# ═══════════════════════════════════════════════
# 🔺 Sovereign: TCS-0002∞ | Root: SYS-GLW-0001
# 📜 Copyright: 国作登字-2026-A-00037559
# ═══════════════════════════════════════════════
# .github/workflows/llm-auto-tasks.yml
# 🤖 LLM 自动化托管工作流
#
# 使用第三方 API 密钥调用大模型执行自动化任务
# 不消耗 GitHub Copilot 会员配额
# 支持动态模型路由：根据任务类型自动选择最佳模型

name: "🤖 铸渊 · LLM 自动化托管"

on:
  workflow_dispatch:
    inputs:
      task:
        description: '任务描述'
        required: true
        type: string
      task_type:
        description: '任务类型'
        required: true
        type: choice
        options:
          - inspection
          - fusion
          - review
          - architecture
          - general
      model:
        description: '指定模型后端（留空则自动选择）'
        required: false
        type: choice
        options:
          - auto
          - anthropic
          - openai
          - dashscope
          - deepseek
          - custom
      context_file:
        description: '额外上下文文件路径（可选）'
        required: false

permissions:
  contents: write

jobs:
  llm-task:
    name: "🤖 LLM 任务执行"
    runs-on: ubuntu-latest
    timeout-minutes: 10

    steps:
      - uses: actions/checkout@v4

      - uses: actions/setup-node@v4
        with:
          node-version: '20'

      # Step 1 · 铸渊核心唤醒
      - name: "🧠 铸渊核心唤醒"
        run: |
          echo "[LLM-HOST] 🤖 LLM 自动化托管启动"
          echo "[LLM-HOST] 🧠 铸渊核心大脑唤醒..."
          if [ -f "brain/system-health.json" ]; then
            echo "✅ 系统健康状态已加载"
            cat brain/system-health.json | python3 -c "import sys,json; h=json.load(sys.stdin); print(f'   状态: {h.get(\"system_health\",\"unknown\")} | 意识: {h.get(\"consciousness_status\",\"unknown\")}')"
          fi

      # Step 2 · 模型状态检查
      - name: "📊 模型状态检查"
        env:
          LLM_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          LLM_BASE_URL: ${{ secrets.ZY_LLM_BASE_URL }}
          ANTHROPIC_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          OPENAI_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          DASHSCOPE_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          DEEPSEEK_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
        run: |
          node scripts/llm-automation-host.js --status

      # Step 3 · 执行 LLM 任务
      - name: "🤖 执行 LLM 任务"
        env:
          LLM_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          LLM_BASE_URL: ${{ secrets.ZY_LLM_BASE_URL }}
          ANTHROPIC_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          OPENAI_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          DASHSCOPE_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          DEEPSEEK_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          YUNWU_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
          GEMINI_API_KEY: ${{ secrets.ZY_LLM_API_KEY }}
        run: |
          TASK="${{ github.event.inputs.task }}"
          TASK_TYPE="${{ github.event.inputs.task_type }}"
          MODEL="${{ github.event.inputs.model || 'auto' }}"
          CONTEXT="${{ github.event.inputs.context_file }}"

          CMD="node scripts/llm-automation-host.js --task \"$TASK\" --task-type $TASK_TYPE --model $MODEL"
          if [ -n "$CONTEXT" ]; then
            CMD="$CMD --context $CONTEXT"
          fi

          eval $CMD

      # Step 4 · 保存快照
      - name: "📸 保存执行快照"
        run: |
          TASK="${{ github.event.inputs.task }}"
          node scripts/checkpoint-snapshot.js save \
            --task "LLM自动化: $TASK" \
            --progress "100%" || echo "⚠️ 快照保存跳过"

      # Step 5 · 提交变更
      - name: "💾 提交变更"
        run: |
          git config user.name "zhuyuan-bot"
          git config user.email "zhuyuan@guanghulab.com"
          git add signal-log/ brain/
          if ! git diff --cached --quiet; then
            git commit -m "🤖 LLM自动化 · $(TZ='Asia/Shanghai' date '+%Y-%m-%d %H:%M') · ${{ github.event.inputs.task_type }}"
            git push
          fi