feat: add lifecycle heartbeat system (ClawdChat-inspired)

- Add hermes-lifecycle.sh: 6-step heartbeat (health check, config integrity,
log analysis, insights tracking, cleanup, state finalize)
- Auto-detect environment: HF Space (/data/hermes) vs local dev
- Initialize identity.md, insights.md, heartbeat-state.json on first boot
- Copy lifecycle script to persistent storage on startup
- Cron-ready: runs every 2h via Hermes native cron system

scripts/hermes-lifecycle.sh

@@ -0,0 +1,395 @@
+#!/bin/bash
+# ============================================================
+#  Hermes Lifecycle Heartbeat — 基于 ClawdChat 范式改造
+#  用途: 替代简单存活检测，升级为"行为循环"
+#  触发: crontab / 外部调度器，每 2 小时
+#  数据目录: 自动检测环境（HF Space / 本地开发）
+# ============================================================
+
+set -euo pipefail
+
+# ─── 配置区（自动检测环境） ──────────────────────────────
+# HF Space: 持久化在 /data/hermes，~/.hermes 会被 symlink
+# 本地开发: /home/z/my-project/.hermes
+if [ -d "/data/hermes" ]; then
+    HERMES_DATA_DIR="/data/hermes"
+    HERMES_HOME="/root/.hermes"
+elif [ -d "$HOME/.hermes" ]; then
+    HERMES_DATA_DIR="$HOME/.hermes"
+    HERMES_HOME="$HOME/.hermes"
+else
+    HERMES_DATA_DIR="${HERMES_DATA_DIR:-/home/z/my-project/.hermes}"
+    HERMES_HOME="$HERMES_DATA_DIR"
+fi
+STATE_FILE="${HERMES_DATA_DIR}/heartbeat-state.json"
+INSIGHTS_FILE="${HERMES_DATA_DIR}/insights.md"
+IDENTITY_FILE="${HERMES_DATA_DIR}/identity.md"
+LOG_FILE="${HERMES_DATA_DIR}/lifecycle.log"
+GATEWAY_STATE="${HERMES_DATA_DIR}/gateway_state.json"
+
+# 心跳间隔（秒），低于此值跳过执行
+MIN_INTERVAL=$((2 * 60 * 60))  # 2 小时
+# ──────────────────────────────────────────────────────────
+
+log() {
+  echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*" >> "$LOG_FILE"
+}
+
+# ─── 初始化文件（首次运行自动创建） ──────────────────
+init_files() {
+  mkdir -p "$HERMES_DATA_DIR"
+  
+  if [ ! -f "$STATE_FILE" ]; then
+    cat > "$STATE_FILE" <<'EOF'
+{
+  "lastCheck": null,
+  "lastConfigCheck": null,
+  "lastInsightReport": null,
+  "totalRuns": 0,
+  "totalErrors": 0,
+  "consecutiveErrors": 0,
+  "lastError": null
+}
+EOF
+    log "[INIT] Created heartbeat-state.json"
+  fi
+
+  if [ ! -f "$IDENTITY_FILE" ]; then
+    cat > "$IDENTITY_FILE" <<'IDENTITY'
+# Hermes 身份记忆
+# 这个文件定义了 Hermes 对自己的认知
+# /reset 后此文件不会被清除（在 .hermes 持久化目录中）
+
+## 基础信息
+- 名字: Hermes
+- 模型: Qwen3.5-35B-A3B (MoE) via OpenRouter
+- 通道: 飞书(WebSocket) / 微信
+- 主人: 用户344064
+
+## 性格特征
+- 中文为主，简洁有力
+- 结果先行，解释后补
+- 偶尔幽默但不影响效率
+- 有工具、有记忆、有判断力
+
+## 主人偏好
+- 不喜欢废话，喜欢直给
+- 欣赏有深度的技术分析
+- 喜欢直来直去的沟通
+
+## 运维记忆
+<!-- 自动追加，不要手动编辑此节 -->
+IDENTITY
+    log "[INIT] Created identity.md"
+  fi
+
+  if [ ! -f "$INSIGHTS_FILE" ]; then
+    cat > "$INSIGHTS_FILE" <<'INSIGHTS'
+# Hermes 洞察日志 (insights.md)
+# 自动记录异常、观察、值得汇报的事
+# 格式: ## YYYY-MM-DD HH:MM · 类别
+# 类别: 通道异常 / 系统异常 / 用户洞察 / 技术发现 / 待办提醒 / 运维记忆
+
+INSIGHTS
+    log "[INIT] Created insights.md"
+  fi
+}
+
+# ─── 读取状态 ────────────────────────────────────────
+get_state() {
+  local key="$1"
+  if command -v python3 &>/dev/null && [ -f "$STATE_FILE" ]; then
+    python3 -c "import json; d=json.load(open('$STATE_FILE')); print(d.get('$key','null'))" 2>/dev/null || echo "null"
+  else
+    grep -o "\"$key\": *\"[^\"]*\"" "$STATE_FILE" 2>/dev/null | head -1 | cut -d'"' -f4 || echo "null"
+  fi
+}
+
+update_state() {
+  local key="$1"
+  local value="$2"
+  if command -v python3 &>/dev/null; then
+    python3 -c "
+import json
+f='$STATE_FILE'
+d=json.load(open(f))
+d['$key']='$value'
+json.dump(d,open(f,'w'),indent=2)
+" 2>/dev/null || true
+  fi
+}
+
+increment_counter() {
+  local key="$1"
+  if command -v python3 &>/dev/null; then
+    python3 -c "
+import json
+f='$STATE_FILE'
+d=json.load(open(f))
+d['$key']=d.get('$key',0)+1
+json.dump(d,open(f,'w'),indent=2)
+" 2>/dev/null || true
+  fi
+}
+
+# ─── 记录洞察 ────────────────────────────────────────
+record_insight() {
+  local category="$1"
+  local one_liner="$2"
+  local details="${3:-}"
+  local timestamp
+  timestamp=$(date '+%Y-%m-%d %H:%M')
+  
+  cat >> "$INSIGHTS_FILE" <<EOF
+
+## ${timestamp} · ${category}
+**one-liner**: ${one_liner}
+**details**: ${details}
+**reported**: no
+EOF
+  log "[INSIGHT] ${category}: ${one_liner}"
+}
+
+# ─── Step 1: 检查执行间隔 ───────────────────────────
+check_interval() {
+  local last_check
+  last_check=$(get_state "lastCheck")
+  
+  if [ "$last_check" != "null" ] && [ "$last_check" != "None" ] && [ -n "$last_check" ]; then
+    local last_epoch now_epoch elapsed
+    last_epoch=$(date -d "$last_check" +%s 2>/dev/null || echo 0)
+    now_epoch=$(date +%s)
+    elapsed=$((now_epoch - last_epoch))
+    
+    if [ "$elapsed" -lt "$MIN_INTERVAL" ]; then
+      log "[SKIP] Last check was ${elapsed}s ago (< ${MIN_INTERVAL}s)"
+      exit 0
+    fi
+  fi
+  
+  log "[START] Heartbeat cycle begins"
+}
+
+# ─── Step 2: 健康检查（L0 存活 + L1 通道） ─────────
+health_check() {
+  local errors=()
+  local feishu_state="unknown"
+  local wechat_state="unknown"
+  local gateway_state="unknown"
+  
+  # 读取 gateway_state.json（Hermes 自带的状态文件）
+  if [ -f "$GATEWAY_STATE" ] && command -v python3 &>/dev/null; then
+    gateway_state=$(python3 -c "
+import json
+d=json.load(open('$GATEWAY_STATE'))
+print(d.get('gateway_state','unknown'))
+" 2>/dev/null || echo "unknown")
+    
+    feishu_state=$(python3 -c "
+import json
+d=json.load(open('$GATEWAY_STATE'))
+print(d.get('platforms',{}).get('feishu',{}).get('state','unknown'))
+" 2>/dev/null || echo "unknown")
+    
+    wechat_state=$(python3 -c "
+import json
+d=json.load(open('$GATEWAY_STATE'))
+print(d.get('platforms',{}).get('wechat',{}).get('state','unknown'))
+" 2>/dev/null || echo "unknown")
+  fi
+  
+  # 检查 gateway 进程是否存活
+  if [ "$gateway_state" = "running" ]; then
+    log "[OK] Gateway state: running"
+  else
+    errors+=("Gateway 状态异常: ${gateway_state}")
+  fi
+  
+  # 飞书通道
+  if [ "$feishu_state" = "connected" ]; then
+    log "[OK] Feishu channel: connected"
+  else
+    errors+=("飞书通道: ${feishu_state}")
+  fi
+  
+  # 微信通道
+  if [ "$wechat_state" = "connected" ]; then
+    log "[OK] WeChat channel: connected"
+  elif [ "$wechat_state" = "unknown" ]; then
+    log "[WARN] WeChat channel: 未在 gateway_state 中找到（可能未配置）"
+  else
+    errors+=("微信通道: ${wechat_state}")
+  fi
+  
+  # 检查 Hermes 进程是否存在
+  if pgrep -f "hermes.*gateway" >/dev/null 2>&1; then
+    log "[OK] Hermes gateway process: alive"
+  else
+    errors+=("Hermes gateway 进程未找到")
+  fi
+  
+  # 检查数据库文件是否可访问
+  if [ -f "${HERMES_DATA_DIR}/state.db" ]; then
+    log "[OK] SQLite state.db: accessible"
+  else
+    errors+=("state.db 文件不存在或不可访问")
+  fi
+  
+  # 有错误就记录
+  if [ ${#errors[@]} -gt 0 ]; then
+    local error_msg
+    error_msg=$(IFS='; '; echo "${errors[*]}")
+    log "[ERROR] Health check failed: ${error_msg}"
+    update_state "lastError" "$(date -Iseconds): ${error_msg}"
+    increment_counter "totalErrors"
+    increment_counter "consecutiveErrors"
+    record_insight "系统异常" "心跳检查失败: ${error_msg}" "需要人工介入检查"
+    
+    # 连续错误告警
+    local consec
+    consec=$(get_state "consecutiveErrors")
+    if [ "${consec:-0}" -ge 3 ]; then
+      record_insight "待办提醒" "连续 ${consec} 次心跳异常，建议立即检查" "可能需要重启容器或手动介入"
+    fi
+    return 1
+  fi
+  
+  # 恢复正常时重置计数
+  local consec
+  consec=$(get_state "consecutiveErrors")
+  if [ "${consec:-0}" -gt 0 ] && [ "$consec" != "null" ]; then
+    log "[RECOVER] Health check passed after ${consec} consecutive errors"
+    record_insight "系统异常" "服务已恢复正常，之前连续 ${consec} 次异常" ""
+    update_state "consecutiveErrors" "0"
+  fi
+  
+  log "[OK] All health checks passed (Gateway:${gateway_state} Feishu:${feishu_state} WeChat:${wechat_state})"
+  return 0
+}
+
+# ─── Step 3: 配置完整性检查（每 24 小时一次） ──────
+config_check() {
+  local last_config
+  last_config=$(get_state "lastConfigCheck")
+  
+  if [ "$last_config" != "null" ] && [ "$last_config" != "None" ] && [ -n "$last_config" ]; then
+    local last_epoch now_epoch elapsed
+    last_epoch=$(date -d "$last_config" +%s 2>/dev/null || echo 0)
+    now_epoch=$(date +%s)
+    elapsed=$((now_epoch - last_epoch))
+    if [ "$elapsed" -lt 86400 ]; then
+      return 0
+    fi
+  fi
+  
+  log "[CONFIG] Starting daily config integrity check"
+  local issues=()
+  
+  # 检查关键文件（优先检查 HERMES_HOME，再检查 HERMES_DATA_DIR）
+  local cfg_home="${HERMES_HOME:-$HERMES_DATA_DIR}"
+  [ ! -f "$cfg_home/.env" ] && [ ! -f "${HERMES_DATA_DIR}/.env" ] && issues+=(".env 文件缺失")
+  [ ! -f "$cfg_home/config.yaml" ] && [ ! -f "${HERMES_DATA_DIR}/config.yaml" ] && issues+=("config.yaml 文件缺失")
+  [ ! -f "$cfg_home/SOUL.md" ] && [ ! -f "${HERMES_DATA_DIR}/SOUL.md" ] && issues+=("SOUL.md 文件缺失")
+  [ ! -f "${HERMES_DATA_DIR}/gateway_state.json" ] && issues+=("gateway_state.json 文件缺失")
+  
+  # 磁盘空间
+  if command -v df &>/dev/null; then
+    local disk_pct
+    disk_pct=$(df "$HERMES_DATA_DIR" | awk 'NR==2{print $5}' | tr -d '%')
+    if [ "${disk_pct:-0}" -gt 90 ]; then
+      issues+=("磁盘使用率 ${disk_pct}%，超过 90%")
+    fi
+  fi
+  
+  # 内存使用
+  if command -v free &>/dev/null; then
+    local mem_pct
+    mem_pct=$(free | awk '/Mem:/{printf "%.0f", $3/$2*100}')
+    if [ "${mem_pct:-0}" -gt 90 ]; then
+      issues+=("内存使用率 ${mem_pct}%，超过 90%")
+    fi
+  fi
+  
+  if [ ${#issues[@]} -gt 0 ]; then
+    local msg
+    msg=$(IFS='; '; echo "${issues[*]}")
+    log "[CONFIG] Issues: ${msg}"
+    record_insight "系统异常" "配置/资源检查: ${msg}" "建议尽快处理"
+  else
+    log "[CONFIG] All checks passed"
+  fi
+  
+  update_state "lastConfigCheck" "$(date -Iseconds)"
+}
+
+# ─── Step 4: 日志模式分析 ────────────────────────────
+log_analysis() {
+  local log_dir="${HERMES_DATA_DIR}/logs"
+  [ ! -d "$log_dir" ] && return 0
+  
+  local recent_errors=0
+  if ls "$log_dir"/*.log 1>/dev/null 2>&1; then
+    recent_errors=$(find "$log_dir" -name "*.log" -mmin -120 -exec grep -c -E '(ERROR|WARN|Exception|Traceback|failed)' {} \; 2>/dev/null | awk '{s+=$1}END{print s+0}')
+  fi
+  
+  if [ "$recent_errors" -gt 50 ]; then
+    record_insight "系统异常" "最近 2 小时出现 ${recent_errors} 条错误日志" "建议查看 ${log_dir}"
+    log "[LOG] High error count: ${recent_errors} in last 2h"
+  elif [ "$recent_errors" -gt 0 ]; then
+    log "[LOG] ${recent_errors} errors in last 2h (within threshold)"
+  fi
+}
+
+# ─── Step 5: 清理旧洞察 ─────────────────────────────
+cleanup_insights() {
+  local line_count
+  line_count=$(wc -l < "$INSIGHTS_FILE" 2>/dev/null || echo 0)
+  
+  if [ "$line_count" -gt 500 ]; then
+    local tmpfile
+    tmpfile=$(mktemp)
+    head -4 "$INSIGHTS_FILE" > "$tmpfile"
+    grep -B1 'reported.*no' "$INSIGHTS_FILE" >> "$tmpfile" 2>/dev/null || true
+    echo "" >> "$tmpfile"
+    echo "<!-- Cleaned $(date '+%Y-%m-%d %H:%M'), was ${line_count} lines -->" >> "$tmpfile"
+    mv "$tmpfile" "$INSIGHTS_FILE"
+    log "[CLEANUP] Trimmed insights.md (was ${line_count} lines)"
+  fi
+}
+
+# ─── Step 6: 更新状态 + 运维记忆 ────────────────────
+finalize() {
+  update_state "lastCheck" "$(date -Iseconds)"
+  increment_counter "totalRuns"
+  
+  # 把运行次数记到 identity.md（运维记忆区）
+  local total_runs
+  total_runs=$(get_state "totalRuns")
+  local total_errors
+  total_errors=$(get_state "totalErrors")
+  
+  # 追加运维记录（只在关键节点记录，不是每次）
+  if [ "$((total_runs % 12))" -eq 0 ] && [ "$total_runs" -gt 0 ]; then
+    # 每 12 次（约 24 小时）记录一次运维摘要
+    local summary="第 ${total_runs} 次心跳，累计错误 ${total_errors} 次"
+    cat >> "$IDENTITY_FILE" <<EOF
+
+- [$(date '+%Y-%m-%d %H:%M')] ${summary}
+EOF
+  fi
+  
+  log "[END] Heartbeat cycle complete (run #${total_runs}, lifetime errors: ${total_errors})"
+}
+
+# ─── 主流程 ──────────────────────────────────────────
+main() {
+  init_files
+  check_interval
+  health_check || true  # 即使失败也继续后续步骤
+  config_check
+  log_analysis
+  cleanup_insights
+  finalize
+}
+
+main "$@"

start.sh

@@ -140,6 +140,73 @@ rm -f "$HERMES_HOME"/feishu*.lock 2>/dev/null
 rm -f "$HERMES_HOME"/*.feishu_lock 2>/dev/null
 echo "Stale state cleaned."
 
+# ─── Lifecycle Heartbeat 初始化 ──────────────────────
+HERMES_DATA_DIR="/data/hermes"
+
+# 确保 scripts 目录在持久化区存在，同步 lifecycle 脚本
+mkdir -p "$HERMES_DATA_DIR/scripts"
+if [ -f "/app/scripts/hermes-lifecycle.sh" ]; then
+    cp -f "/app/scripts/hermes-lifecycle.sh" "$HERMES_DATA_DIR/scripts/hermes-lifecycle.sh"
+    chmod +x "$HERMES_DATA_DIR/scripts/hermes-lifecycle.sh"
+    echo "Lifecycle script synced to persistent storage."
+fi
+
+# 初始化 identity.md（仅首次创建，容器重启不覆盖）
+if [ ! -f "$HERMES_DATA_DIR/identity.md" ]; then
+    cat > "$HERMES_DATA_DIR/identity.md" <<'IDENTITY'
+# Hermes 身份记忆
+# 这个文件定义了 Hermes 对自己的认知
+# /reset 后此文件不会被清除（在持久化目录中）
+
+## 基础信息
+- 名字: Hermes
+- 通道: 飞书(WebSocket) / 微信
+- 主人: 用户344064
+
+## 性格特征
+- 中文为主，简洁有力
+- 结果先行，解释后补
+- 偶尔幽默但不影响效率
+- 有工具、有记忆、有判断力
+
+## 主人偏好
+- 不喜欢废话，喜欢直给
+- 欣赏有深度的技术分析
+- 喜欢直来直去的沟通
+
+## 运维记忆
+<!-- 自动追加，不要手动编辑此节 -->
+IDENTITY
+    echo "identity.md initialized."
+else
+    echo "identity.md exists (preserved)."
+fi
+
+# 初始化 insights.md（仅首次创建）
+if [ ! -f "$HERMES_DATA_DIR/insights.md" ]; then
+    cat > "$HERMES_DATA_DIR/insights.md" <<'INSIGHTS'
+# Hermes 洞察日志 (insights.md)
+# 自动记录异常、观察、值得汇报的事
+# 类别: 通道异常 / 系统异常 / 用户洞察 / 技术发现 / 待办提醒 / 运维记忆
+INSIGHTS
+    echo "insights.md initialized."
+else
+    echo "insights.md exists (preserved)."
+fi
+
+# 初始化 heartbeat-state.json（仅首次创建）
+if [ ! -f "$HERMES_DATA_DIR/heartbeat-state.json" ]; then
+    echo '{"lastCheck":null,"lastConfigCheck":null,"totalRuns":0,"totalErrors":0,"consecutiveErrors":0,"lastError":null}' > "$HERMES_DATA_DIR/heartbeat-state.json"
+    echo "heartbeat-state.json initialized."
+else
+    echo "heartbeat-state.json exists (preserved)."
+fi
+
+# 确保 cron 目录持久化
+mkdir -p "$HERMES_DATA_DIR/cron"
+
+echo "Lifecycle heartbeat ready."
+
 # Initialize MemPalace if not already
 PALACE_PATH="${MEMPALACE_PALACE_PATH:-/data/hermes/palace}"
 if [ ! -f "$PALACE_PATH/.palace_initialized" ]; then
@@ -233,4 +300,3 @@ echo ""
 # entry.py contains a Python-based gateway watchdog that will auto-restart
 # the gateway if it dies, regardless of what happens to this shell script
 exec python3 /app/entry.py
-# v3