Merge pull request #52 from hobbytp/feat/complete-digital-twin

feat: complete digital twin chatbot implementation

Author: hobbytp

.github/workflows/rag-ingest.yml

@@ -0,0 +1,47 @@
+name: RAG Ingestion
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - "content/**"
+  workflow_dispatch:
+
+jobs:
+  ingest:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r scripts/requirements.txt
+
+      - name: Run Ingestion Script
+        env:
+          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }}
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+          CLOUDFLARE_VECTORIZE_INDEX_NAME: "blog-index"
+        run: |
+          python scripts/ingest.py
+
+      - name: Commit State File
+        run: |
+          git config --local user.email "action@github.com"
+          git config --local user.name "GitHub Action"
+          git add .ingest_state.json
+          git diff --staged --quiet || git commit -m "🤖 Update RAG ingestion state"
+          git push

.ingest_state.json

@@ -28,9 +28,9 @@ ${contextText}
 
 Instructions:
 1. Use a professional, friendly, slightly geeky tone.
-2. If the answer is not in the Context, say "My memory bank doesn't have info on this specific topic based on the blog posts."
-3. Answer in the same language as the user's question (Default to Chinese).
-4. Keep answers concise.`;
+2. If the answer is not in the Context, you MUST reply exactly with: "没有找到相关内容".
+3. Answer in Chinese (Simplified) by default, unless the user asks in another language.
+4. Keep answers concise and helpful.`;
 }
 
 /**

assets/css/chat.css assets/css/extended/chat.cssassets/css/chat.css renamed to assets/css/extended/chat.css

@@ -1,3 +1,4 @@
+<script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
 <div id="chat-launcher" 
      class="chat-launcher" 
      role="button" 

functions/api/chat.js

@@ -0,0 +1,31 @@
+# 完善博客数字分身助手 (Digital Twin Chatbot)
+
+## 目标
+完成基于 RAG 的博客数字分身助手功能，修复 CSS 加载问题，引入必要的 JS 依赖，并优化提示词以支持中文回复。
+
+## 变更内容
+
+### 1. 修复样式加载
+- **操作**: 将 `assets/css/chat.css` 移动到 `assets/css/extended/chat.css`。
+- **原因**: Hugo PaperMod 主题的 CSS 打包逻辑（在 `head.html` 中）只会自动加载 `assets/css/extended/` 目录下的扩展样式文件。当前文件位于 `assets/css/` 根目录，因此未被加载，导致聊天窗口无样式。
+
+### 2. 引入 Marked.js 依赖
+- **操作**: 在 `layouts/partials/chatbox.html` 顶部添加 Marked.js 的 CDN 引用。
+  ```html
+  <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
+  ```
+- **原因**: 前端代码依赖 `marked` 库来渲染 AI 返回的 Markdown 格式回复。目前该库未被引入，导致 Markdown 渲染功能失效。
+
+### 3. 优化后端提示词 (System Prompt)
+- **操作**: 修改 `functions/api/chat.js` 中的 `buildSystemPrompt` 函数。
+- **变更点**:
+  - 将 System Prompt 的指令语言调整为更自然的中文设定。
+  - 明确指定“找不到内容”时的回复语为用户要求的：“没有找到相关内容”。
+  - 确保 AI 默认使用中文回答。
+
+## 验证计划
+1.  **样式验证**: 部署后检查右下角是否出现悬浮聊天按钮，点击是否能正常展开聊天窗口。
+2.  **功能验证**: 发送测试消息（如“你好”），确认是否收到回复。
+3.  **Markdown 验证**: 询问需要代码块或列表回复的问题，确认渲染是否正常。
+4.  **RAG 验证**: 询问博客相关问题，确认回复基于上下文。
+5.  **Fallback 验证**: 询问无关问题（如“番茄炒蛋怎么做”），确认回复“没有找到相关内容”。

layouts/partials/chatbox.html

@@ -0,0 +1,7 @@
+# Tasks
+
+- [ ] Create OpenSpec proposal file <!-- id: 0 -->
+- [ ] Move `assets/css/chat.css` to `assets/css/extended/chat.css` <!-- id: 1 -->
+- [ ] Add marked.js CDN to `layouts/partials/chatbox.html` <!-- id: 2 -->
+- [ ] Update system prompt in `functions/api/chat.js` <!-- id: 3 -->
+- [ ] Commit and push changes <!-- id: 4 -->