations, queries
   services/     Document parsing, RAG, item and note storage
-  providers/    LLM provider abstraction and registry
+  models/       Model Connection resolution and API protocol adapters
 src/preload/    IPC bridge
 src/renderer/   React UI
 src/shared/     Types and utilities used by more than one process

README.md

@@ -46,7 +46,7 @@ It’s still early, but the core idea works — and I’m excited to share it an
 
 - 📚 Build a local knowledge base from documents and notes
 - 💬 Chat, summarize, and reason with your content using LLMs
-- 🔌 Provider-based LLM design (OpenAI, DeepSeek, Ollama, and more)
+- 🔌 Bring your own API: any OpenAI-, Anthropic- or Google-compatible endpoint, or a local server
 - 🔍 RAG-powered retrieval with precise source traceability
 - 🖥️ Desktop app built with Electron — no Docker, no server setup
 
@@ -83,7 +83,7 @@ KnowNote is for you if:
 ### 🤖 AI-Powered Q&A
 
 - Retrieval-Augmented Generation (RAG)
-- Multiple LLM providers
+- Any OpenAI / Anthropic / Google-compatible endpoint, or local models
 - Answers with precise source references
 
 ### 🔒 Local-First by Design
@@ -123,7 +123,7 @@ Feedback and suggestions are very welcome.
 
 ### ✅ Completed
 
-- AI LLM conversation with multiple providers
+- AI LLM conversation with any compatible endpoint, local or remote
 - Structured note generation
 - One-click mind map generation
 - RAG-based document retrieval
@@ -177,7 +177,7 @@ KnowNote/
 │   ├── main/              # Electron main process
 │   │   ├── db/            # Database configuration and schema
 │   │   ├── services/      # Core logic (document parsing, RAG, etc.)
-│   │   └── providers/     # LLM provider abstraction
+│   │   └── models/        # Model connection resolution and API protocol adapters
 │   ├── renderer/          # React renderer process
 │   ├── preload/           # Electron preload scripts
 │   └── shared/            # Shared types and utilities
@@ -196,7 +196,7 @@ If you have ideas about:
 
 - learning workflows
 - knowledge visualization
-- model/provider abstraction
+- model connection / API protocol abstraction
 
 I’d love to hear them.
 

README_CN.md

@@ -46,7 +46,7 @@
 
 - 📚 从文档和笔记构建本地知识库
 - 💬 使用 LLM 对内容进行聊天、总结和推理
-- 🔌 基于提供商的 LLM 设计（OpenAI、DeepSeek、Ollama 等）
+- 🔌 自带 API：任何兼容 OpenAI、Anthropic 或 Google 协议的端点，或本地部署
 - 🔍 基于 RAG 的检索，具有精确的来源追溯能力
 - 🖥️ 使用 Electron 构建的桌面应用——无需 Docker，无需服务器设置
 
@@ -83,7 +83,7 @@
 ### 🤖 AI 驱动的问答
 
 - 检索增强生成（RAG）
-- 多种 LLM 提供商
+- 任何兼容 OpenAI / Anthropic / Google 协议的端点，或本地模型
 - 带有精确来源引用的答案
 
 ### 🔒 本地优先设计
@@ -123,7 +123,7 @@ KnowNote 是一个早期阶段的项目。
 
 ### ✅ 已完成
 
-- 支持多个提供商的 AI LLM 对话
+- 支持任意兼容端点的 AI LLM 对话（本地或远程）
 - 结构化笔记生成
 - 一键生成思维导图
 - 基于 RAG 的文档检索
@@ -177,7 +177,7 @@ KnowNote/
 │   ├── main/              # Electron 主进程
 │   │   ├── db/            # 数据库配置和架构
 │   │   ├── services/      # 核心逻辑（文档解析、RAG 等）
-│   │   └── providers/     # LLM 提供商抽象
+│   │   └── models/        # 模型连接解析与 API 协议适配器
 │   ├── renderer/          # React 渲染进程
 │   ├── preload/           # Electron 预加载脚本
 │   └── shared/            # 共享类型和工具
@@ -196,7 +196,7 @@ KnowNote/
 
 - 学习工作流程
 - 知识可视化
-- 模型/提供商抽象
+- 模型连接 / API 协议抽象
 
 我很乐意听取。
 

package.json

@@ -18,7 +18,9 @@
     "lint": "eslint --cache .",
     "typecheck:node": "tsc --noEmit -p tsconfig.node.json --composite false",
     "typecheck:web": "tsc --noEmit -p tsconfig.web.json --composite false",
-    "typecheck": "npm run typecheck:node && npm run typecheck:web",
+    "typecheck:test": "tsc -p tsconfig.test.json",
+    "typecheck": "npm run typecheck:node && npm run typecheck:web && npm run typecheck:test",
+    "test": "node --disable-warning=MODULE_TYPELESS_PACKAGE_JSON --test \"test/**/*.test.ts\"",
     "start": "electron-vite preview",
     "dev": "electron-vite dev",
     "build": "npm run typecheck && electron-vite build",
@@ -90,8 +92,8 @@
     "rather than in app.asar."
   ],
   "devDependencies": {
-    "@ai-sdk/alibaba": "^0.0.31",
-    "@ai-sdk/deepseek": "^1.0.33",
+    "@ai-sdk/anthropic": "^2.0.105",
+    "@ai-sdk/google": "^2.0.99",
     "@ai-sdk/openai": "^2.0.89",
     "@ai-sdk/openai-compatible": "^1.0.31",
     "@dagrejs/dagre": "^1.1.8",
@@ -161,7 +163,6 @@
     "nanoid": "^3.3.11",
     "next-themes": "^0.4.6",
     "officeparser": "^5.2.2",
-    "ollama-ai-provider-v2": "^1.5.5",
     "postcss": "^8.5.6",
     "prettier": "^3.8.1",
     "react": "^19.2.4",

src/main/config/ConnectionConfigManager.ts

@@ -0,0 +1,90 @@
+/**
+ * ConnectionConfigManager
+ * 持久化 Model Connection 配置（apiKey 使用 safeStorage 加密）。
+ */
+
+import type Store from 'electron-store'
+import type { ConnectionMap, ModelCapability, ModelConnection } from '../../shared/types'
+import { MODEL_CAPABILITIES } from '../../shared/types'
+import type { StoreSchema } from './types'
+import { encryptObjectFields, decryptObjectFields, isEncryptionAvailable } from './secureStorage'
+import Logger from '../../shared/utils/logger'
+
+const SENSITIVE_FIELDS = ['apiKey']
+
+export class ConnectionConfigManager {
+  private getStore: () => Promise<Store<StoreSchema>>
+
+  constructor(getStore: () => Promise<Store<StoreSchema>>) {
+    this.getStore = getStore
+  }
+
+  /**
+   * 读取全部连接（自动解密 apiKey）
+   */
+  async getConnections(): Promise<ConnectionMap> {
+    const store = await this.getStore()
+    const stored = store.get('connections', {}) as ConnectionMap
+    const result: ConnectionMap = {}
+
+    for (const capability of MODEL_CAPABILITIES) {
+      const connection = stored?.[capability]
+      if (connection) {
+        result[capability] = decryptObjectFields({ ...connection }, SENSITIVE_FIELDS)
+      }
+    }
+
+    return result
+  }
+
+  /**
+   * 读取单个连接
+   */
+  async getConnection(capability: ModelCapability): Promise<ModelConnection | null> {
+    const connections = await this.getConnections()
+    return connections[capability] ?? null
+  }
+
+  /**
+   * 保存单个连接（覆盖同一 capability 的已有连接）
+   */
+  async saveConnection(capability: ModelCapability, connection: ModelConnection): Promise<void> {
+    const store = await this.getStore()
+    const connections = { ...((store.get('connections', {}) as ConnectionMap) || {}) }
+
+    let stored: ModelConnection = { ...connection }
+    if (isEncryptionAvailable()) {
+      stored = encryptObjectFields(stored, SENSITIVE_FIELDS)
+    } else {
+      Logger.warn(
+        'ConnectionConfigManager',
+        `Encryption not available, storing ${capability} apiKey in plain text`
+      )
+    }
+
+    connections[capability] = stored
+    store.set('connections', connections)
+    Logger.info('ConnectionConfigManager', `Saved ${capability} connection`)
+  }
+
+  /**
+   * 删除单个连接
+   */
+  async deleteConnection(capability: ModelCapability): Promise<void> {
+    const store = await this.getStore()
+    const connections = { ...((store.get('connections', {}) as ConnectionMap) || {}) }
+    delete connections[capability]
+    store.set('connections', connections)
+    Logger.info('ConnectionConfigManager', `Deleted ${capability} connection`)
+  }
+
+  /**
+   * 监听连接配置变化
+   */
+  async onConnectionsChange(callback: (connections: ConnectionMap) => void): Promise<() => void> {
+    const store = await this.getStore()
+    return store.onDidChange('connections', () => {
+      void this.getConnections().then(callback)
+    })
+  }
+}