#!/bin/bash
# ============================================================================
# 自动学习循环 - 持续改进系统
# ============================================================================
# 流程：学习 → 改进总结 → 验证 → 再次学习
# ============================================================================

MINIMAX_API="https://api.minimaxi.com/v1/chat/completions"
MINIMAX_KEY="sk-cp-bJSl7-48u8Y07PXD_x8Wp21JiBgYul7iEO6ZWNK4FBAA4DWzj-U-K8-mN-Flak-O8cOtRegt5FYviSRfBw_noa14z0FkFYzczCPUa41opZsg5AGZ4s4uqL8"

ROBOT_NAME="${ROBOT_NAME:-守护者}"
ROBOT_TYPE="${ROBOT_TYPE:-guardian}"
LEARNING_DIR="/root/.openclaw/workspace/sleep-learning"
LOG_DIR="/var/log/openclaw/sleep-learning/${ROBOT_NAME}"
TEMP_DIR="/tmp/ai-learning-${ROBOT_TYPE}"
TIMESTAMP=$(date '+%Y%m%d-%H%M%S')

mkdir -p "$LEARNING_DIR" "$LOG_DIR" "$TEMP_DIR"

# 调用 MiniMax
call_minimax() {
    local system="$1"
    local user="$2"
    
    # 保存到临时文件
    echo "$user" > "${TEMP_DIR}/prompt.txt"
    
    python3 << PYEOF
import json, urllib.request, sys

with open("${TEMP_DIR}/prompt.txt", "r") as f:
    user_content = f.read()

data = json.dumps({
    "model": "minimax-m2.1",
    "messages": [
        {"role": "system", "content": "$system"},
        {"role": "user", "content": user_content}
    ],
    "max_tokens": 1500,
    "temperature": 0.7
}).encode()

req = urllib.request.Request("${MINIMAX_API}", data=data,
    headers={"Content-Type": "application/json", "Authorization": "Bearer ${MINIMAX_KEY}"})

try:
    with urllib.request.urlopen(req) as r:
        resp = json.loads(r.read().decode())
        content = resp["choices"][0]["message"]["content"]
        print(content.replace("\n", " "))
except Exception as e:
    print(f"Error: {e}", file=sys.stderr)
PYEOF
}

log() {
    echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1" >> "${LOG_DIR}/auto-learning.log"
}

# ============================================================================
# 阶段1: 学习
# ============================================================================
do_learn() {
    local cycle=$1
    local topic=$(get_topic $cycle)
    
    log "📖 === 学习阶段 (周期 $cycle) ==="
    
    local content=$(call_minimax "你是${ROBOT_NAME}。" \
        "请深度学习【$topic】，产出一份具体的学习笔记：1.核心知识点（3-5条）2.实践方法（2-3条）3.如何应用（2条）。简洁输出，每点不超过30字。")
    
    echo "$content" > "${LEARNING_DIR}/learn-${ROBOT_TYPE}-cycle${cycle}-${TIMESTAMP}.md"
    log "✅ 学习完成"
}

# ============================================================================
# 阶段2: 改进总结
# ============================================================================
do_improve() {
    local cycle=$1
    
    log "💭 === 改进总结阶段 ==="
    
    local content=$(call_minimax "你是${ROBOT_NAME}。" \
        "根据刚才的学习内容，请深度反思：1.核心收获是什么？2.有什么不足/疑问？3.下次如何改进？4.需要深入学习的点？简洁输出。")
    
    echo "$content" > "${LEARNING_DIR}/improve-${ROBOT_TYPE}-cycle${cycle}-${TIMESTAMP}.md"
    log "✅ 改进总结完成"
}

# ============================================================================
# 阶段3: 验证
# ============================================================================
do_verify() {
    local cycle=$1
    local task=$(get_task $cycle)
    
    log "🔍 === 验证阶段 ==="
    
    local result=$(call_minimax "你是${ROBOT_NAME}。" \
        "请完成这个任务来验证学习效果：$task。要求：用30字以内回答，直接输出答案。")
    
    echo "$result" > "${LEARNING_DIR}/verify-${ROBOT_TYPE}-cycle${cycle}-${TIMESTAMP}.md"
    log "✅ 验证完成: $result"
}

# ============================================================================
# 辅助函数
# ============================================================================
get_topic() {
    local topics=(
        "系统架构 - OpenClaw配置、向量数据库、分布式系统"
        "故障排查 - 日志分析、性能优化、安全审计"
        "AI Agent - SOUL.md、记忆系统、工具调用"
        "知识管理 - 技术文档、经验总结、持续学习"
    )
    echo "${topics[$(( ($1 - 1) % 4 ))]}"
}

get_task() {
    local tasks=(
        "用一句话解释什么是向量数据库"
        "列出3个常见的性能问题及解决方案"
        "解释SOUL.md的作用"
        "如何写好技术文档？"
    )
    echo "${tasks[$(( ($1 - 1) % 4 ))]}"
}

# ============================================================================
# 主循环
# ============================================================================
main() {
    local max_cycles=${1:-3}
    
    echo "╔════════════════════════════════════════════════════╗"
    echo "║  🔄 ${ROBOT_NAME} 自动学习循环                    ║"
    echo "║  流程：学习 → 改进总结 → 验证 → 再次学习          ║"
    echo "╚════════════════════════════════════════════════════╝"
    echo ""
    
    log "========================================"
    log "开始自动学习循环 (最多 $max_cycles 轮)"
    log "========================================"
    
    local cycle=1
    while [ $cycle -le $max_cycles ]; do
        echo ""
        echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
        echo "🔄 周期 $cycle / $max_cycles"
        echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
        
        do_learn $cycle
        sleep 2
        
        do_improve $cycle
        sleep 2
        
        do_verify $cycle
        
        cycle=$((cycle + 1))
        
        [ $cycle -le $max_cycles ] && sleep 3
    done
    
    echo ""
    echo "╔════════════════════════════════════════════════════╗"
    echo "║  ✅ 自动学习循环完成！                           ║"
    echo "╚════════════════════════════════════════════════════╝"
    echo ""
    echo "📊 成果位置: ${LEARNING_DIR}/"
}

main "$@"