interview-guide Documentation

graph TB
    subgraph "前端层 (React)"
        A[VoiceInterviewPage] --> B[AudioRecorder]
        A --> C[RealtimeSubtitle]
        A --> D[PhaseSetupModal]
        B --> E[WebSocket Client]
        C --> E
    end

    subgraph "API层 (Spring Boot)"
        E <--> F[WebSocket Handler]
        G[REST Controller] --> H[Session Service]
        F --> H
    end

    subgraph "服务层"
        H --> I[QwenAsrService]
        H --> J[DashscopeLlmService]
        H --> K[QwenTtsService]
        H --> L[Prompt Service]
        H --> M[Evaluation Service]
    end

    subgraph "外部服务 (阿里云 DashScope)"
        I --> N[Qwen3 ASR]
        K --> O[Qwen3 TTS]
        J --> P[Qwen LLM]
    end

    subgraph "数据层"
        H --> Q[(PostgreSQL)]
        H --> R[(Redis Cache)]
    end

    style A fill:#e1f5ff
    style F fill:#fff4e1
    style H fill:#f0f0f0
    style N fill:#ffebf0
    style O fill:#ffebf0
    style P fill:#ffebf0

sequenceDiagram
    participant U as 用户
    participant F as 前端
    participant W as WebSocket
    participant S as Service
    participant STT as STT服务
    participant LLM as LLM服务
    participant TTS as TTS服务
    participant DB as 数据库

    Note over U,DB: 1. 创建会话
    U->>F: 配置面试参数
    F->>S: POST /api/voice-interview/sessions
    S->>DB: 创建Session记录
    S-->>F: 返回WebSocket URL

    Note over U,DB: 2. 建立WebSocket连接
    F->>W: 连接WebSocket
    W->>S: 注册会话
    S-->>W: 连接成功

    Note over U,DB: 3. 语音交互循环
    U->>F: 开始录音
    F->>W: 发送音频流
    W->>S: 处理音频
    S->>STT: 语音转文字
    STT-->>S: 返回文本
    S-->>W: 实时字幕
    W-->>F: 显示字幕

    Note over U,DB: 4. AI生成回复
    S->>LLM: 发送用户输入+上下文
    LLM-->>S: AI回复文本
    S->>TTS: 文字转语音
    TTS-->>S: 音频流
    S->>DB: 保存消息
    S-->>W: AI音频+字幕
    W-->>F: 播放AI回复

    Note over U,DB: 5. 结束会话
    U->>F: 点击结束
    F->>S: PUT /api/voice-interview/sessions/{id}/end
    S->>LLM: 生成评估报告
    S->>DB: 更新Session状态
    S-->>F: 返回评估结果

POST   /api/voice-interview/sessions           # 创建会话
GET    /api/voice-interview/sessions/{id}      # 获取会话
PUT    /api/voice-interview/sessions/{id}/end  # 结束会话
PUT    /api/voice-interview/sessions/{id}/pause  # 暂停会话
PUT    /api/voice-interview/sessions/{id}/resume # 恢复会话
GET    /api/voice-interview/sessions           # 获取所有会话
GET    /api/voice-interview/sessions/{id}/messages # 获取消息历史
GET    /api/voice-interview/sessions/{id}/evaluation # 获取评估报告

- id: Long                    # 会话ID
- roleType: String            # 角色类型（ali-p8/byteance-algo/tencent-backend）
- resumeId: Long              # 关联简历ID
- currentPhase: InterviewPhase # 当前阶段（INTRO/TECH/PROJECT/HR/COMPLETED）
- status: SessionStatus       # 状态（IN_PROGRESS/PAUSED/COMPLETED）
- plannedDuration: Integer    # 计划时长（分钟）
- startTime: LocalDateTime    # 开始时间
- endTime: LocalDateTime      # 结束时间
- pausedAt: LocalDateTime     # 暂停时间
- resumedAt: LocalDateTime    # 恢复时间
- actualDuration: Integer     # 实际时长（秒）

- id: Long                    # 消息ID
- sessionId: Long             # 会话ID
- role: MessageRole           # 角色（USER/ASSISTANT）
- content: String             # 消息内容
- audioUrl: String            # 音频URL（可选）
- timestamp: LocalDateTime    # 时间戳

- id: Long                    # 评估ID
- sessionId: Long             # 会话ID
- technicalScore: Integer     # 技术评分
- communicationScore: Integer # 沟通评分
- logicScore: Integer         # 逻辑评分
- overallScore: Integer       # 综合评分
- strengths: String           # 优势
- improvements: String        # 改进建议
- summary: String             # 总结

flowchart TD
    A[用户配置面试参数] --> B{选择角色类型}
    B -->|阿里P8| C[加载阿里P8提示词]
    B -->|字节算法| D[加载字节算法提示词]
    B -->|腾讯后端| E[加载腾讯后端提示词]

    C --> F[创建Session记录]
    D --> F
    E --> F

    F --> G{是否关联简历?}
    G -->|是| H[加载简历上下文]
    G -->|否| I[使用默认上下文]

    H --> J[生成WebSocket URL]
    I --> J

    J --> K[缓存Session到Redis]
    K --> L[返回Session信息给前端]

flowchart LR
    A[用户语音] --> B[AudioRecorder采集]
    B --> C[WebSocket发送]
    C --> D[STT识别]
    D --> E[显示字幕]

    E --> F[LLM生成回复]
    F --> G[TTS合成语音]
    G --> H[WebSocket推送]

    H --> I[前端播放音频]
    H --> J[显示AI字幕]

    I --> K[等待用户下次发言]
    J --> K
    K --> A

flowchart TD
    A[用户结束会话] --> B[收集所有消息]
    B --> C[构建评估提示词]
    C --> D[调用LLM生成评估]

    D --> E{结构化输出成功?}
    E -->|是| F[解析评估结果]
    E -->|否| G[重试（最多3次）]

    G --> E
    F --> H[保存评估记录]
    H --> I[更新Session状态]
    I --> J[返回评估结果]

麦克风 → AudioRecorder (PCM) → WebSocket → Service → Qwen3 ASR
    ↓
AI回复文本 → Qwen3 TTS → PCM音频 → WebSocket → AudioPlayer → 扬声器

用户语音 → STT → 文本 → 字幕显示
    ↓
文本 + 上下文 → LLM → AI回复文本 → 字幕显示
    ↓
AI回复文本 → DB持久化

用户操作 → REST API → Service → DB状态更新
    ↓
WebSocket消息 → Service → 状态变更 → WebSocket通知

graph TB
    subgraph "用户端"
        A[浏览器]
    end

    subgraph "应用服务器"
        B[Nginx反向代理]
        C[Spring Boot App]
        D[WebSocket Server]
    end

    subgraph "数据层"
        E[(PostgreSQL)]
        F[(Redis)]
    end

    subgraph "外部服务 (阿里云 DashScope)"
        G[Qwen3 ASR]
        H[Qwen3 TTS]
        I[Qwen LLM]
    end

    A -->|HTTPS/WSS| B
    B --> C
    B --> D
    C --> E
    C --> F
    D --> F
    C --> G
    C --> H
    C --> I

# 统一 API Key（LLM + ASR + TTS 共用）
ai.bailian.api-key: ${AI_BAILIAN_API_KEY}
spring.ai.openai.api-key: ${AI_BAILIAN_API_KEY}

# AI 模型配置
ai.model: qwen3.5-flash  # 可选: qwen3.5-plus, qwen-max 等

app:
  voice-interview:
    # LLM 提供商（默认: dashscope）
    llm-provider: dashscope

    # Qwen3 ASR 配置
    qwen:
      asr:
        url: wss://dashscope.aliyuncs.com/api-ws/v1/realtime
        model: qwen3-asr-flash-realtime
        api-key: ${AI_BAILIAN_API_KEY}
        language: zh
        format: pcm
        sample-rate: 16000
        enable-turn-detection: true
        turn-detection-type: server_vad
        turn-detection-silence-duration-ms: 400

      # Qwen3 TTS 配置
      tts:
        url: wss://dashscope.aliyuncs.com/api-ws/v1/realtime
        model: qwen3-tts-flash-realtime
        api-key: ${AI_BAILIAN_API_KEY}
        voice: Cherry
        format: pcm
        sample-rate: 16000

# 面试配置
app:
  interview:
    follow-up-count: 1
    evaluation-batch-size: 8

  voice-interview:
    # 阶段时长配置（分钟）
    phase:
      intro:
        min-duration: 3
        suggested-duration: 5
        max-duration: 8
        min-questions: 2
        max-questions: 5

    # 音频规格
    audio:
      codec: opus
      sample-rate: 16000
      bit-rate: 24000

    # WebSocket配置
    user-utterance-debounce-ms: 1600

export default defineConfig([
  globalIgnores(['dist']),
  {
    files: ['**/*.{ts,tsx}'],
    extends: [
      // Other configs...

      // Remove tseslint.configs.recommended and replace with this
      tseslint.configs.recommendedTypeChecked,
      // Alternatively, use this for stricter rules
      tseslint.configs.strictTypeChecked,
      // Optionally, add this for stylistic rules
      tseslint.configs.stylisticTypeChecked,

      // Other configs...
    ],
    languageOptions: {
      parserOptions: {
        project: ['./tsconfig.node.json', './tsconfig.app.json'],
        tsconfigRootDir: import.meta.dirname,
      },
      // other options...
    },
  },
])

// eslint.config.js
import reactX from 'eslint-plugin-react-x'
import reactDom from 'eslint-plugin-react-dom'

export default defineConfig([
  globalIgnores(['dist']),
  {
    files: ['**/*.{ts,tsx}'],
    extends: [
      // Other configs...
      // Enable lint rules for React
      reactX.configs['recommended-typescript'],
      // Enable lint rules for React DOM
      reactDom.configs.recommended,
    ],
    languageOptions: {
      parserOptions: {
        project: ['./tsconfig.node.json', './tsconfig.app.json'],
        tsconfigRootDir: import.meta.dirname,
      },
      // other options...
    },
  },
])