getDomain/domain-api/app/services/ops_agent_service.py

from __future__ import annotations

import hashlib
import json
import secrets
import shlex
import threading
from datetime import datetime, timedelta

from app.core.config import settings
from app.core.db import get_db
from app.services.ops_command_service import build_bash_command
from app.services.ops_job_service import create_ops_job, ensure_ops_schema, get_ops_job
from app.services.ops_template_service import build_ops_template_payload, get_ops_action_template
from app.services.runtime_settings_service import get_runtime_settings
from app.services.sensitive_words_service import get_sensitive_words_payload
from app.services.settings_service import get_settings_payload


_AGENT_SCHEMA_SQL = """
CREATE TABLE IF NOT EXISTS ops_node_tokens (
    id BIGSERIAL PRIMARY KEY,
    token_hash VARCHAR(128) NOT NULL UNIQUE,
    node_code VARCHAR(64) NOT NULL DEFAULT '',
    purpose VARCHAR(32) NOT NULL DEFAULT 'agent',
    issued_by VARCHAR(64) NOT NULL DEFAULT 'system',
    is_enabled BOOLEAN NOT NULL DEFAULT TRUE,
    expires_at TIMESTAMP,
    last_used_at TIMESTAMP,
    metadata_json JSONB,
    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
);

CREATE INDEX IF NOT EXISTS idx_ops_node_tokens_node_code
ON ops_node_tokens(node_code, is_enabled, purpose);

ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS node_code VARCHAR(64) NOT NULL DEFAULT '';
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS purpose VARCHAR(32) NOT NULL DEFAULT 'agent';
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS issued_by VARCHAR(64) NOT NULL DEFAULT 'system';
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS is_enabled BOOLEAN NOT NULL DEFAULT TRUE;
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS expires_at TIMESTAMP;
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS last_used_at TIMESTAMP;
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS metadata_json JSONB;
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP;
ALTER TABLE ops_node_tokens ADD COLUMN IF NOT EXISTS updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP;

CREATE TABLE IF NOT EXISTS ops_job_events (
    id BIGSERIAL PRIMARY KEY,
    job_id BIGINT REFERENCES ops_jobs(id) ON DELETE CASCADE,
    step_id BIGINT REFERENCES ops_job_steps(id) ON DELETE SET NULL,
    node_code VARCHAR(64) NOT NULL DEFAULT '',
    client_event_id VARCHAR(128) NOT NULL DEFAULT '',
    event_type VARCHAR(64) NOT NULL DEFAULT '',
    level VARCHAR(16) NOT NULL DEFAULT 'info',
    message TEXT NOT NULL DEFAULT '',
    payload_json JSONB,
    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
);

ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS job_id BIGINT;
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS step_id BIGINT;
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS node_code VARCHAR(64) NOT NULL DEFAULT '';
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS client_event_id VARCHAR(128) NOT NULL DEFAULT '';
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS event_type VARCHAR(64) NOT NULL DEFAULT '';
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS level VARCHAR(16) NOT NULL DEFAULT 'info';
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS message TEXT NOT NULL DEFAULT '';
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS payload_json JSONB;
ALTER TABLE ops_job_events ADD COLUMN IF NOT EXISTS created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP;

CREATE INDEX IF NOT EXISTS idx_ops_job_events_job_created
ON ops_job_events(job_id, created_at DESC);

CREATE UNIQUE INDEX IF NOT EXISTS uq_ops_job_events_job_client_event
ON ops_job_events(job_id, client_event_id)
WHERE client_event_id <> '';

ALTER TABLE ops_jobs ADD COLUMN IF NOT EXISTS last_agent_complete_request_id VARCHAR(128) NOT NULL DEFAULT '';
"""
_OPS_AGENT_SCHEMA_LOCK = threading.Lock()
_OPS_AGENT_SCHEMA_READY = False
_OPS_AGENT_SCHEMA_ADVISORY_LOCK_KEY = 90421802


def ensure_ops_agent_schema() -> None:
    global _OPS_AGENT_SCHEMA_READY
    ensure_ops_schema()
    if _OPS_AGENT_SCHEMA_READY:
        return
    with _OPS_AGENT_SCHEMA_LOCK:
        if _OPS_AGENT_SCHEMA_READY:
            return
        with get_db() as conn:
            conn.autocommit = False
            with conn.cursor() as cur:
                cur.execute("SELECT pg_advisory_xact_lock(%s)", (_OPS_AGENT_SCHEMA_ADVISORY_LOCK_KEY,))
                cur.execute(_AGENT_SCHEMA_SQL)
            conn.commit()
        _OPS_AGENT_SCHEMA_READY = True


def _hash_token(token: str) -> str:
    return hashlib.sha256(str(token or "").encode("utf-8")).hexdigest()


def _decode_json(value: object) -> dict:
    if isinstance(value, dict):
        return value
    if value in (None, ""):
        return {}
    try:
        return json.loads(value)
    except Exception:
        return {}


def _format_time(value: object) -> str:
    if isinstance(value, datetime):
        return value.isoformat(sep=" ", timespec="seconds")
    return ""


def _prefer_non_loopback_identity(primary: object, fallback: object) -> str:
    primary_value = str(primary or "").strip()
    fallback_value = str(fallback or "").strip()
    invalid_values = {"", "localhost", "localhost.localdomain", "127.0.0.1", "::1"}
    if primary_value and primary_value.lower() not in invalid_values:
        return primary_value
    return fallback_value


def _merge_detect_runtime_snapshot(cluster_metadata: dict, metadata: dict, current_load: int) -> dict:
    active_threads = int(cluster_metadata.get("active_threads", metadata.get("active_threads", 0)) or 0)
    max_threads = int(cluster_metadata.get("max_threads", metadata.get("max_threads", 0)) or 0)
    inferred_worker_online = bool(
        cluster_metadata.get("worker_online", metadata.get("worker_online", False))
        or cluster_metadata.get("service_running", metadata.get("service_running", False))
        or current_load > 0
        or active_threads > 0
    )
    inferred_detect_participating = bool(
        cluster_metadata.get("detect_participating", metadata.get("detect_participating", False))
        or current_load > 0
        or active_threads > 0
    )
    return {
        "worker_online": inferred_worker_online,
        "detect_participating": inferred_detect_participating,
        "active_threads": active_threads,
        "max_threads": max_threads,
        "current_load": current_load,
        "phase_label": str(cluster_metadata.get("phase_label", metadata.get("phase_label", "")) or "").strip(),
        "phase_detail": str(cluster_metadata.get("phase_detail", metadata.get("phase_detail", "")) or "").strip(),
        "recent_warning": str(cluster_metadata.get("recent_warning", metadata.get("recent_warning", "")) or "").strip(),
        "proxy_runtime_label": str(
            cluster_metadata.get("proxy_runtime_label", metadata.get("proxy_runtime_label", "")) or ""
        ).strip(),
        "proxy_runtime_reason": str(
            cluster_metadata.get("proxy_runtime_reason", metadata.get("proxy_runtime_reason", "")) or ""
        ).strip(),
        "updated_at": str(cluster_metadata.get("updated_at", metadata.get("updated_at", "")) or "").strip(),
    }


def _ops_job_event_level_label(level: object) -> str:
    normalized_level = str(level or "").strip().lower()
    mapping = {
        "debug": "调试",
        "info": "信息",
        "success": "成功",
        "warning": "警告",
        "error": "错误",
        "critical": "严重",
    }
    return mapping.get(normalized_level, normalized_level or "信息")


def _build_ops_job_event_row(row: tuple) -> dict:
    created_at = _format_time(row[9])
    payload = _decode_json(row[8])
    job_id = int(row[1]) if row[1] else 0
    event_id = int(row[0])
    event_type = str(row[5] or "")
    level = str(row[6] or "")
    message = str(row[7] or "")
    occurred_at = str(payload.get("occurred_at") or "").strip() or created_at
    summary_text = str(payload.get("summary_text") or "").strip() or message
    focus_ref = payload.get("focus_ref") if isinstance(payload.get("focus_ref"), dict) else {}
    if not focus_ref:
        focus_ref = {
            "kind": "ops_job_event",
            "job_id": job_id,
            "focus_event_id": event_id,
            "event_type": event_type,
        }
    return {
        "id": event_id,
        "event_key": f"job-event:{event_id}",
        "job_id": job_id,
        "step_id": int(row[2]) if row[2] else 0,
        "node_code": str(row[3] or ""),
        "client_event_id": str(row[4] or ""),
        "event_type": event_type,
        "level": level,
        "level_label": _ops_job_event_level_label(level),
        "message": message,
        "summary": message,
        "summary_text": summary_text,
        "payload": payload,
        "has_payload": bool(payload),
        "created_at": created_at,
        "occurred_at": occurred_at,
        "focus_ref": focus_ref,
        "ui_intent": {
            "kind": "job_detail",
            "job_id": job_id,
            "focus_event_id": event_id,
            "event_type": event_type,
        },
    }


def summarize_ops_job_events(events: list[dict], *, job_id: int = 0, limit: int = 50) -> dict:
    level_counts: dict[str, int] = {}
    event_type_counts: dict[str, int] = {}
    node_counts: dict[str, int] = {}
    latest_at = ""
    for event in list(events or []):
        event_level = str(event.get("level") or "").strip() or "info"
        event_type = str(event.get("event_type") or "").strip() or "event"
        node_code = str(event.get("node_code") or "").strip()
        occurred_at = str(event.get("occurred_at") or event.get("created_at") or "").strip()
        level_counts[event_level] = int(level_counts.get(event_level, 0) or 0) + 1
        event_type_counts[event_type] = int(event_type_counts.get(event_type, 0) or 0) + 1
        if node_code:
            node_counts[node_code] = int(node_counts.get(node_code, 0) or 0) + 1
        if occurred_at and occurred_at > latest_at:
            latest_at = occurred_at
    return {
        "job_id": int(job_id or 0),
        "returned_total": len(events or []),
        "level_counts": level_counts,
        "event_type_counts": event_type_counts,
        "node_counts": node_counts,
        "latest_at": latest_at,
        "filters": {
            "limit": int(limit or 50),
        },
    }


def _parse_time(value: object) -> datetime | None:
    if isinstance(value, datetime):
        return value
    raw_value = str(value or "").strip()
    if not raw_value:
        return None
    normalized = raw_value.replace("Z", "+00:00")
    try:
        return datetime.fromisoformat(normalized)
    except Exception:
        pass
    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d %H:%M:%S.%f"):
        try:
            return datetime.strptime(raw_value, fmt)
        except ValueError:
            continue
    return None


def _agent_error(message: str, detail_code: str, data: dict | None = None) -> tuple[bool, str, dict]:
    payload = dict(data or {})
    payload["detail_code"] = str(detail_code or "").strip()
    return False, message, payload


def _agent_state_from_context(
    *,
    last_seen_at: str,
    latest_token: dict,
    cluster_status: str,
    current_load: int,
) -> tuple[str, str, str]:
    now = datetime.now()
    last_seen_dt = _parse_time(last_seen_at)
    token_status = str(latest_token.get("token_status") or "absent").strip() or "absent"

    if last_seen_dt:
        current_time = datetime.now(last_seen_dt.tzinfo) if last_seen_dt.tzinfo else now
        age_seconds = max(0, int((current_time - last_seen_dt).total_seconds()))
        if age_seconds <= 180:
            if cluster_status == "busy" or int(current_load or 0) > 0:
                return "online_busy", "已接管", f"Agent 最近心跳 {age_seconds} 秒前，节点当前负载 {int(current_load or 0)}。"
            return "online", "已接管", f"Agent 最近心跳 {age_seconds} 秒前。"
        return "stale", "心跳过期", f"Agent 最近心跳 {age_seconds} 秒前，已超过在线阈值。"

    if token_status == "active":
        return "pending_bootstrap", "待接入", "已签发有效 Agent Token，但节点尚未完成 register/heartbeat。"
    if token_status == "expired":
        return "token_expired", "Token 过期", "已签发 Agent Token，但该 Token 已过期。"
    if token_status == "disabled":
        return "token_disabled", "Token 已禁用", "最近一枚 Agent Token 已被禁用，需重新签发。"

    if cluster_status in {"online", "busy"}:
        return "runtime_only", "仅运行态在线", "控制面能看到 runtime 心跳，但 Node Agent 尚未接管。"
    return "unmanaged", "未接管", "当前节点尚未生成有效 Agent 接入状态。"


def _ssh_access_context(node: dict) -> tuple[bool, str, str]:
    ssh_host = str(node.get("ssh_host") or "").strip()
    ssh_user = str(node.get("ssh_user") or "").strip()
    if ssh_host and ssh_user:
        return True, "SSH 已备好", f"已保存 SSH 入口 {ssh_user}@{ssh_host}，后续可由海外控制面 SSH 执行器接管。"
    if ssh_host and not ssh_user:
        return False, "SSH 待完善", "已填写 SSH 主机，但还缺少 ssh_user。"
    if ssh_user and not ssh_host:
        return False, "SSH 待完善", "已填写 ssh_user，但还缺少 ssh_host。"
    return False, "SSH 未配置", "当前还没有保存 SSH 入口。"


def _remote_access_context(node: dict) -> tuple[str, str, str, bool]:
    is_managed = bool(node.get("is_managed", False))
    is_enabled = bool(node.get("is_enabled", False))
    is_agent_online = bool(node.get("is_agent_online", False))
    ssh_ready, ssh_label, ssh_reason = _ssh_access_context(node)

    if not is_managed:
        return "unmanaged", "未纳管", "当前节点还没有进入海外控制面的托管清单。", False
    if not is_enabled:
        return "disabled", "已停用", "当前节点已纳管，但被标记为停用。", False
    if is_agent_online and ssh_ready:
        return "hybrid_ready", "Agent + SSH", "Node Agent 已在线，SSH 入口也已备好。", True
    if is_agent_online:
        return "agent_ready", "Agent 在线", "当前节点已经进入标准自动执行器就绪状态。", True
    if ssh_ready:
        return (
            "ssh_ready",
            "SSH 可执行",
            f"{ssh_reason} 当前已可执行日志、诊断和部分服务控制；正式发布与标准 Rollout 仍优先依赖 Node Agent。",
            True,
        )
    return "agent_pending", "待接入", "当前还没有可用 Agent，也没有保存 SSH 入口。", False


def _delivery_queue_context(metadata: dict) -> tuple[str, str, str, dict]:
    snapshot = dict((metadata or {}).get("delivery_queue") or {})
    pending_count = int(snapshot.get("pending_count", 0) or 0)
    dead_letter_count = int(snapshot.get("dead_letter_count", 0) or 0)
    last_flush_at = str(snapshot.get("last_flush_at") or "").strip()
    oldest_pending_at = str(snapshot.get("oldest_pending_at") or "").strip()
    oldest_dead_letter_at = str(snapshot.get("oldest_dead_letter_at") or "").strip()

    if not snapshot:
        return "unknown", "未上报", "Node Agent 尚未上报回执队列状态。", {}
    if dead_letter_count > 0:
        reason = f"当前存在 {dead_letter_count} 条死信记录，建议优先查看节点日志或诊断编排。"
        if oldest_dead_letter_at:
            reason += f" 最早死信时间：{oldest_dead_letter_at}。"
        return "dead_letter", f"死信 {dead_letter_count}", reason, snapshot
    if pending_count > 0:
        reason = f"当前存在 {pending_count} 条待重试回执，Node Agent 会继续自动回放。"
        if oldest_pending_at:
            reason += f" 最早积压时间：{oldest_pending_at}。"
        return "retrying", f"待重试 {pending_count}", reason, snapshot
    reason = "当前没有待重试回执，也没有死信记录。"
    if last_flush_at:
        reason += f" 最近一次队列冲刷：{last_flush_at}。"
    return "healthy", "正常", reason, snapshot


def _summarize_managed_node_states(nodes: list[dict]) -> dict:
    status_counts: dict[str, int] = {}
    remote_access_counts: dict[str, int] = {}
    delivery_queue_state_counts: dict[str, int] = {}
    for item in nodes:
        state = str(item.get("agent_state") or "").strip() or "unknown"
        status_counts[state] = int(status_counts.get(state, 0) or 0) + 1
        remote_state = str(item.get("remote_access_state") or "").strip() or "unknown"
        remote_access_counts[remote_state] = int(remote_access_counts.get(remote_state, 0) or 0) + 1
        queue_state = str(item.get("delivery_queue_state") or "").strip() or "unknown"
        delivery_queue_state_counts[queue_state] = int(delivery_queue_state_counts.get(queue_state, 0) or 0) + 1
    managed_nodes = [item for item in nodes if bool(item.get("is_managed", False))]
    participating_nodes = [item for item in nodes if bool(item.get("is_current_participant", False))]
    dispatch_active_nodes = [item for item in nodes if bool(item.get("is_dispatch_active", False))]
    recent_only_nodes = [
        item for item in nodes if str(item.get("participation_state") or "").strip() == "recent_throughput"
    ]
    standby_nodes = [item for item in nodes if str(item.get("participation_state") or "").strip() == "standby"]
    load_syncing_nodes = [item for item in nodes if str(item.get("participation_state") or "").strip() == "load_syncing"]
    return {
        "status_counts": status_counts,
        "online": int(status_counts.get("online", 0) or 0) + int(status_counts.get("online_busy", 0) or 0),
        "stale": int(status_counts.get("stale", 0) or 0),
        "pending_bootstrap": int(status_counts.get("pending_bootstrap", 0) or 0),
        "runtime_only": int(status_counts.get("runtime_only", 0) or 0),
        "unmanaged": int(status_counts.get("unmanaged", 0) or 0),
        "token_issue": int(status_counts.get("token_expired", 0) or 0) + int(status_counts.get("token_disabled", 0) or 0),
        "ssh_ready": int(remote_access_counts.get("ssh_ready", 0) or 0) + int(remote_access_counts.get("hybrid_ready", 0) or 0),
        "agent_ready": int(remote_access_counts.get("agent_ready", 0) or 0) + int(remote_access_counts.get("hybrid_ready", 0) or 0),
        "remote_access_ready": sum(1 for item in nodes if bool(item.get("is_remote_access_ready", False))),
        "managed_total": len(managed_nodes),
        "managed_enabled": sum(1 for item in managed_nodes if bool(item.get("is_enabled", False))),
        "queue_retrying_nodes": sum(
            1 for item in managed_nodes if str(item.get("delivery_queue_state") or "").strip() == "retrying"
        ),
        "queue_dead_letter_nodes": sum(
            1 for item in managed_nodes if str(item.get("delivery_queue_state") or "").strip() == "dead_letter"
        ),
        "queue_pending_records": sum(int(item.get("delivery_queue_pending_count", 0) or 0) for item in managed_nodes),
        "queue_dead_letter_records": sum(
            int(item.get("delivery_queue_dead_letter_count", 0) or 0) for item in managed_nodes
        ),
        "participating": len(participating_nodes),
        "dispatch_active": len(dispatch_active_nodes),
        "recent_only": len(recent_only_nodes),
        "standby": len(standby_nodes),
        "load_syncing": len(load_syncing_nodes),
        "non_participating": len(standby_nodes) + len(load_syncing_nodes),
        "total": len(nodes),
        "remote_access_state_counts": remote_access_counts,
        "delivery_queue_state_counts": delivery_queue_state_counts,
    }


def _normalize_detect_participation_rows(payload: dict | None = None) -> dict[str, dict]:
    normalized_payload = dict(payload or {})
    if "detect" in normalized_payload and isinstance(normalized_payload.get("detect"), dict):
        normalized_payload = dict(normalized_payload.get("detect") or {})

    rows: dict[str, dict] = {}
    for collection_key in ("participating_nodes", "non_participating_nodes", "standby_nodes"):
        for item in list(normalized_payload.get(collection_key) or []):
            node_code = str((item or {}).get("node_code") or "").strip()
            if not node_code:
                continue
            rows[node_code] = dict(item or {})
    return rows


def append_ops_job_event(
    *,
    job_id: int | None,
    step_id: int | None = None,
    node_code: str = "",
    client_event_id: str = "",
    event_type: str,
    message: str,
    level: str = "info",
    payload: dict | None = None,
) -> int:
    ensure_ops_agent_schema()
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                INSERT INTO ops_job_events (
                    job_id, step_id, node_code, client_event_id, event_type, level, message, payload_json, created_at
                ) VALUES (%s, %s, %s, %s, %s, %s, %s, %s, CURRENT_TIMESTAMP)
                ON CONFLICT (job_id, client_event_id) WHERE client_event_id <> ''
                DO UPDATE SET
                    step_id = COALESCE(EXCLUDED.step_id, ops_job_events.step_id),
                    node_code = CASE WHEN EXCLUDED.node_code <> '' THEN EXCLUDED.node_code ELSE ops_job_events.node_code END,
                    event_type = EXCLUDED.event_type,
                    level = EXCLUDED.level,
                    message = EXCLUDED.message,
                    payload_json = EXCLUDED.payload_json
                RETURNING id
                """,
                (
                    int(job_id) if job_id else None,
                    int(step_id) if step_id else None,
                    str(node_code or "").strip(),
                    str(client_event_id or "").strip()[:128],
                    str(event_type or "").strip() or "event",
                    str(level or "info").strip() or "info",
                    str(message or "").strip()[:2000],
                    json.dumps(payload or {}, ensure_ascii=False),
                ),
            )
            event_id = int(cur.fetchone()[0])
        conn.commit()
    return event_id


def list_ops_job_events(job_id: int, limit: int = 50) -> list[dict]:
    ensure_ops_agent_schema()
    safe_limit = min(max(int(limit or 50), 1), 200)
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT id, job_id, step_id, node_code, client_event_id, event_type, level, message, payload_json, created_at
                FROM ops_job_events
                WHERE job_id = %s
                ORDER BY id DESC
                LIMIT %s
                """,
                (int(job_id), safe_limit),
            )
            rows = cur.fetchall()
    return [_build_ops_job_event_row(row) for row in rows]


def get_ops_job_event(event_id: int) -> dict:
    ensure_ops_agent_schema()
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT id, job_id, step_id, node_code, client_event_id, event_type, level, message, payload_json, created_at
                FROM ops_job_events
                WHERE id = %s
                LIMIT 1
                """,
                (int(event_id),),
            )
            row = cur.fetchone()
    return _build_ops_job_event_row(row) if row else {}


def list_ops_job_events_for_jobs(job_ids: list[int], limit: int = 120) -> list[dict]:
    ensure_ops_agent_schema()
    normalized_job_ids = sorted({int(item) for item in list(job_ids or []) if int(item or 0) > 0})
    if not normalized_job_ids:
        return []
    safe_limit = min(max(int(limit or 120), 1), 800)
    placeholders = ", ".join(["%s"] * len(normalized_job_ids))
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                f"""
                SELECT id, job_id, step_id, node_code, client_event_id, event_type, level, message, payload_json, created_at
                FROM ops_job_events
                WHERE job_id IN ({placeholders})
                ORDER BY id DESC
                LIMIT %s
                """,
                (*normalized_job_ids, safe_limit),
            )
            rows = cur.fetchall()
    return [_build_ops_job_event_row(row) for row in rows]


def list_managed_nodes_with_agent_state(*, participation_payload: dict | None = None) -> dict:
    from app.services.cluster_runtime_service import get_cluster_snapshot
    from app.services.ops_job_service import list_managed_nodes

    ensure_ops_agent_schema()
    managed_nodes = list_managed_nodes()
    cluster_nodes = list((get_cluster_snapshot().get("nodes") or []))
    cluster_map = {
        str(item.get("node_code") or "").strip(): item
        for item in cluster_nodes
        if str(item.get("node_code") or "").strip()
    }
    detect_participation_map = _normalize_detect_participation_rows(participation_payload)
    if not detect_participation_map:
        try:
            from app.services.runtime_status_service import get_runtime_status

            runtime_payload = get_runtime_status()
            detect_participation_map = _normalize_detect_participation_rows(runtime_payload.get("detect") or {})
        except Exception:
            detect_participation_map = {}

    latest_tokens: dict[str, dict] = {}
    latest_jobs: dict[str, dict] = {}
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT DISTINCT ON (node_code)
                    node_code, id, issued_by, is_enabled, expires_at, last_used_at, metadata_json, created_at, updated_at
                FROM ops_node_tokens
                WHERE purpose = 'agent'
                ORDER BY node_code ASC, created_at DESC, id DESC
                """
            )
            for row in cur.fetchall():
                node_code = str(row[0] or "").strip()
                if not node_code:
                    continue
                expires_at = _format_time(row[4])
                token_status = "active"
                expires_dt = _parse_time(row[4])
                if not bool(row[3]):
                    token_status = "disabled"
                elif expires_dt:
                    current_time = datetime.now(expires_dt.tzinfo) if expires_dt.tzinfo else datetime.now()
                    if expires_dt < current_time:
                        token_status = "expired"
                latest_tokens[node_code] = {
                    "record_id": int(row[1]),
                    "issued_by": str(row[2] or ""),
                    "is_enabled": bool(row[3]),
                    "expires_at": expires_at,
                    "last_used_at": _format_time(row[5]),
                    "metadata": _decode_json(row[6]),
                    "created_at": _format_time(row[7]),
                    "updated_at": _format_time(row[8]),
                    "token_status": token_status,
                }

            cur.execute(
                """
                SELECT DISTINCT ON (target_node_code)
                    target_node_code, id, job_code, action, status, execution_mode, created_at, updated_at
                FROM ops_jobs
                WHERE target_node_code <> ''
                ORDER BY target_node_code ASC, created_at DESC, id DESC
                """
            )
            for row in cur.fetchall():
                node_code = str(row[0] or "").strip()
                if not node_code:
                    continue
                latest_jobs[node_code] = {
                    "id": int(row[1]),
                    "job_code": str(row[2] or ""),
                    "action": str(row[3] or ""),
                    "status": str(row[4] or ""),
                    "execution_mode": str(row[5] or ""),
                    "created_at": _format_time(row[6]),
                    "updated_at": _format_time(row[7]),
                }

    enriched_nodes: list[dict] = []
    seen_node_codes: set[str] = set()
    for node in managed_nodes:
        node_code = str(node.get("node_code") or "").strip()
        if not node_code:
            continue
        seen_node_codes.add(node_code)
        metadata = dict(node.get("metadata") or {})
        cluster_node = cluster_map.get(node_code, {})
        cluster_metadata = dict(cluster_node.get("metadata") or {})
        latest_token = latest_tokens.get(node_code, {})
        latest_job = latest_jobs.get(node_code, {})
        participation_row = dict(detect_participation_map.get(node_code) or {})
        last_seen_at = str(node.get("last_seen_at") or "").strip() or str(metadata.get("last_seen_at") or "").strip()
        cluster_status = str(cluster_node.get("status") or metadata.get("cluster_status") or "").strip()
        current_load = int(cluster_node.get("current_load", 0) or 0)
        (
            delivery_queue_state,
            delivery_queue_label,
            delivery_queue_reason,
            delivery_queue_snapshot,
        ) = _delivery_queue_context(metadata)
        agent_state, agent_state_label, agent_state_reason = _agent_state_from_context(
            last_seen_at=last_seen_at,
            latest_token=latest_token,
            cluster_status=cluster_status,
            current_load=current_load,
        )
        capabilities = list(metadata.get("capabilities") or [])
        detect_runtime = _merge_detect_runtime_snapshot(cluster_metadata, metadata, current_load)
        active_job_snapshot = {
            "job_code": str(cluster_metadata.get("active_job_code", metadata.get("active_job_code", "")) or "").strip(),
            "status": str(cluster_metadata.get("active_job_status", metadata.get("active_job_status", "")) or "").strip(),
            "items_total": int(cluster_metadata.get("job_items_total", metadata.get("job_items_total", 0)) or 0),
            "items_claimed": int(cluster_metadata.get("job_items_claimed", metadata.get("job_items_claimed", 0)) or 0),
            "items_running": int(cluster_metadata.get("job_items_running", metadata.get("job_items_running", 0)) or 0),
            "items_completed": int(cluster_metadata.get("job_items_completed", metadata.get("job_items_completed", 0)) or 0),
        }
        merged_node = {
            **node,
            "is_managed": True,
            "last_seen_at": last_seen_at,
            "agent_state": agent_state,
            "agent_state_label": agent_state_label,
            "agent_state_reason": agent_state_reason,
            "is_agent_online": agent_state in {"online", "online_busy"},
            "capabilities": capabilities,
            "capabilities_count": len(capabilities),
            "agent_version": str(metadata.get("agent_version") or "").strip(),
            "agent_hostname": str(metadata.get("hostname") or "").strip(),
            "agent_ip": str(metadata.get("ip") or "").strip(),
            "cluster_hostname": _prefer_non_loopback_identity(cluster_node.get("hostname"), metadata.get("hostname")),
            "cluster_ip": _prefer_non_loopback_identity(cluster_node.get("ip"), metadata.get("ip")),
            "cluster_status": cluster_status,
            "cluster_current_load": current_load,
            "current_load": current_load,
            "detect_runtime": detect_runtime,
            "runtime_state": detect_runtime,
            "active_job": active_job_snapshot,
            "cluster_last_heartbeat_at": str(cluster_node.get("last_heartbeat_at") or metadata.get("last_heartbeat_at") or "").strip(),
            "cluster_is_effective_worker": bool(cluster_node.get("is_effective_worker", metadata.get("is_effective_worker", False))),
            "cluster_detect_participating": bool(
                participation_row.get("is_current_participant", False)
                or cluster_node.get("detect_participating", metadata.get("detect_participating", False))
            ),
            "participation_state": str(participation_row.get("participation_state") or "").strip(),
            "participation_label": str(participation_row.get("participation_label") or "").strip(),
            "participation_reason": str(participation_row.get("participation_reason") or "").strip(),
            "is_current_participant": bool(participation_row.get("is_current_participant", False)),
            "is_dispatch_active": bool(participation_row.get("is_dispatch_active", False)),
            "items_total": int(participation_row.get("items_total", 0) or 0),
            "items_pending": int(participation_row.get("items_pending", 0) or 0),
            "items_claimed": int(participation_row.get("items_claimed", 0) or 0),
            "items_running": int(participation_row.get("items_running", 0) or 0),
            "items_completed": int(participation_row.get("items_completed", 0) or 0),
            "items_failed": int(participation_row.get("items_failed", 0) or 0),
            "processed_recent": int(participation_row.get("processed_recent", 0) or 0),
            "processed_per_minute": float(participation_row.get("processed_per_minute", 0) or 0),
            "delivery_queue_state": delivery_queue_state,
            "delivery_queue_label": delivery_queue_label,
            "delivery_queue_reason": delivery_queue_reason,
            "delivery_queue_pending_count": int(delivery_queue_snapshot.get("pending_count", 0) or 0),
            "delivery_queue_dead_letter_count": int(delivery_queue_snapshot.get("dead_letter_count", 0) or 0),
            "delivery_queue_last_flush_at": str(delivery_queue_snapshot.get("last_flush_at") or "").strip(),
            "delivery_queue_oldest_pending_at": str(delivery_queue_snapshot.get("oldest_pending_at") or "").strip(),
            "delivery_queue_oldest_pending_request_id": str(
                delivery_queue_snapshot.get("oldest_pending_request_id") or ""
            ).strip(),
            "delivery_queue_oldest_pending_kind": str(delivery_queue_snapshot.get("oldest_pending_kind") or "").strip(),
            "delivery_queue_oldest_dead_letter_at": str(
                delivery_queue_snapshot.get("oldest_dead_letter_at") or ""
            ).strip(),
            "delivery_queue_oldest_dead_letter_request_id": str(
                delivery_queue_snapshot.get("oldest_dead_letter_request_id") or ""
            ).strip(),
            "delivery_queue_oldest_dead_letter_kind": str(
                delivery_queue_snapshot.get("oldest_dead_letter_kind") or ""
            ).strip(),
            "latest_token": latest_token,
            "latest_job": latest_job,
        }
        ssh_ready, ssh_access_label, ssh_access_reason = _ssh_access_context(merged_node)
        remote_access_state, remote_access_label, remote_access_reason, is_remote_access_ready = _remote_access_context(
            merged_node
        )

        enriched_nodes.append(
            {
                **merged_node,
                "has_ssh_access": ssh_ready,
                "ssh_access_label": ssh_access_label,
                "ssh_access_reason": ssh_access_reason,
                "remote_access_state": remote_access_state,
                "remote_access_label": remote_access_label,
                "remote_access_reason": remote_access_reason,
                "is_remote_access_ready": is_remote_access_ready,
            }
        )

    for node_code, cluster_node in cluster_map.items():
        if not node_code or node_code in seen_node_codes:
            continue
        metadata = dict(cluster_node.get("metadata") or {})
        cluster_metadata = dict(cluster_node.get("metadata") or {})
        latest_token = latest_tokens.get(node_code, {})
        latest_job = latest_jobs.get(node_code, {})
        participation_row = dict(detect_participation_map.get(node_code) or {})
        cluster_status = str(cluster_node.get("status") or "").strip()
        current_load = int(cluster_node.get("current_load", 0) or 0)
        (
            delivery_queue_state,
            delivery_queue_label,
            delivery_queue_reason,
            delivery_queue_snapshot,
        ) = _delivery_queue_context(metadata)
        agent_state, agent_state_label, agent_state_reason = _agent_state_from_context(
            last_seen_at="",
            latest_token=latest_token,
            cluster_status=cluster_status,
            current_load=current_load,
        )
        capabilities = list(metadata.get("capabilities") or [])
        detect_runtime = _merge_detect_runtime_snapshot(cluster_metadata, metadata, current_load)
        active_job_snapshot = {
            "job_code": str(cluster_metadata.get("active_job_code", metadata.get("active_job_code", "")) or "").strip(),
            "status": str(cluster_metadata.get("active_job_status", metadata.get("active_job_status", "")) or "").strip(),
            "items_total": int(cluster_metadata.get("job_items_total", metadata.get("job_items_total", 0)) or 0),
            "items_claimed": int(cluster_metadata.get("job_items_claimed", metadata.get("job_items_claimed", 0)) or 0),
            "items_running": int(cluster_metadata.get("job_items_running", metadata.get("job_items_running", 0)) or 0),
            "items_completed": int(cluster_metadata.get("job_items_completed", metadata.get("job_items_completed", 0)) or 0),
        }
        fallback_node = {
            "node_code": node_code,
            "region": str(cluster_node.get("region") or "").strip(),
            "role": str(cluster_node.get("role") or "").strip(),
            "title": str(cluster_node.get("hostname") or node_code).strip(),
            "ssh_host": "",
            "ssh_port": 22,
            "ssh_user": "",
            "auth_mode": "key",
            "deploy_channel": "",
            "is_enabled": False,
            "is_managed": False,
            "metadata": {
                **metadata,
                "imported_from_cluster_fallback": True,
            },
            "created_at": "",
            "last_seen_at": "",
            "updated_at": "",
            "agent_state": agent_state,
            "agent_state_label": agent_state_label,
            "agent_state_reason": agent_state_reason,
            "is_agent_online": agent_state in {"online", "online_busy"},
            "capabilities": capabilities,
            "capabilities_count": len(capabilities),
            "agent_version": str(metadata.get("agent_version") or "").strip(),
            "agent_hostname": _prefer_non_loopback_identity(cluster_node.get("hostname"), metadata.get("hostname")),
            "agent_ip": _prefer_non_loopback_identity(cluster_node.get("ip"), metadata.get("ip")),
            "cluster_hostname": _prefer_non_loopback_identity(cluster_node.get("hostname"), metadata.get("hostname")),
            "cluster_ip": _prefer_non_loopback_identity(cluster_node.get("ip"), metadata.get("ip")),
            "cluster_status": cluster_status,
            "cluster_current_load": current_load,
            "current_load": current_load,
            "detect_runtime": detect_runtime,
            "runtime_state": detect_runtime,
            "active_job": active_job_snapshot,
            "cluster_last_heartbeat_at": str(cluster_node.get("last_heartbeat_at") or "").strip(),
            "cluster_is_effective_worker": bool(cluster_node.get("is_effective_worker", False)),
            "cluster_detect_participating": bool(
                participation_row.get("is_current_participant", False) or cluster_node.get("detect_participating", False)
            ),
            "participation_state": str(participation_row.get("participation_state") or "").strip(),
            "participation_label": str(participation_row.get("participation_label") or "").strip(),
            "participation_reason": str(participation_row.get("participation_reason") or "").strip(),
            "is_current_participant": bool(participation_row.get("is_current_participant", False)),
            "is_dispatch_active": bool(participation_row.get("is_dispatch_active", False)),
            "items_total": int(participation_row.get("items_total", 0) or 0),
            "items_pending": int(participation_row.get("items_pending", 0) or 0),
            "items_claimed": int(participation_row.get("items_claimed", 0) or 0),
            "items_running": int(participation_row.get("items_running", 0) or 0),
            "items_completed": int(participation_row.get("items_completed", 0) or 0),
            "items_failed": int(participation_row.get("items_failed", 0) or 0),
            "processed_recent": int(participation_row.get("processed_recent", 0) or 0),
            "processed_per_minute": float(participation_row.get("processed_per_minute", 0) or 0),
            "delivery_queue_state": delivery_queue_state,
            "delivery_queue_label": delivery_queue_label,
            "delivery_queue_reason": delivery_queue_reason,
            "delivery_queue_pending_count": int(delivery_queue_snapshot.get("pending_count", 0) or 0),
            "delivery_queue_dead_letter_count": int(delivery_queue_snapshot.get("dead_letter_count", 0) or 0),
            "delivery_queue_last_flush_at": str(delivery_queue_snapshot.get("last_flush_at") or "").strip(),
            "delivery_queue_oldest_pending_at": str(delivery_queue_snapshot.get("oldest_pending_at") or "").strip(),
            "delivery_queue_oldest_pending_request_id": str(
                delivery_queue_snapshot.get("oldest_pending_request_id") or ""
            ).strip(),
            "delivery_queue_oldest_pending_kind": str(delivery_queue_snapshot.get("oldest_pending_kind") or "").strip(),
            "delivery_queue_oldest_dead_letter_at": str(
                delivery_queue_snapshot.get("oldest_dead_letter_at") or ""
            ).strip(),
            "delivery_queue_oldest_dead_letter_request_id": str(
                delivery_queue_snapshot.get("oldest_dead_letter_request_id") or ""
            ).strip(),
            "delivery_queue_oldest_dead_letter_kind": str(
                delivery_queue_snapshot.get("oldest_dead_letter_kind") or ""
            ).strip(),
            "latest_token": latest_token,
            "latest_job": latest_job,
        }
        ssh_ready, ssh_access_label, ssh_access_reason = _ssh_access_context(fallback_node)
        remote_access_state, remote_access_label, remote_access_reason, is_remote_access_ready = _remote_access_context(
            fallback_node
        )
        enriched_nodes.append(
            {
                **fallback_node,
                "has_ssh_access": ssh_ready,
                "ssh_access_label": ssh_access_label,
                "ssh_access_reason": ssh_access_reason,
                "remote_access_state": remote_access_state,
                "remote_access_label": remote_access_label,
                "remote_access_reason": remote_access_reason,
                "is_remote_access_ready": is_remote_access_ready,
            }
        )

    enriched_nodes.sort(
        key=lambda item: (
            0 if bool(item.get("is_managed", False)) else 1,
            str(item.get("region") or ""),
            str(item.get("role") or ""),
            str(item.get("node_code") or ""),
        )
    )

    return {
        "nodes": enriched_nodes,
        "summary": _summarize_managed_node_states(enriched_nodes),
    }


def _build_managed_node_handover_stage(node: dict) -> dict:
    normalized_node = dict(node or {})
    node_code = str(normalized_node.get("node_code") or "").strip()
    agent_state = str(normalized_node.get("agent_state") or "").strip()
    agent_state_label = str(normalized_node.get("agent_state_label") or "").strip() or "未知"
    agent_state_reason = str(normalized_node.get("agent_state_reason") or "").strip()
    remote_access_state = str(normalized_node.get("remote_access_state") or "").strip()
    remote_access_label = str(normalized_node.get("remote_access_label") or "").strip() or "未就绪"
    remote_access_reason = str(normalized_node.get("remote_access_reason") or "").strip()
    token_status = str(((normalized_node.get("latest_token") or {}).get("token_status") or "")).strip()
    is_managed = bool(normalized_node.get("is_managed", False))
    is_enabled = bool(normalized_node.get("is_enabled", False))
    has_ssh_access = bool(normalized_node.get("has_ssh_access", False))

    if not is_managed:
        return {
            "code": "unmanaged",
            "label": "未纳管",
            "summary": "节点已出现在集群视图，但还没保存到海外控制面的托管清单。",
            "next_step": {
                "code": "save_managed_node",
                "label": "先保存节点资料",
                "summary": "先补齐 SSH 入口、身份和部署通道，再生成 Node Agent 接入方案。",
                "ui_intent": {
                    "kind": "managed_node_edit",
                    "node_code": node_code,
                },
            },
        }
    if not is_enabled:
        return {
            "code": "disabled",
            "label": "已停用",
            "summary": "节点虽然已经纳管，但当前被标记为停用，标准运维动作不会继续下发。",
            "next_step": {
                "code": "enable_managed_node",
                "label": "检查并重新启用",
                "summary": "确认该节点仍应参与运维后，先在托管节点面板重新启用。",
                "ui_intent": {
                    "kind": "managed_node_edit",
                    "node_code": node_code,
                },
            },
        }
    if agent_state in {"online", "online_busy"}:
        return {
            "code": "ready",
            "label": "接管完成",
            "summary": "Node Agent 已经在线，节点已经进入标准自动执行链路。",
            "next_step": {
                "code": "run_acceptance",
                "label": "执行接管验收",
                "summary": "建议立刻跑一轮 onboarding.acceptance 或标准巡检，确认日志、健康快照和发布链路都打通。",
                "ui_intent": {
                    "kind": "open_playbook_dialog",
                    "playbook_key": "onboarding.acceptance",
                    "target_node_codes": [node_code] if node_code else [],
                    "execution_mode": "remote-agent",
                    "auto_approve": True,
                },
            },
        }
    if agent_state == "stale":
        return {
            "code": "stale",
            "label": "心跳过期",
            "summary": "节点曾经接入过，但 Node Agent 心跳已经过期，当前不能视为稳定可控。",
            "next_step": {
                "code": "recover_agent",
                "label": "优先恢复 Agent",
                "summary": "先检查 domaincheck-node-agent 服务、网络连通性和控制面地址，再决定是否重新签发接入方案。",
                "ui_intent": {
                    "kind": "managed_node_handover",
                    "node_code": node_code,
                },
            },
        }
    if token_status in {"expired", "disabled"} or agent_state in {"token_expired", "token_disabled"}:
        return {
            "code": "token_issue",
            "label": "Token 失效",
            "summary": "最近一枚 Node Agent Token 已不可继续使用，需要重新生成接入方案。",
            "next_step": {
                "code": "reissue_bootstrap_plan",
                "label": "重新签发接入方案",
                "summary": "重新生成 bootstrap env 和一键落地片段，再到目标节点执行。",
                "ui_intent": {
                    "kind": "managed_node_handover",
                    "node_code": node_code,
                },
            },
        }
    if agent_state == "pending_bootstrap":
        return {
            "code": "pending_bootstrap",
            "label": "待执行接入",
            "summary": "控制面已经签发了有效 Token，当前卡在目标节点尚未执行 bootstrap 片段。",
            "next_step": {
                "code": "execute_bootstrap_plan",
                "label": "去目标机执行接入片段",
                "summary": "把 env、脚本或一键落地片段拷到目标机执行，然后观察 register / heartbeat 是否建立。",
                "ui_intent": {
                    "kind": "managed_node_handover",
                    "node_code": node_code,
                },
            },
        }
    if remote_access_state == "ssh_ready" or has_ssh_access:
        return {
            "code": "ssh_ready",
            "label": remote_access_label or "SSH 已备好",
            "summary": "SSH 入口已经可用，可以从海外控制面统一生成并下发 Node Agent 接入方案。",
            "next_step": {
                "code": "issue_bootstrap_plan",
                "label": "生成接入方案",
                "summary": "下一步就是签发 bootstrap env、脚本和一键落地命令，然后在目标节点启动 domaincheck-node-agent。",
                "ui_intent": {
                    "kind": "managed_node_handover",
                    "node_code": node_code,
                },
            },
        }
    if agent_state == "runtime_only":
        return {
            "code": "runtime_only",
            "label": agent_state_label or "仅运行态在线",
            "summary": "控制面能看到 runtime 心跳，但 Node Agent 还没进入标准接管链路。",
            "next_step": {
                "code": "complete_ssh_profile",
                "label": "补齐 SSH 入口并生成接入方案",
                "summary": "先把 SSH 主机和账号补齐，再从海外控制面统一签发 Node Agent 接入方案。",
                "ui_intent": {
                    "kind": "managed_node_edit",
                    "node_code": node_code,
                },
            },
        }
    return {
        "code": "profile_incomplete",
        "label": remote_access_label or agent_state_label or "待完善",
        "summary": remote_access_reason or agent_state_reason or "当前接管资料还不完整，尚不能进入标准自动执行链路。",
        "next_step": {
            "code": "complete_ssh_profile",
            "label": "补齐节点接入资料",
            "summary": "至少补齐 SSH 入口，再生成标准 Node Agent 接入方案。",
            "ui_intent": {
                "kind": "managed_node_edit",
                "node_code": node_code,
            },
        },
    }


def _build_managed_node_handover_payload(
    *,
    node: dict,
    control_plane_base_url: str = "",
    root_dir: str = "",
) -> dict:
    normalized_node = dict(node or {})
    normalized_node_code = str(normalized_node.get("node_code") or "").strip()
    if not normalized_node_code:
        return {}

    node = normalized_node
    metadata = dict(node.get("metadata") or {})
    latest_token = dict(node.get("latest_token") or {})
    latest_job = dict(node.get("latest_job") or {})
    control_plane_value = _normalize_control_plane_base_url(control_plane_base_url)
    if not control_plane_value:
        control_plane_value = (
            _normalize_control_plane_base_url(str(metadata.get("ops_control_plane_base_url") or ""))
            or f"http://127.0.0.1:{settings.api_port}"
        )
    resolved_root_dir = (
        str(root_dir or metadata.get("root_dir") or metadata.get("deploy_root_dir") or "/opt/domaincheck").strip()
        or "/opt/domaincheck"
    )

    stage = _build_managed_node_handover_stage(node)

    blocking_items: list[str] = []
    attention_items: list[str] = []

    if not bool(node.get("is_managed", False)):
        blocking_items.append("当前节点还未纳入海外控制面的托管清单。")
    if bool(node.get("is_managed", False)) and not bool(node.get("is_enabled", False)):
        blocking_items.append("当前节点已纳管，但被标记为停用。")
    if not bool(node.get("has_ssh_access", False)):
        attention_items.append("当前还没有完整 SSH 入口，无法从海外控制面直接接管该节点。")

    agent_state = str(node.get("agent_state") or "").strip()
    if agent_state == "pending_bootstrap":
        blocking_items.append("已签发有效 Token，但目标节点尚未执行 bootstrap 片段。")
    elif agent_state == "stale":
        blocking_items.append("Node Agent 心跳已过期，需先恢复服务或重新接入。")
    elif agent_state in {"token_expired", "token_disabled"}:
        blocking_items.append("最近一枚 Node Agent Token 已失效，需要重新签发。")

    if str(node.get("delivery_queue_state") or "").strip() == "dead_letter":
        attention_items.append(
            f"当前节点存在 {int(node.get('delivery_queue_dead_letter_count', 0) or 0)} 条死信回执，建议先查看 delivery queue。"
        )
    elif str(node.get("delivery_queue_state") or "").strip() == "retrying":
        attention_items.append(
            f"当前节点还有 {int(node.get('delivery_queue_pending_count', 0) or 0)} 条待重试回执，Node Agent 会继续自动回放。"
        )

    return {
        "node_code": normalized_node_code,
        "summary": str(stage.get("summary") or "").strip(),
        "stage": stage,
        "node": node,
        "control_plane_base_url": control_plane_value,
        "root_dir": resolved_root_dir,
        "readiness": {
            "is_managed": bool(node.get("is_managed", False)),
            "is_enabled": bool(node.get("is_enabled", False)),
            "has_ssh_access": bool(node.get("has_ssh_access", False)),
            "is_agent_online": bool(node.get("is_agent_online", False)),
            "is_remote_access_ready": bool(node.get("is_remote_access_ready", False)),
        },
        "ssh_profile": {
            "host": str(node.get("ssh_host") or "").strip(),
            "port": int(node.get("ssh_port", 22) or 22),
            "user": str(node.get("ssh_user") or "").strip(),
            "auth_mode": str(node.get("auth_mode") or "").strip(),
            "label": str(node.get("ssh_access_label") or "").strip(),
            "reason": str(node.get("ssh_access_reason") or "").strip(),
            "ssh_command_hint": (
                f"ssh -p {int(node.get('ssh_port', 22) or 22)} "
                f"{str(node.get('ssh_user') or '').strip()}@{str(node.get('ssh_host') or '').strip()}"
                if str(node.get("ssh_host") or "").strip() and str(node.get("ssh_user") or "").strip()
                else ""
            ),
        },
        "cluster": {
            "status": str(node.get("cluster_status") or "").strip(),
            "current_load": int(node.get("cluster_current_load", 0) or 0),
            "last_heartbeat_at": str(node.get("cluster_last_heartbeat_at") or "").strip(),
            "hostname": str(node.get("cluster_hostname") or "").strip(),
            "ip": str(node.get("cluster_ip") or "").strip(),
            "is_effective_worker": bool(node.get("cluster_is_effective_worker", False)),
        },
        "participation": {
            "state": str(node.get("participation_state") or "").strip(),
            "label": str(node.get("participation_label") or "").strip(),
            "reason": str(node.get("participation_reason") or "").strip(),
            "is_current_participant": bool(node.get("is_current_participant", False)),
            "is_dispatch_active": bool(node.get("is_dispatch_active", False)),
            "items_claimed": int(node.get("items_claimed", 0) or 0),
            "items_running": int(node.get("items_running", 0) or 0),
            "items_completed": int(node.get("items_completed", 0) or 0),
            "processed_recent": int(node.get("processed_recent", 0) or 0),
            "processed_per_minute": float(node.get("processed_per_minute", 0) or 0),
        },
        "agent_token": latest_token,
        "latest_job": latest_job,
        "delivery_queue": {
            "state": str(node.get("delivery_queue_state") or "").strip(),
            "label": str(node.get("delivery_queue_label") or "").strip(),
            "reason": str(node.get("delivery_queue_reason") or "").strip(),
            "pending_count": int(node.get("delivery_queue_pending_count", 0) or 0),
            "dead_letter_count": int(node.get("delivery_queue_dead_letter_count", 0) or 0),
            "last_flush_at": str(node.get("delivery_queue_last_flush_at") or "").strip(),
        },
        "blocking_items": blocking_items,
        "attention_items": attention_items,
        "bootstrap_defaults": {
            "node_code": normalized_node_code,
            "node_region": str(node.get("region") or "mainland").strip() or "mainland",
            "node_role": str(node.get("role") or "worker").strip() or "worker",
            "control_plane_base_url": control_plane_value,
            "root_dir": resolved_root_dir,
            "expires_in_hours": 72,
            "issued_by": "api",
            "metadata": {
                "issued_from": "ops-managed-node-handover",
                "node_region": str(node.get("region") or "mainland").strip() or "mainland",
                "node_role": str(node.get("role") or "worker").strip() or "worker",
                "remote_access_state": str(node.get("remote_access_state") or "").strip(),
                "agent_state": agent_state,
            },
        },
        "endpoints": {
            "detail": f"/api/v1/ops/nodes/{normalized_node_code}/handover",
            "bootstrap_plan": f"/api/v1/ops/nodes/{normalized_node_code}/handover/bootstrap-plan",
        },
    }


def get_managed_node_handover(
    node_code: str,
    *,
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> dict:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return {}

    resolved_nodes_payload = dict(nodes_payload or {})
    if not resolved_nodes_payload:
        resolved_nodes_payload = list_managed_nodes_with_agent_state(participation_payload=participation_payload)
    nodes = list(resolved_nodes_payload.get("nodes") or [])
    node = next(
        (dict(item or {}) for item in nodes if str(item.get("node_code") or "").strip() == normalized_node_code),
        {},
    )
    if not node:
        return {}
    return _build_managed_node_handover_payload(
        node=node,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
    )


def get_managed_node_onboarding(
    node_code: str,
    *,
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> dict:
    handover = get_managed_node_handover(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not handover:
        return {}

    node = dict(handover.get("node") or {})
    stage = dict(handover.get("stage") or {})
    readiness = dict(handover.get("readiness") or {})
    normalized_node_code = str(handover.get("node_code") or node_code or "").strip()
    stage_code = str(stage.get("code") or "").strip()
    has_ssh_access = bool(readiness.get("has_ssh_access", False))
    is_agent_online = bool(readiness.get("is_agent_online", False))
    is_remote_access_ready = bool(readiness.get("is_remote_access_ready", False))

    acceptance_execution_mode = ""
    acceptance_status_code = "blocked"
    acceptance_status_label = "暂不可验收"
    acceptance_summary = "当前还没进入标准接管就绪状态，暂不建议直接跑 onboarding.acceptance。"
    if is_agent_online:
        acceptance_execution_mode = "remote-agent"
        acceptance_status_code = "ready"
        acceptance_status_label = "可远端验收"
        acceptance_summary = "Node Agent 已在线，可以直接跑 onboarding.acceptance 做标准接管验收。"
    elif has_ssh_access:
        acceptance_execution_mode = "ssh"
        acceptance_status_code = "ssh_only"
        acceptance_status_label = "可走 SSH 验收"
        acceptance_summary = "当前 Agent 还没在线，但 SSH 已备好，可先走 SSH 模式做接管验收。"

    onboarding_stage_code = stage_code or "unknown"
    onboarding_stage_label = str(stage.get("label") or "").strip() or "待判断"
    onboarding_summary = str(handover.get("summary") or stage.get("summary") or "").strip()
    if acceptance_status_code in {"ready", "ssh_only"} and stage_code == "ready":
        onboarding_stage_code = "acceptance_ready"
        onboarding_stage_label = "待接管验收"
        onboarding_summary = acceptance_summary

    acceptance_payload = {
        "playbook_key": "onboarding.acceptance",
        "target_node_codes": [normalized_node_code] if normalized_node_code else [],
        "execution_mode": acceptance_execution_mode or "remote-agent",
        "auto_approve": True,
    }
    bootstrap_payload = {
        "node_code": normalized_node_code,
        "node_region": str(node.get("region") or "mainland").strip() or "mainland",
        "node_role": str(node.get("role") or "worker").strip() or "worker",
        "control_plane_base_url": str(handover.get("control_plane_base_url") or "").strip(),
        "root_dir": str(handover.get("root_dir") or "/opt/domaincheck").strip() or "/opt/domaincheck",
    }
    next_actions = []
    if stage_code in {"pending_bootstrap", "runtime_only", "profile_incomplete", "ssh_ready", "token_issue", "stale"}:
        next_actions.append(
            {
                "key": "bootstrap_plan",
                "label": "生成接入方案",
                "summary": "先生成并执行 Node Agent bootstrap 方案，再观察 register / heartbeat 是否建立。",
                "command_hint": (
                    build_bash_command("drive_ops_center.sh", "node-bootstrap-plan", "http://127.0.0.1:8100", normalized_node_code)
                    if normalized_node_code
                    else ""
                ),
            }
        )
        next_actions.append(
            {
                "key": "bootstrap_run",
                "label": "签发接入工单",
                "summary": "把本节点 bootstrap 收编进 ops playbook / ops job，形成标准可追踪的纳管回执。",
                "command_hint": (
                    build_bash_command("drive_ops_center.sh", "node-bootstrap-run", "http://127.0.0.1:8100", normalized_node_code)
                    if normalized_node_code
                    else ""
                ),
            }
        )
    if acceptance_status_code in {"ready", "ssh_only"}:
        next_actions.append(
            {
                "key": "run_acceptance",
                "label": "执行接管验收",
                "summary": acceptance_summary,
                "command_hint": (
                    "通过 driver / runbook 执行 onboarding.acceptance，确认 health / node-agent / worker 三段验收全部通过。"
                ),
            }
        )

    recovery_action = "noop"
    recovery_label = "当前无需额外恢复动作"
    recovery_summary = onboarding_summary or "当前节点暂无需要执行的接管恢复动作。"
    recovery_command_hint = ""
    recovery_needs_confirmation = True
    recovery_window = "none"
    if onboarding_stage_code == "acceptance_ready" or is_agent_online:
        recovery_action = "run_acceptance"
        recovery_label = "执行接管验收"
        recovery_summary = acceptance_summary or "当前节点已经进入验收窗口，建议直接执行 onboarding.acceptance。"
        recovery_command_hint = (
            build_bash_command("drive_ops_center.sh", "node-acceptance-run", "http://127.0.0.1:8100", normalized_node_code, "cli")
            if normalized_node_code
            else ""
        )
        recovery_window = "acceptance"
    elif onboarding_stage_code in {
        "pending_bootstrap",
        "runtime_only",
        "profile_incomplete",
        "ssh_ready",
        "token_issue",
        "stale",
        "agent_pending",
        "unknown",
        "",
    }:
        recovery_action = "bootstrap_run"
        recovery_label = "签发接入工单"
        recovery_summary = "当前节点仍处于接入阶段，建议先签发 onboarding.bootstrap 并等待 register / heartbeat 建立。"
        recovery_command_hint = (
            build_bash_command("drive_ops_center.sh", "node-bootstrap-run", "http://127.0.0.1:8100", normalized_node_code, "cli")
            if normalized_node_code
            else ""
        )
        recovery_window = "bootstrap"

    return {
        "node_code": normalized_node_code,
        "summary": onboarding_summary,
        "onboarding_stage": {
            "code": onboarding_stage_code,
            "label": onboarding_stage_label,
            "summary": onboarding_summary,
        },
        "handover": handover,
        "acceptance": {
            "status_code": acceptance_status_code,
            "status_label": acceptance_status_label,
            "summary": acceptance_summary,
            "execution_mode": acceptance_execution_mode,
            "is_agent_online": is_agent_online,
            "has_ssh_access": has_ssh_access,
            "is_remote_access_ready": is_remote_access_ready,
            "playbook": acceptance_payload,
        },
        "recommended_actions": next_actions,
        "recovery_decision": {
            "action": recovery_action,
            "label": recovery_label,
            "summary": recovery_summary,
            "command_hint": recovery_command_hint,
            "needs_confirmation": recovery_needs_confirmation,
            "window": recovery_window,
            "stage_code": onboarding_stage_code,
            "acceptance_status_code": acceptance_status_code,
        },
        "bootstrap_plan_request": bootstrap_payload,
    }


def preview_managed_node_onboarding_bootstrap(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    normalized_node_code = str(onboarding.get("node_code") or node_code or "").strip()
    bootstrap_plan_request = dict(onboarding.get("bootstrap_plan_request") or {})
    normalized_requested_by = str(requested_by or "api").strip() or "api"

    from app.services.ops_playbook_service import preview_ops_playbook

    ok, message, data = preview_ops_playbook(
        {
            "playbook_key": "onboarding.bootstrap",
            "node_codes": [normalized_node_code] if normalized_node_code else [],
            "execution_mode": "control-plane",
            "auto_approve": True,
            "requested_by": normalized_requested_by,
            "payload": bootstrap_plan_request,
        }
    )
    follow_up = {
        "next_stage_code": "acceptance_ready",
        "summary": (
            str((((data or {}).get("playbook_run") or {}).get("focus_summary") or "")).strip()
            or "接入工单预览已生成；正式执行后应继续观察 Node Agent register / heartbeat，并确认 onboarding_stage 是否进入 acceptance_ready。"
        ),
        "recommended_commands": [
            build_bash_command("drive_ops_center.sh", "node-recover", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "node-onboarding", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "node-handover", "http://127.0.0.1:8100", normalized_node_code),
        ] if normalized_node_code else [],
        "focus_ref": dict((((data or {}).get("playbook_preview") or {}).get("focus_ref") or {})),
    }
    return ok, message, {
        "node_code": normalized_node_code,
        "requested_by": normalized_requested_by,
        "execution_mode": "control-plane",
        "bootstrap_plan_request": bootstrap_plan_request,
        "onboarding": onboarding,
        "playbook_preview": data,
        "follow_up": follow_up,
    }


def execute_managed_node_onboarding_bootstrap(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    normalized_node_code = str(onboarding.get("node_code") or node_code or "").strip()
    bootstrap_plan_request = dict(onboarding.get("bootstrap_plan_request") or {})
    normalized_requested_by = str(requested_by or "api").strip() or "api"

    from app.services.ops_playbook_service import execute_ops_playbook

    ok, message, data = execute_ops_playbook(
        {
            "playbook_key": "onboarding.bootstrap",
            "node_codes": [normalized_node_code] if normalized_node_code else [],
            "execution_mode": "control-plane",
            "auto_approve": True,
            "requested_by": normalized_requested_by,
            "payload": bootstrap_plan_request,
        }
    )
    follow_up = {
        "next_stage_code": "acceptance_ready",
        "summary": (
            str((((data or {}).get("playbook_run") or {}).get("focus_summary") or "")).strip()
            or "接入工单已签发；接下来应继续观察 Node Agent register / heartbeat，并确认 onboarding_stage 是否进入 acceptance_ready。"
        ),
        "recommended_commands": [
            build_bash_command("drive_ops_center.sh", "node-recover", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "node-onboarding", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "node-handover", "http://127.0.0.1:8100", normalized_node_code),
        ] if normalized_node_code else [],
        "focus_ref": dict((((data or {}).get("playbook_run") or {}).get("focus_ref") or {})),
    }
    return ok, message, {
        "node_code": normalized_node_code,
        "requested_by": normalized_requested_by,
        "execution_mode": "control-plane",
        "bootstrap_plan_request": bootstrap_plan_request,
        "onboarding": onboarding,
        "playbook_run": data,
        "follow_up": follow_up,
    }


def preview_managed_node_onboarding_acceptance(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    acceptance = dict(onboarding.get("acceptance") or {})
    acceptance_status_code = str(acceptance.get("status_code") or "").strip()
    if acceptance_status_code not in {"ready", "ssh_only"}:
        return False, (
            str(acceptance.get("summary") or "当前节点暂不可执行接管验收").strip() or "当前节点暂不可执行接管验收"
        ), {"onboarding": onboarding}

    from app.services.ops_playbook_service import preview_ops_playbook

    normalized_node_code = str(onboarding.get("node_code") or node_code or "").strip()
    normalized_requested_by = str(requested_by or "api").strip() or "api"
    execution_mode = str(acceptance.get("execution_mode") or "remote-agent").strip() or "remote-agent"
    ok, message, data = preview_ops_playbook(
        {
            "playbook_key": "onboarding.acceptance",
            "node_codes": [normalized_node_code] if normalized_node_code else [],
            "execution_mode": execution_mode,
            "auto_approve": True,
            "requested_by": normalized_requested_by,
        }
    )
    follow_up = {
        "next_stage_code": "ready",
        "summary": (
            str((((data or {}).get("playbook_run") or {}).get("focus_summary") or "")).strip()
            or "接管验收预览已生成；正式执行后应继续确认 health / node-agent / worker 三段验收全部通过。"
        ),
        "recommended_commands": [
            build_bash_command("drive_ops_center.sh", "node-onboarding", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "doctor", "http://127.0.0.1:8100"),
        ] if normalized_node_code else [],
        "focus_ref": dict((((data or {}).get("playbook_preview") or {}).get("focus_ref") or {})),
    }
    return ok, message, {
        "node_code": normalized_node_code,
        "requested_by": normalized_requested_by,
        "execution_mode": execution_mode,
        "onboarding": onboarding,
        "acceptance": acceptance,
        "playbook_preview": data,
        "follow_up": follow_up,
    }


def execute_managed_node_onboarding_acceptance(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    acceptance = dict(onboarding.get("acceptance") or {})
    acceptance_status_code = str(acceptance.get("status_code") or "").strip()
    if acceptance_status_code not in {"ready", "ssh_only"}:
        return False, (
            str(acceptance.get("summary") or "当前节点暂不可执行接管验收").strip() or "当前节点暂不可执行接管验收"
        ), {"onboarding": onboarding}

    from app.services.ops_playbook_service import execute_ops_playbook

    normalized_node_code = str(onboarding.get("node_code") or node_code or "").strip()
    normalized_requested_by = str(requested_by or "api").strip() or "api"
    execution_mode = str(acceptance.get("execution_mode") or "remote-agent").strip() or "remote-agent"
    ok, message, data = execute_ops_playbook(
        {
            "playbook_key": "onboarding.acceptance",
            "node_codes": [normalized_node_code] if normalized_node_code else [],
            "execution_mode": execution_mode,
            "auto_approve": True,
            "requested_by": normalized_requested_by,
        }
    )
    follow_up = {
        "next_stage_code": "ready",
        "summary": (
            str((((data or {}).get("playbook_run") or {}).get("focus_summary") or "")).strip()
            or "接管验收已发起；接下来应继续确认 health / node-agent / worker 三段验收全部通过。"
        ),
        "recommended_commands": [
            build_bash_command("drive_ops_center.sh", "node-onboarding", "http://127.0.0.1:8100", normalized_node_code),
            build_bash_command("drive_ops_center.sh", "doctor", "http://127.0.0.1:8100"),
        ] if normalized_node_code else [],
        "focus_ref": dict((((data or {}).get("playbook_run") or {}).get("focus_ref") or {})),
    }
    return ok, message, {
        "node_code": normalized_node_code,
        "requested_by": normalized_requested_by,
        "execution_mode": execution_mode,
        "onboarding": onboarding,
        "acceptance": acceptance,
        "playbook_run": data,
        "follow_up": follow_up,
    }


def preview_managed_node_onboarding_recovery(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    recovery = dict(onboarding.get("recovery_decision") or {})
    recovery_action = str(recovery.get("action") or "").strip()
    if recovery_action == "bootstrap_run":
        ok, message, data = preview_managed_node_onboarding_bootstrap(
            node_code=str(onboarding.get("node_code") or node_code or "").strip(),
            requested_by=requested_by,
            control_plane_base_url=control_plane_base_url,
            root_dir=root_dir,
            participation_payload=participation_payload,
            nodes_payload=nodes_payload,
        )
        return ok, message, {
            "recovery_action": recovery_action,
            "recovery_decision": recovery,
            "onboarding": onboarding,
            "selected_preview": data,
        }
    if recovery_action == "run_acceptance":
        ok, message, data = preview_managed_node_onboarding_acceptance(
            node_code=str(onboarding.get("node_code") or node_code or "").strip(),
            requested_by=requested_by,
            control_plane_base_url=control_plane_base_url,
            root_dir=root_dir,
            participation_payload=participation_payload,
            nodes_payload=nodes_payload,
        )
        return ok, message, {
            "recovery_action": recovery_action,
            "recovery_decision": recovery,
            "onboarding": onboarding,
            "selected_preview": data,
        }
    return False, (
        str(recovery.get("summary") or "当前节点暂无可执行恢复动作").strip() or "当前节点暂无可执行恢复动作"
    ), {
        "recovery_action": recovery_action,
        "recovery_decision": recovery,
        "onboarding": onboarding,
    }


def execute_managed_node_onboarding_recovery(
    *,
    node_code: str,
    requested_by: str = "api",
    control_plane_base_url: str = "",
    root_dir: str = "",
    participation_payload: dict | None = None,
    nodes_payload: dict | None = None,
) -> tuple[bool, str, dict]:
    onboarding = get_managed_node_onboarding(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
        participation_payload=participation_payload,
        nodes_payload=nodes_payload,
    )
    if not onboarding:
        return False, "节点不存在", {}

    recovery = dict(onboarding.get("recovery_decision") or {})
    recovery_action = str(recovery.get("action") or "").strip()
    if recovery_action == "bootstrap_run":
        ok, message, data = execute_managed_node_onboarding_bootstrap(
            node_code=str(onboarding.get("node_code") or node_code or "").strip(),
            requested_by=requested_by,
            control_plane_base_url=control_plane_base_url,
            root_dir=root_dir,
            participation_payload=participation_payload,
            nodes_payload=nodes_payload,
        )
        return ok, message, {
            "recovery_action": recovery_action,
            "recovery_decision": recovery,
            "onboarding": onboarding,
            "selected_run": data,
        }
    if recovery_action == "run_acceptance":
        ok, message, data = execute_managed_node_onboarding_acceptance(
            node_code=str(onboarding.get("node_code") or node_code or "").strip(),
            requested_by=requested_by,
            control_plane_base_url=control_plane_base_url,
            root_dir=root_dir,
            participation_payload=participation_payload,
            nodes_payload=nodes_payload,
        )
        return ok, message, {
            "recovery_action": recovery_action,
            "recovery_decision": recovery,
            "onboarding": onboarding,
            "selected_run": data,
        }
    return False, (
        str(recovery.get("summary") or "当前节点暂无可执行恢复动作").strip() or "当前节点暂无可执行恢复动作"
    ), {
        "recovery_action": recovery_action,
        "recovery_decision": recovery,
        "onboarding": onboarding,
    }


def build_managed_node_handover_bootstrap_plan(
    *,
    node_code: str,
    issued_by: str = "api",
    expires_in_hours: int = 72,
    control_plane_base_url: str = "",
    root_dir: str = "",
    metadata: dict | None = None,
) -> tuple[bool, str, dict]:
    handover = get_managed_node_handover(
        node_code,
        control_plane_base_url=control_plane_base_url,
        root_dir=root_dir,
    )
    if not handover:
        return False, "节点不存在", {}

    node = dict(handover.get("node") or {})
    bootstrap_defaults = dict(handover.get("bootstrap_defaults") or {})
    normalized_node_code = str(node.get("node_code") or node_code or "").strip()
    ok, message, data = build_node_agent_bootstrap_plan(
        node_code=normalized_node_code,
        node_region=str(
            bootstrap_defaults.get("node_region") or node.get("region") or "mainland"
        ).strip()
        or "mainland",
        node_role=str(bootstrap_defaults.get("node_role") or node.get("role") or "worker").strip() or "worker",
        issued_by=str(issued_by or "api").strip() or "api",
        expires_in_hours=int(expires_in_hours or bootstrap_defaults.get("expires_in_hours") or 72),
        control_plane_base_url=str(
            control_plane_base_url or bootstrap_defaults.get("control_plane_base_url") or handover.get("control_plane_base_url") or ""
        ).strip(),
        root_dir=str(root_dir or bootstrap_defaults.get("root_dir") or handover.get("root_dir") or "/opt/domaincheck").strip()
        or "/opt/domaincheck",
        metadata={
            **dict(bootstrap_defaults.get("metadata") or {}),
            **dict(metadata or {}),
            "issued_from": str((metadata or {}).get("issued_from") or "ops-managed-node-handover").strip()
            or "ops-managed-node-handover",
        },
    )
    if not ok:
        return False, message, data

    refreshed_handover = get_managed_node_handover(
        normalized_node_code,
        control_plane_base_url=str(handover.get("control_plane_base_url") or "").strip(),
        root_dir=str(handover.get("root_dir") or "").strip(),
    )
    return True, message, {
        **data,
        "handover": refreshed_handover,
        "requested_via": "managed-node-handover",
    }


def _agent_job_type(job: dict) -> str:
    normalized_job = dict(job or {})
    metadata = dict(normalized_job.get("metadata") or {})
    payload = dict(normalized_job.get("payload") or {})
    explicit_job_type = str(metadata.get("job_type") or payload.get("job_type") or "").strip()
    if explicit_job_type:
        return explicit_job_type
    action = str(normalized_job.get("action") or "").strip()
    rollout_id = int(normalized_job.get("rollout_id") or 0)
    if rollout_id > 0:
        return "release_rollout"
    if action == "deploy.release":
        return "release_deploy"
    if action in {"health.snapshot", "logs.collect", "diagnostics.collect"}:
        return "inspection"
    return "ops_action"


def _agent_release_context(job: dict) -> dict:
    normalized_job = dict(job or {})
    metadata = dict(normalized_job.get("metadata") or {})
    payload = dict(normalized_job.get("payload") or {})
    return {
        "release_id": int(payload.get("release_id") or metadata.get("release_id") or 0),
        "release_version": str(payload.get("release_version") or metadata.get("release_version") or "").strip(),
        "rollout_id": int(
            normalized_job.get("rollout_id")
            or payload.get("rollout_id")
            or metadata.get("rollout_id")
            or 0
        ),
        "rollout_code": str(payload.get("rollout_code") or metadata.get("rollout_code") or "").strip(),
    }


def _agent_job_envelope(job: dict) -> dict:
    normalized_job = dict(job or {})
    steps = list(normalized_job.get("steps") or [])
    first_step = dict(steps[0] or {}) if steps else {}
    metadata = dict(normalized_job.get("metadata") or {})
    focus_ref = dict(normalized_job.get("focus_ref") or {})
    if not focus_ref:
        focus_ref = {
            "kind": "ops_job",
            "job_id": int(normalized_job.get("id") or 0),
            "job_code": str(normalized_job.get("job_code") or "").strip(),
            "action": str(normalized_job.get("action") or "").strip(),
            "target_node_code": str(normalized_job.get("target_node_code") or "").strip(),
        }
    step_key = str(first_step.get("step_key") or metadata.get("step_key") or "dispatch").strip() or "dispatch"
    step_title = (
        str(
            first_step.get("title")
            or metadata.get("step_title")
            or normalized_job.get("summary")
            or normalized_job.get("action")
            or ""
        ).strip()
        or step_key
    )
    envelope = {
        **normalized_job,
        "protocol_version": "ops-agent/v1",
        "envelope_type": "agent_job",
        "job_id": int(normalized_job.get("id") or 0),
        "job_type": _agent_job_type(normalized_job),
        "step_key": step_key,
        "step_title": step_title,
        "policy": dict(normalized_job.get("policy") or {}),
        "release_context": _agent_release_context(normalized_job),
        "focus_ref": focus_ref,
        "job_ref": {
            "job_id": int(normalized_job.get("id") or 0),
            "job_code": str(normalized_job.get("job_code") or "").strip(),
            "action": str(normalized_job.get("action") or "").strip(),
            "target_node_code": str(normalized_job.get("target_node_code") or "").strip(),
        },
        "step_ref": {
            "step_id": int(first_step.get("id") or 0),
            "step_key": step_key,
            "step_title": step_title,
        },
    }
    return envelope


def _completion_summary(job: dict, *, client_request_id: str = "", deduplicated: bool = False) -> dict:
    normalized_job = dict(job or {})
    result = dict(normalized_job.get("result") or {})
    return {
        "job_id": int(normalized_job.get("id") or 0),
        "job_code": str(normalized_job.get("job_code") or "").strip(),
        "status": str(normalized_job.get("status") or "").strip(),
        "status_label": str(normalized_job.get("status_label") or "").strip(),
        "deduplicated": bool(deduplicated),
        "client_request_id": str(client_request_id or "").strip(),
        "result_summary_text": str(result.get("summary_text") or result.get("summary") or "").strip(),
        "focus_ref": dict(normalized_job.get("focus_ref") or {}),
    }


def _slim_delivery_queue_head_record(node: dict, *, state: str) -> dict:
    normalized_state = str(state or "").strip()
    if normalized_state not in {"pending", "dead_letter"}:
        return {}

    if normalized_state == "pending":
        created_at = str(node.get("delivery_queue_oldest_pending_at") or "").strip()
        request_id = str(node.get("delivery_queue_oldest_pending_request_id") or "").strip()
        request_kind = str(node.get("delivery_queue_oldest_pending_kind") or "").strip()
    else:
        created_at = str(node.get("delivery_queue_oldest_dead_letter_at") or "").strip()
        request_id = str(node.get("delivery_queue_oldest_dead_letter_request_id") or "").strip()
        request_kind = str(node.get("delivery_queue_oldest_dead_letter_kind") or "").strip()

    if not any((created_at, request_id, request_kind)):
        return {}

    node_code = str(node.get("node_code") or "").strip()
    return {
        "record_key": f"{node_code}:{normalized_state}:{request_id or request_kind or created_at or 'head'}",
        "record_id": request_id,
        "node_code": node_code,
        "state": normalized_state,
        "label": "待重试头部记录" if normalized_state == "pending" else "死信头部记录",
        "request_kind": request_kind,
        "client_request_id": request_id,
        "created_at": created_at if normalized_state == "pending" else "",
        "dead_letter_at": created_at if normalized_state == "dead_letter" else "",
        "detail_code": "",
        "error_message": "",
        "record_source": "agent-heartbeat-snapshot",
        "record_visibility": "head_only",
        "payload_visibility": "unavailable",
        "response_visibility": "unavailable",
    }


def get_managed_node_delivery_queue(node_code: str) -> dict:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return {}

    payload = list_managed_nodes_with_agent_state()
    node = next(
        (
            dict(item or {})
            for item in list(payload.get("nodes") or [])
            if str(item.get("node_code") or "").strip() == normalized_node_code
        ),
        {},
    )
    if not node:
        return {}

    pending_head = _slim_delivery_queue_head_record(node, state="pending")
    dead_letter_head = _slim_delivery_queue_head_record(node, state="dead_letter")
    record_visibility = "head_only"
    capabilities = {
        "can_flush": True,
        "can_replay": True,
        "can_discard": True,
    }

    return {
        "node_code": normalized_node_code,
        "node": {
            "node_code": normalized_node_code,
            "title": str(node.get("title") or "").strip(),
            "region": str(node.get("region") or "").strip(),
            "role": str(node.get("role") or "").strip(),
            "is_managed": bool(node.get("is_managed", False)),
            "is_enabled": bool(node.get("is_enabled", False)),
            "is_agent_online": bool(node.get("is_agent_online", False)),
            "agent_state": str(node.get("agent_state") or "").strip(),
            "agent_state_label": str(node.get("agent_state_label") or "").strip(),
            "remote_access_state": str(node.get("remote_access_state") or "").strip(),
            "remote_access_label": str(node.get("remote_access_label") or "").strip(),
        },
        "summary": {
            "state": str(node.get("delivery_queue_state") or "").strip(),
            "label": str(node.get("delivery_queue_label") or "").strip(),
            "reason": str(node.get("delivery_queue_reason") or "").strip(),
            "pending_count": int(node.get("delivery_queue_pending_count", 0) or 0),
            "dead_letter_count": int(node.get("delivery_queue_dead_letter_count", 0) or 0),
            "last_flush_at": str(node.get("delivery_queue_last_flush_at") or "").strip(),
            "oldest_pending_at": str(node.get("delivery_queue_oldest_pending_at") or "").strip(),
            "oldest_dead_letter_at": str(node.get("delivery_queue_oldest_dead_letter_at") or "").strip(),
        },
        "record_visibility": record_visibility,
        "capabilities": capabilities,
        "records_capability": {
            "record_source": "agent-heartbeat-snapshot",
            "record_visibility": record_visibility,
            "full_records_available": False,
            "available_actions": [
                {
                    "key": "delivery.queue.flush",
                    "label": "立即冲刷",
                    "endpoint": f"/api/v1/ops/nodes/{normalized_node_code}/delivery-queue/flush",
                    "record_scope": "pending",
                    "execution_mode": "remote-agent",
                },
                {
                    "key": "delivery.queue.replay",
                    "label": "重放死信",
                    "endpoint": f"/api/v1/ops/nodes/{normalized_node_code}/delivery-queue/replay",
                    "record_scope": "dead_letter",
                    "execution_mode": "remote-agent",
                },
                {
                    "key": "delivery.queue.discard",
                    "label": "丢弃死信",
                    "endpoint": f"/api/v1/ops/nodes/{normalized_node_code}/delivery-queue/records/{{record_id}}/discard",
                    "record_scope": "dead_letter",
                    "execution_mode": "remote-agent",
                },
            ],
            "next_step": "当前控制面仍以头部记录视角展示队列；已支持通过正式 Ops Job 触发冲刷、重放和丢弃动作。",
        },
        "head_records": {
            "pending": pending_head,
            "dead_letter": dead_letter_head,
        },
    }


def list_managed_node_delivery_queue_records(
    node_code: str,
    *,
    state: str = "",
    limit: int = 50,
) -> dict:
    normalized_node_code = str(node_code or "").strip()
    normalized_state = str(state or "").strip()
    safe_limit = min(max(int(limit or 50), 1), 200)
    queue_payload = get_managed_node_delivery_queue(normalized_node_code)
    if not queue_payload:
        return {}

    candidate_records = [
        record
        for record in (
            dict((queue_payload.get("head_records") or {}).get("pending") or {}),
            dict((queue_payload.get("head_records") or {}).get("dead_letter") or {}),
        )
        if record
    ]
    if normalized_state:
        candidate_records = [
            record for record in candidate_records if str(record.get("state") or "").strip() == normalized_state
        ]
    records = candidate_records[:safe_limit]

    available_state_counts: dict[str, int] = {}
    for record in candidate_records:
        record_state = str(record.get("state") or "").strip() or "unknown"
        available_state_counts[record_state] = int(available_state_counts.get(record_state, 0) or 0) + 1

    return {
        "node_code": normalized_node_code,
        "records": records,
        "summary": {
            "total": len(records),
            "returned_records": len(records),
            "available_total": len(candidate_records),
            "available_state_counts": available_state_counts,
            "record_visibility": "head_only",
            "filters": {
                "state": normalized_state,
                "limit": safe_limit,
            },
        },
        "records_capability": dict(queue_payload.get("records_capability") or {}),
        "record_visibility": str((queue_payload.get("record_visibility") or "head_only")).strip() or "head_only",
        "capabilities": dict(queue_payload.get("capabilities") or {}),
    }


def _delivery_queue_visible_record_id(record: dict | None = None) -> str:
    normalized_record = dict(record or {})
    return (
        str(normalized_record.get("record_id") or "").strip()
        or str(normalized_record.get("client_request_id") or "").strip()
        or str(normalized_record.get("record_key") or "").strip()
    )


def _find_visible_delivery_queue_record(queue_payload: dict, record_id: str) -> dict:
    normalized_record_id = str(record_id or "").strip()
    if not normalized_record_id:
        return {}
    for state in ("pending", "dead_letter"):
        record = dict(((queue_payload.get("head_records") or {}).get(state) or {}))
        if record and _delivery_queue_visible_record_id(record) == normalized_record_id:
            return record
    return {}


def _build_delivery_queue_selector_from_template_payload(
    template_key: str,
    normalized_payload: dict,
    *,
    force_state: str = "",
) -> dict:
    selector: dict[str, object] = {}
    if str(force_state or "").strip():
        selector["state"] = str(force_state or "").strip()
    elif template_key in {"delivery.queue.replay", "delivery.queue.discard"}:
        selector["state"] = "dead_letter"

    record_id = str(normalized_payload.get("record_id") or "").strip()
    request_kind = str(normalized_payload.get("request_kind") or "").strip()
    detail_code = str(normalized_payload.get("detail_code") or "").strip()
    if record_id:
        selector["record_id"] = record_id
    if request_kind:
        selector["request_kind"] = request_kind
    if detail_code:
        selector["detail_code"] = detail_code
    return selector


def request_managed_node_delivery_queue_action(
    node_code: str,
    action_key: str,
    *,
    payload: dict | None = None,
    record_id: str = "",
) -> tuple[bool, str, dict]:
    normalized_node_code = str(node_code or "").strip()
    normalized_action_key = str(action_key or "").strip()
    raw_payload = dict(payload or {})
    if not normalized_node_code:
        return False, "node_code 不能为空", {}
    if normalized_action_key not in {"delivery.queue.flush", "delivery.queue.replay", "delivery.queue.discard"}:
        return False, "当前不支持该 delivery queue 动作", {"action_key": normalized_action_key}

    queue_payload = get_managed_node_delivery_queue(normalized_node_code)
    if not queue_payload:
        return False, "节点不存在", {}

    template = get_ops_action_template(normalized_action_key)
    if not template:
        return False, "delivery queue 动作模板不存在", {"action_key": normalized_action_key}

    effective_payload = dict(raw_payload)
    if str(record_id or "").strip():
        effective_payload["record_id"] = str(record_id or "").strip()
    normalize_ok, normalize_message, normalized_action_payload = build_ops_template_payload(
        normalized_action_key,
        effective_payload,
    )
    if not normalize_ok:
        return False, normalize_message, {"action_key": normalized_action_key}

    visible_record = {}
    if str(record_id or "").strip():
        visible_record = _find_visible_delivery_queue_record(queue_payload, str(record_id or "").strip())
        if not visible_record:
            return False, "当前控制面仅支持对可见头部记录执行单条动作", {
                "action_key": normalized_action_key,
                "node_code": normalized_node_code,
                "record_id": str(record_id or "").strip(),
                "record_visibility": str(((queue_payload.get("records_capability") or {}).get("record_visibility") or "")),
            }
        if normalized_action_key in {"delivery.queue.replay", "delivery.queue.discard"}:
            if str(visible_record.get("state") or "").strip() != "dead_letter":
                return False, "单条重放/丢弃当前仅支持死信头部记录", {
                    "action_key": normalized_action_key,
                    "node_code": normalized_node_code,
                    "record": visible_record,
                }
        normalized_action_payload["record_id"] = _delivery_queue_visible_record_id(visible_record)

    selector = _build_delivery_queue_selector_from_template_payload(
        normalized_action_key,
        normalized_action_payload,
    )
    action_payload: dict[str, object] = {}
    if normalized_action_key == "delivery.queue.flush":
        action_payload["limit"] = int(normalized_action_payload.get("limit") or 20)
    elif normalized_action_key == "delivery.queue.replay":
        action_payload["selector"] = selector
        action_payload["limit"] = int(normalized_action_payload.get("limit") or 20)
        action_payload["flush_after_replay"] = bool(normalized_action_payload.get("flush_after_replay", True))
        if str(normalized_action_payload.get("reason") or "").strip():
            action_payload["reason"] = str(normalized_action_payload.get("reason") or "").strip()
    elif normalized_action_key == "delivery.queue.discard":
        action_payload["selector"] = selector
        action_payload["limit"] = int(normalized_action_payload.get("limit") or 20)
        action_payload["discarded_by"] = (
            str(raw_payload.get("discarded_by") or raw_payload.get("requested_by") or "web-ui").strip() or "web-ui"
        )
        action_payload["reason"] = str(normalized_action_payload.get("reason") or "").strip()

    requested_by = str(raw_payload.get("requested_by") or "web-ui").strip() or "web-ui"
    ok, message, data = create_ops_job(
        {
            "action": normalized_action_key,
            "target_type": "node",
            "target_node_code": normalized_node_code,
            "requested_by": requested_by,
            "execution_mode": "remote-agent",
            "auto_approve": bool(template.get("default_auto_approve", False)),
            "payload": action_payload,
            "metadata": {
                "delivery_queue_action": {
                    "action_key": normalized_action_key,
                    "template_key": normalized_action_key,
                    "selector": selector,
                    "record_visibility": str(((queue_payload.get("records_capability") or {}).get("record_visibility") or "")),
                    "requested_from": "ops-delivery-queue",
                    "visible_record": visible_record,
                    "queue_summary": dict(queue_payload.get("summary") or {}),
                }
            },
        }
    )
    if not ok:
        return False, message, data

    action_label = str(template.get("title") or normalized_action_key).strip()
    return True, f"{action_label}任务已创建", {
        "node_code": normalized_node_code,
        "action_key": normalized_action_key,
        "action_label": action_label,
        "queue": {
            "summary": dict(queue_payload.get("summary") or {}),
            "records_capability": dict(queue_payload.get("records_capability") or {}),
        },
        "selector": selector,
        "visible_record": visible_record,
        "job": dict(data.get("job") or {}),
        "executed_immediately": bool(data.get("executed_immediately", False)),
    }


def issue_node_agent_token(
    *,
    node_code: str,
    issued_by: str = "api",
    expires_in_hours: int = 72,
    metadata: dict | None = None,
) -> tuple[bool, str, dict]:
    ensure_ops_agent_schema()
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return False, "node_code 不能为空", {}

    token = secrets.token_urlsafe(32)
    token_hash = _hash_token(token)
    expires_at = datetime.now() + timedelta(hours=max(1, min(int(expires_in_hours or 72), 24 * 30)))
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                INSERT INTO ops_node_tokens (
                    token_hash, node_code, purpose, issued_by, is_enabled, expires_at, metadata_json, created_at, updated_at
                ) VALUES (%s, %s, 'agent', %s, TRUE, %s, %s, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)
                RETURNING id, expires_at, created_at
                """,
                (
                    token_hash,
                    normalized_node_code,
                    str(issued_by or "api").strip() or "api",
                    expires_at,
                    json.dumps(metadata or {}, ensure_ascii=False),
                ),
            )
            row = cur.fetchone()
        conn.commit()

    return True, "节点 Agent 令牌已签发", {
        "token": token,
        "token_preview": f"{token[:6]}...{token[-4:]}",
        "record_id": int(row[0]),
        "node_code": normalized_node_code,
        "expires_at": _format_time(row[1]),
        "created_at": _format_time(row[2]),
    }


def _normalize_control_plane_base_url(raw_value: str) -> str:
    normalized = str(raw_value or "").strip().rstrip("/")
    if normalized.endswith("/api/v1"):
        normalized = normalized[: -len("/api/v1")]
    return normalized


def _candidate_node_agent_install_paths(root_dir: str) -> list[str]:
    normalized_root_dir = str(root_dir or "/opt/domaincheck").strip().rstrip("/") or "/opt/domaincheck"
    return [
        f"{normalized_root_dir}/domainCheck/deploy/multi-region/bootstrap_node_agent.sh",
        f"{normalized_root_dir}/domain-api/deploy/multi-region/bootstrap_node_agent.sh",
        f"{normalized_root_dir}/deploy/multi-region/bootstrap_node_agent.sh",
    ]


def _build_node_agent_install_block(root_dir: str) -> tuple[str, str]:
    candidates = _candidate_node_agent_install_paths(root_dir)
    preferred_path = candidates[0]
    lines = [
        f"ROOT_DIR={shlex.quote(root_dir)}",
        "INSTALL_SCRIPT=",
        "for candidate in \\",
    ]
    for index, candidate in enumerate(candidates):
        suffix = " \\" if index < len(candidates) - 1 else ""
        lines.append(f"  {shlex.quote(candidate)}{suffix}")
    lines.extend(
        [
            "do",
            "  if [[ -f \"${candidate}\" ]]; then",
            "    INSTALL_SCRIPT=\"${candidate}\"",
            "    break",
            "  fi",
            "done",
            "",
            "if [[ -z \"${INSTALL_SCRIPT}\" ]]; then",
            "  echo \"bootstrap_node_agent.sh not found under known layouts\" >&2",
            "  exit 1",
            "fi",
            "",
            "bash \"${INSTALL_SCRIPT}\" \"${ROOT_DIR}\"",
        ]
    )
    return preferred_path, "\n".join(lines)


def _build_node_agent_bootstrap_script(
    *,
    node_code: str,
    root_dir: str,
    install_script_path: str,
    install_command_block: str,
    env_content: str,
    health_checks: list[str],
) -> tuple[str, str, str]:
    script_name = f"bootstrap-node-agent-{node_code}.sh"
    script_path = f"/tmp/{script_name}"
    script_lines = [
        "#!/usr/bin/env bash",
        "set -euo pipefail",
        "",
        f"ROOT_DIR={shlex.quote(root_dir)}",
        f"INSTALL_SCRIPT={shlex.quote(install_script_path)}",
        "ENV_FILE=/etc/default/domaincheck-node-agent",
        "SERVICE_NAME=domaincheck-node-agent",
        "",
        "# Resolve bootstrap_node_agent.sh across supported repo layouts.",
        *install_command_block.splitlines(),
        "",
        "cat >\"${ENV_FILE}\" <<'EOF'",
        env_content,
        "EOF",
        "",
        "systemctl restart \"${SERVICE_NAME}\"",
        "systemctl status \"${SERVICE_NAME}\" --no-pager -l",
        "journalctl -u \"${SERVICE_NAME}\" -n 60 --no-pager",
    ]
    if health_checks:
        script_lines.extend(
            [
                "",
                "echo",
                "echo '[health-checks]'",
            ]
        )
        for item in health_checks:
            normalized_item = str(item or "").strip()
            if normalized_item:
                script_lines.append(normalized_item)
    script_content = "\n".join(script_lines).strip() + "\n"
    write_script_block = "\n".join(
        [
            f"cat >{script_path} <<'EOF'",
            script_content.rstrip("\n"),
            "EOF",
            f"chmod +x {script_path}",
        ]
    )
    run_script_block = "\n".join(
        [
            write_script_block,
            f"bash {script_path}",
        ]
    )
    return script_name, script_content, run_script_block


def build_node_agent_bootstrap_plan(
    *,
    node_code: str,
    node_region: str = "mainland",
    node_role: str = "worker",
    issued_by: str = "api",
    expires_in_hours: int = 72,
    control_plane_base_url: str = "",
    root_dir: str = "/opt/domaincheck",
    metadata: dict | None = None,
) -> tuple[bool, str, dict]:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return False, "node_code 不能为空", {}

    normalized_region = str(node_region or "mainland").strip() or "mainland"
    normalized_role = str(node_role or "worker").strip() or "worker"
    normalized_root_dir = str(root_dir or "/opt/domaincheck").strip() or "/opt/domaincheck"
    normalized_control_plane_base_url = _normalize_control_plane_base_url(control_plane_base_url)
    if not normalized_control_plane_base_url:
        normalized_control_plane_base_url = f"http://127.0.0.1:{settings.api_port}"

    token_ok, token_message, token_data = issue_node_agent_token(
        node_code=normalized_node_code,
        issued_by=issued_by,
        expires_in_hours=expires_in_hours,
        metadata=metadata or {},
    )
    if not token_ok:
        return False, token_message, {}

    env_lines = [
        f"OPS_CONTROL_PLANE_BASE_URL={normalized_control_plane_base_url}",
        f"OPS_AGENT_TOKEN={token_data.get('token') or ''}",
        "OPS_AGENT_POLL_INTERVAL_SECONDS=5",
        "",
        f"NODE_CODE={normalized_node_code}",
        f"NODE_REGION={normalized_region}",
        f"NODE_ROLE={normalized_role}",
        "",
        f"WORKER_SERVICE_NAME={settings.worker_service_name}",
        f"API_SERVICE_NAME={settings.api_service_name}",
        f"SYNC_AGENT_SERVICE_NAME={settings.sync_agent_service_name}",
        "NODE_AGENT_SERVICE_NAME=domaincheck-node-agent",
        "",
        'OPS_AGENT_CAPABILITIES=["service.start","service.stop","service.restart","service.status","runtime.start_worker","runtime.stop_worker","runtime.start_detection","runtime.stop_detection","runtime.pull_tasks","runtime.restart_api","runtime.start_sync_agent","runtime.stop_sync_agent","runtime.reset_lab_state","health.snapshot","logs.collect","diagnostics.collect","deploy.release"]',
        "OPS_AGENT_LABELS={}",
    ]
    env_content = "\n".join(env_lines)
    install_script_path, install_command_block = _build_node_agent_install_block(normalized_root_dir)
    command_lines = [
        install_command_block,
        "cat >/etc/default/domaincheck-node-agent <<'EOF'",
        env_content,
        "EOF",
        "systemctl restart domaincheck-node-agent",
        "systemctl status domaincheck-node-agent --no-pager -l",
        "journalctl -u domaincheck-node-agent -n 60 --no-pager",
    ]
    command_block = "\n".join(command_lines)
    health_checks = [
        "systemctl status domaincheck-node-agent --no-pager -l",
        "journalctl -u domaincheck-node-agent -n 60 --no-pager",
        f"curl -s {normalized_control_plane_base_url}/api/v1/ops/overview",
    ]
    bootstrap_script_name, bootstrap_script_content, bootstrap_run_script_block = _build_node_agent_bootstrap_script(
        node_code=normalized_node_code,
        root_dir=normalized_root_dir,
        install_script_path=install_script_path,
        install_command_block=install_command_block,
        env_content=env_content,
        health_checks=health_checks,
    )

    return True, "节点 Agent 接入方案已生成", {
        **token_data,
        "control_plane_base_url": normalized_control_plane_base_url,
        "bootstrap_plan": {
            "node_code": normalized_node_code,
            "node_region": normalized_region,
            "node_role": normalized_role,
            "root_dir": normalized_root_dir,
            "env_file": "/etc/default/domaincheck-node-agent",
            "service_name": "domaincheck-node-agent",
            "service_file": "/etc/systemd/system/domaincheck-node-agent.service",
            "install_script_path": install_script_path,
            "install_script_candidates": _candidate_node_agent_install_paths(normalized_root_dir),
            "install_command_block": install_command_block,
            "env_content": env_content,
            "command_lines": command_lines,
            "command_block": command_block,
            "health_checks": health_checks,
            "health_check_block": "\n".join(health_checks),
            "bootstrap_script_name": bootstrap_script_name,
            "bootstrap_script_path": f"/tmp/{bootstrap_script_name}",
            "bootstrap_script_content": bootstrap_script_content,
            "bootstrap_run_script_block": bootstrap_run_script_block,
        },
    }


def _authenticate_agent_token(token: str, *, expected_node_code: str | None = None) -> tuple[bool, str, dict]:
    ensure_ops_agent_schema()
    token_hash = _hash_token(token)
    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT id, node_code, purpose, is_enabled, expires_at, metadata_json
                FROM ops_node_tokens
                WHERE token_hash = %s
                LIMIT 1
                """,
                (token_hash,),
            )
            row = cur.fetchone()
            if not row:
                return _agent_error("agent token 无效", "agent_token_invalid")
            record_id = int(row[0])
            node_code = str(row[1] or "")
            purpose = str(row[2] or "")
            is_enabled = bool(row[3])
            expires_at = row[4]
            metadata = _decode_json(row[5])
            if purpose != "agent" or not is_enabled:
                return _agent_error("agent token 不可用", "agent_token_unavailable")
            if isinstance(expires_at, datetime):
                current_time = datetime.now(expires_at.tzinfo) if expires_at.tzinfo else datetime.now()
                if expires_at < current_time:
                    return _agent_error("agent token 已过期", "agent_token_expired")
            normalized_expected = str(expected_node_code or "").strip()
            if normalized_expected and normalized_expected != node_code:
                return _agent_error(
                    "agent token 与节点不匹配",
                    "agent_token_node_mismatch",
                    {"expected_node_code": normalized_expected, "token_node_code": node_code},
                )
            cur.execute(
                """
                UPDATE ops_node_tokens
                SET last_used_at = CURRENT_TIMESTAMP, updated_at = CURRENT_TIMESTAMP
                WHERE id = %s
                """,
                (record_id,),
            )
        conn.commit()
    return True, "ok", {
        "record_id": record_id,
        "node_code": node_code,
        "metadata": metadata,
        "expires_at": _format_time(expires_at),
    }


def _upsert_agent_runtime(node_code: str, payload: dict) -> None:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return

    region = str(payload.get("region") or "unknown").strip() or "unknown"
    role = str(payload.get("role") or "worker").strip() or "worker"
    title = str(payload.get("title") or normalized_node_code).strip() or normalized_node_code
    metadata = payload.get("metadata") or {}
    ssh_host = str(payload.get("ssh_host") or "").strip()
    ssh_user = str(payload.get("ssh_user") or "").strip()
    ssh_port = max(1, int(payload.get("ssh_port") or 22))
    auth_mode = str(payload.get("auth_mode") or "key").strip() or "key"
    deploy_channel = str(payload.get("deploy_channel") or "stable").strip() or "stable"

    merged_metadata = {
        **metadata,
        "agent_version": str(payload.get("agent_version") or metadata.get("agent_version") or "").strip(),
        "last_seen_at": datetime.now().isoformat(timespec="seconds"),
        "capabilities": payload.get("capabilities") or metadata.get("capabilities") or [],
        "labels": payload.get("labels") or metadata.get("labels") or {},
        "hostname": str(payload.get("hostname") or metadata.get("hostname") or "").strip(),
        "ip": str(payload.get("ip") or metadata.get("ip") or "").strip(),
    }

    with get_db() as conn:
        with conn.cursor() as cur:
            cur.execute(
                """
                INSERT INTO ops_managed_nodes (
                    node_code, region, role, title, ssh_host, ssh_port, ssh_user, auth_mode, deploy_channel, is_enabled, metadata_json, created_at, last_seen_at, updated_at
                ) VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, TRUE, %s, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)
                ON CONFLICT (node_code) DO UPDATE SET
                    region = EXCLUDED.region,
                    role = EXCLUDED.role,
                    title = EXCLUDED.title,
                    ssh_host = CASE WHEN EXCLUDED.ssh_host <> '' THEN EXCLUDED.ssh_host ELSE ops_managed_nodes.ssh_host END,
                    ssh_port = EXCLUDED.ssh_port,
                    ssh_user = CASE WHEN EXCLUDED.ssh_user <> '' THEN EXCLUDED.ssh_user ELSE ops_managed_nodes.ssh_user END,
                    auth_mode = EXCLUDED.auth_mode,
                    deploy_channel = EXCLUDED.deploy_channel,
                    metadata_json = EXCLUDED.metadata_json,
                    last_seen_at = CURRENT_TIMESTAMP,
                    updated_at = CURRENT_TIMESTAMP
                """,
                (
                    normalized_node_code,
                    region,
                    role,
                    title,
                    ssh_host,
                    ssh_port,
                    ssh_user,
                    auth_mode,
                    deploy_channel,
                    json.dumps(merged_metadata, ensure_ascii=False),
                ),
            )
        conn.commit()


def _load_existing_detect_node_runtime(node_code: str) -> dict:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return {}

    try:
        with get_db() as conn:
            with conn.cursor() as cur:
                cur.execute(
                    """
                    SELECT status, current_load, metadata_json, last_heartbeat_at
                    FROM detect_worker_nodes
                    WHERE node_code = %s
                    LIMIT 1
                    """,
                    (normalized_node_code,),
                )
                row = cur.fetchone()
    except Exception:
        return {}

    if not row:
        return {}
    metadata = row[2] if isinstance(row[2], dict) else {}
    return {
        "status": str(row[0] or "").strip(),
        "current_load": int(row[1] or 0),
        "metadata": metadata,
        "last_heartbeat_at": row[3],
    }


def _upsert_agent_detect_runtime(node_code: str, payload: dict) -> None:
    normalized_node_code = str(node_code or "").strip()
    if not normalized_node_code:
        return

    metadata = payload.get("metadata") if isinstance(payload.get("metadata"), dict) else {}
    detect_runtime = metadata.get("detect_runtime") if isinstance(metadata.get("detect_runtime"), dict) else {}
    if not detect_runtime:
        return

    try:
        from app.services.cluster_runtime_service import register_node_heartbeat

        region = str(payload.get("region") or "unknown").strip() or "unknown"
        role = str(payload.get("role") or "worker").strip() or "worker"
        active_threads = max(0, int(detect_runtime.get("active_threads") or 0))
        max_threads = max(0, int(detect_runtime.get("max_threads") or 0))
        current_load = max(
            0,
            int(detect_runtime.get("current_load") or 0),
            active_threads,
        )
        worker_online = bool(detect_runtime.get("worker_online", False) or detect_runtime.get("service_running", False))
        detect_participating = bool(detect_runtime.get("detect_participating", False) or current_load > 0)
        phase_label = str(detect_runtime.get("phase_label") or "").strip()
        phase_detail = str(detect_runtime.get("phase_detail") or "").strip()
        recent_warning = str(detect_runtime.get("recent_warning") or "").strip()
        has_runtime_signal = any(
            [
                worker_online,
                detect_participating,
                current_load > 0,
                active_threads > 0,
                max_threads > 0,
                bool(phase_label),
                bool(phase_detail),
                bool(recent_warning),
            ]
        )
        if not has_runtime_signal:
            return
        existing_runtime = _load_existing_detect_node_runtime(normalized_node_code)
        existing_metadata = existing_runtime.get("metadata") if isinstance(existing_runtime.get("metadata"), dict) else {}
        existing_active_threads = max(0, int(existing_metadata.get("active_threads", 0) or 0))
        existing_current_load = max(0, int(existing_runtime.get("current_load", 0) or 0), existing_active_threads)
        existing_status = str(existing_runtime.get("status") or "").strip()
        existing_last_heartbeat_at = existing_runtime.get("last_heartbeat_at")
        existing_is_recent = False
        if isinstance(existing_last_heartbeat_at, datetime):
            current_time = (
                datetime.now(existing_last_heartbeat_at.tzinfo)
                if existing_last_heartbeat_at.tzinfo
                else datetime.now()
            )
            existing_is_recent = (current_time - existing_last_heartbeat_at) <= timedelta(seconds=120)
        normalized_phase_detail = phase_detail.lower()
        generic_phase_detail = bool(normalized_phase_detail) and "/" in normalized_phase_detail and " " not in normalized_phase_detail
        weak_agent_snapshot = bool(
            worker_online
            and current_load <= 0
            and active_threads <= 0
            and max_threads > 0
            and not phase_label
            and (not phase_detail or generic_phase_detail)
            and not recent_warning
        )
        if weak_agent_snapshot and existing_is_recent and (
            existing_current_load > 0 or existing_status == "busy"
        ):
            return
        status = "busy" if current_load > 0 else ("online" if worker_online else "unknown")
        heartbeat_metadata = {
            "service": "agent-heartbeat",
            "worker_online": worker_online,
            "detect_participating": detect_participating,
            "active_threads": active_threads,
            "max_threads": max_threads,
            "phase_label": phase_label,
            "phase_detail": phase_detail,
            "recent_warning": recent_warning,
            "updated_at": str(detect_runtime.get("updated_at") or "").strip(),
            "agent_heartbeat_at": datetime.now().isoformat(timespec="seconds"),
        }
        register_node_heartbeat(
            node_code=normalized_node_code,
            region=region,
            role=role,
            status=status,
            current_load=current_load,
            metadata=heartbeat_metadata,
            hostname_override=str(payload.get("hostname") or "").strip(),
            ip_override=str(payload.get("ip") or "").strip(),
        )
    except Exception:
        return


def agent_register(payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, auth
    _upsert_agent_runtime(node_code, payload)
    _upsert_agent_detect_runtime(node_code, payload)
    return True, "Agent 注册成功", {
        "node_code": node_code,
        "expires_at": auth.get("expires_at", ""),
        "capabilities": payload.get("capabilities") or [],
    }


def agent_heartbeat(payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, auth
    _upsert_agent_runtime(node_code, payload)
    _upsert_agent_detect_runtime(node_code, payload)
    return True, "heartbeat ok", {
        "node_code": node_code,
        "server_time": _format_time(datetime.now()),
        "expires_at": auth.get("expires_at", ""),
    }


def _runtime_config_bundle_hash_payload(bundle: dict | None) -> dict:
    normalized = dict(bundle or {})
    normalized.pop("config_hash", None)
    normalized.pop("generated_at", None)
    return normalized


def _build_agent_runtime_config_bundle(node_code: str) -> dict:
    settings_payload = get_settings_payload()
    runtime_settings = get_runtime_settings()
    sensitive_words_payload = get_sensitive_words_payload()
    bundle = {
        "node_code": str(node_code or "").strip(),
        "detect_options": dict(settings_payload.get("detect_options") or {}),
        "proxy_config": dict(settings_payload.get("proxy_config") or {}),
        "thread_count": int(settings_payload.get("thread_count", 2) or 2),
        "node_thread_counts": dict(settings_payload.get("node_thread_counts") or {}),
        "runtime_settings": dict(runtime_settings or {}),
        "sensitive_words": {
            "text": str(sensitive_words_payload.get("text") or ""),
            "total": int(sensitive_words_payload.get("total", 0) or 0),
            "items": list(sensitive_words_payload.get("items") or []),
        },
        "generated_at": _format_time(datetime.now()),
    }
    bundle["config_hash"] = hashlib.sha256(
        json.dumps(
            _runtime_config_bundle_hash_payload(bundle),
            ensure_ascii=False,
            sort_keys=True,
        ).encode("utf-8")
    ).hexdigest()
    return bundle


def agent_pull_runtime_config(payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, auth
    bundle = _build_agent_runtime_config_bundle(node_code)
    return True, "runtime config ok", {
        "node_code": node_code,
        "server_time": _format_time(datetime.now()),
        "expires_at": auth.get("expires_at", ""),
        "bundle": bundle,
    }


def agent_pull_jobs(payload: dict, *, token: str, limit: int = 1) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, _auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, _auth

    safe_limit = min(max(int(limit or 1), 1), 10)
    dispatched_events: list[dict] = []
    with get_db() as conn:
        conn.autocommit = False
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT id
                FROM ops_jobs
                WHERE target_node_code = %s
                  AND status = 'queued'
                  AND execution_mode = 'remote-agent'
                ORDER BY created_at ASC, id ASC
                LIMIT %s
                FOR UPDATE SKIP LOCKED
                """,
                (node_code, safe_limit),
            )
            rows = cur.fetchall()
            job_ids = [int(row[0]) for row in rows]
            jobs: list[dict] = []
            for job_id in job_ids:
                cur.execute(
                    """
                    UPDATE ops_jobs
                    SET status = 'dispatching', started_at = COALESCE(started_at, CURRENT_TIMESTAMP), dispatched_at = CURRENT_TIMESTAMP, updated_at = CURRENT_TIMESTAMP
                    WHERE id = %s
                    """,
                    (job_id,),
                )
                cur.execute(
                    """
                    UPDATE ops_job_steps
                    SET status = 'dispatching', started_at = COALESCE(started_at, CURRENT_TIMESTAMP), updated_at = CURRENT_TIMESTAMP
                    WHERE job_id = %s
                    """,
                    (job_id,),
                )
                dispatched_events.append(
                    {
                        "job_id": job_id,
                        "node_code": node_code,
                        "event_type": "agent_dispatched",
                        "message": f"任务已派发给节点 {node_code}",
                        "payload": {"node_code": node_code},
                    }
                )
                jobs.append(_agent_job_envelope(get_ops_job(job_id)))
        conn.commit()
    for event in dispatched_events:
        append_ops_job_event(**event)
    return True, "ok", {
        "jobs": jobs,
        "count": len(jobs),
        "node_code": node_code,
        "limit": safe_limit,
        "protocol_version": "ops-agent/v1",
        "envelope_type": "agent_job",
    }


def agent_mark_job_started(job_id: int, payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, _auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, _auth

    started_event: dict | None = None
    with get_db() as conn:
        conn.autocommit = False
        with conn.cursor() as cur:
            cur.execute(
                """
                UPDATE ops_jobs
                SET status = 'running', started_at = COALESCE(started_at, CURRENT_TIMESTAMP), updated_at = CURRENT_TIMESTAMP
                WHERE id = %s AND target_node_code = %s
                RETURNING id
                """,
                (int(job_id), node_code),
            )
            row = cur.fetchone()
            if not row:
                conn.rollback()
                return _agent_error(
                    "任务不存在或不属于当前节点",
                    "ops_job_not_owned_by_agent",
                    {"job_id": int(job_id), "node_code": node_code},
                )
            cur.execute(
                """
                UPDATE ops_job_steps
                SET status = 'running', started_at = COALESCE(started_at, CURRENT_TIMESTAMP), updated_at = CURRENT_TIMESTAMP
                WHERE job_id = %s
                RETURNING id
                """,
                (int(job_id),),
            )
            step_rows = cur.fetchall()
            step_ids = [int(item[0]) for item in step_rows]
            started_event = {
                "job_id": int(job_id),
                "node_code": node_code,
                "event_type": "agent_started",
                "message": f"节点 {node_code} 已开始执行任务",
                "payload": {"step_ids": step_ids},
            }
        conn.commit()
    if started_event:
        append_ops_job_event(**started_event)
    job = get_ops_job(int(job_id))
    return True, "任务已标记为运行中", {
        "job": job,
        "accepted_job": _agent_job_envelope(job),
        "focus_ref": dict(job.get("focus_ref") or {}),
    }


def agent_complete_job(job_id: int, payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, _auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, _auth

    job_status = str(payload.get("status") or "").strip().lower()
    if job_status not in {"success", "failed", "partially_succeeded"}:
        return _agent_error(
            "status 必须是 success / failed / partially_succeeded",
            "ops_job_invalid_status",
            {"allowed_statuses": ["success", "failed", "partially_succeeded"]},
        )

    stdout_text = str(payload.get("stdout") or "")
    stderr_text = str(payload.get("stderr") or "")
    result = dict(payload.get("result") or {})
    error_message = str(payload.get("error_message") or "").strip()
    client_request_id = str(payload.get("client_request_id") or "").strip()[:128]
    duration_ms = max(0, int(payload.get("duration_ms") or 0))
    summary_text = str(payload.get("summary_text") or "").strip()
    focus_ref = payload.get("focus_ref") if isinstance(payload.get("focus_ref"), dict) else {}
    if duration_ms and "duration_ms" not in result:
        result["duration_ms"] = duration_ms
    if summary_text and "summary_text" not in result and "summary" not in result:
        result["summary_text"] = summary_text
    if focus_ref and "focus_ref" not in result:
        result["focus_ref"] = focus_ref
    event_level = "info" if job_status == "success" else ("warning" if job_status == "partially_succeeded" else "error")

    completed_event: dict | None = None
    with get_db() as conn:
        conn.autocommit = False
        with conn.cursor() as cur:
            cur.execute(
                """
                SELECT id, COALESCE(last_agent_complete_request_id, '')
                FROM ops_jobs
                WHERE id = %s AND target_node_code = %s
                LIMIT 1
                """,
                (int(job_id), node_code),
            )
            job_row = cur.fetchone()
            if not job_row:
                conn.rollback()
                return _agent_error(
                    "任务不存在或不属于当前节点",
                    "ops_job_not_owned_by_agent",
                    {"job_id": int(job_id), "node_code": node_code},
                )
            current_request_id = str(job_row[1] or "")
            if client_request_id and current_request_id == client_request_id:
                conn.rollback()
                deduplicated_job = get_ops_job(int(job_id))
                return True, "任务结果已幂等回写", {
                    "job": deduplicated_job,
                    "deduplicated": True,
                    "client_request_id": client_request_id,
                    "completion_summary": _completion_summary(
                        deduplicated_job,
                        client_request_id=client_request_id,
                        deduplicated=True,
                    ),
                }

            cur.execute(
                """
                UPDATE ops_jobs
                SET
                    status = %s,
                    result_json = %s,
                    error_message = %s,
                    last_agent_complete_request_id = CASE WHEN %s <> '' THEN %s ELSE last_agent_complete_request_id END,
                    finished_at = CURRENT_TIMESTAMP,
                    updated_at = CURRENT_TIMESTAMP
                WHERE id = %s
                """,
                (
                    job_status,
                    json.dumps(result, ensure_ascii=False),
                    error_message if job_status != "success" else "",
                    client_request_id,
                    client_request_id,
                    int(job_id),
                ),
            )
            cur.execute(
                """
                UPDATE ops_job_steps
                SET
                    status = %s,
                    stdout_text = %s,
                    stderr_text = %s,
                    result_json = %s,
                    finished_at = CURRENT_TIMESTAMP,
                    updated_at = CURRENT_TIMESTAMP
                WHERE job_id = %s
                RETURNING id
                """,
                (
                    job_status,
                    stdout_text,
                    stderr_text,
                    json.dumps(result, ensure_ascii=False),
                    int(job_id),
                ),
            )
            step_rows = cur.fetchall()
            step_ids = [int(item[0]) for item in step_rows]
            completed_event = {
                "job_id": int(job_id),
                "node_code": node_code,
                "client_event_id": (f"complete:{client_request_id}" if client_request_id else ""),
                "event_type": "agent_completed",
                "message": f"节点 {node_code} 已完成任务，状态: {job_status}",
                "level": event_level,
                "payload": {
                    "step_ids": step_ids,
                    "result": result,
                    "duration_ms": duration_ms,
                    "summary_text": str(result.get("summary_text") or result.get("summary") or summary_text).strip(),
                    "focus_ref": focus_ref,
                },
            }
        conn.commit()
    if completed_event:
        append_ops_job_event(**completed_event)
    from app.services.ops_release_service import refresh_release_rollout_for_job

    refresh_release_rollout_for_job(int(job_id))
    job = get_ops_job(int(job_id))
    return True, "任务结果已回写", {
        "job": job,
        "deduplicated": False,
        "client_request_id": client_request_id,
        "completion_summary": _completion_summary(job, client_request_id=client_request_id),
    }


def agent_append_job_event(job_id: int, payload: dict, *, token: str) -> tuple[bool, str, dict]:
    node_code = str(payload.get("node_code") or "").strip()
    if not node_code:
        return _agent_error("node_code 不能为空", "agent_node_code_required")
    ok, message, _auth = _authenticate_agent_token(token, expected_node_code=node_code)
    if not ok:
        return False, message, _auth
    client_event_id = str(payload.get("client_event_id") or "").strip()[:128]
    event_payload = dict(payload.get("payload") or {})
    occurred_at = str(payload.get("occurred_at") or "").strip()
    summary_text = str(payload.get("summary_text") or "").strip()
    focus_ref = payload.get("focus_ref") if isinstance(payload.get("focus_ref"), dict) else {}
    if occurred_at and "occurred_at" not in event_payload:
        event_payload["occurred_at"] = occurred_at
    if summary_text and "summary_text" not in event_payload:
        event_payload["summary_text"] = summary_text
    if focus_ref and "focus_ref" not in event_payload:
        event_payload["focus_ref"] = focus_ref
    event_id = append_ops_job_event(
        job_id=int(job_id),
        step_id=int(payload.get("step_id") or 0) or None,
        node_code=node_code,
        client_event_id=client_event_id,
        event_type=str(payload.get("event_type") or "agent_event").strip() or "agent_event",
        level=str(payload.get("level") or "info").strip() or "info",
        message=str(payload.get("message") or "").strip()[:2000] or "agent event",
        payload=event_payload,
    )
    return True, "事件已写入", {
        "event_id": event_id,
        "client_event_id": client_event_id,
        "event": get_ops_job_event(event_id),
    }