ForcePilot/backend/server/routers/external_systems/metric_router.py

"""Metric 子域 Router。

外部系统限界上下文的指标查询 API，覆盖桶列表 / 聚合 / 时间序列 /
跨系统排名 / 按环境分解 / 最新快照 / Prometheus 导出 / 过期清理。
所有端点通过 ``create_use_cases_from_db`` 装配 use_cases，
经 ``metric_service`` 端口调用用例。

路径顺序约束：静态路径（aggregate/timeseries/by-system/by-env/latest/prometheus/buckets）
必须在根路径之前声明。

认证策略：
- 查询类端点使用 get_required_user（支持 JWT + API Key 双模认证）
- 清理端点使用 get_admin_user
- Prometheus 端点复用 get_required_user，通过 Authorization: Bearer yxkey_<key> 支持 API Key 抓取

时间范围约束：
- aggregate / timeseries / by-system / by-env 均要求 start_time + end_time，避免全表扫描
- timeseries 额外限制最大范围（hour→31 天，day→366 天），防止结果集膨胀
- by-system / by-env / aggregate 限制最大范围 366 天
"""

from __future__ import annotations

from datetime import datetime, timedelta
from typing import Any, Literal

from fastapi import APIRouter, Depends, Query, Response
from yuxi.external_systems.exceptions import DomainValidationError
from yuxi.external_systems.infrastructure.container import UseCases
from yuxi.external_systems.use_cases.dto.metric import (
    DeleteOldBucketsInput,
    GetLatestBucketInput,
    ListMetricBucketsInput,
    MetricAggregateInput,
    MetricByEnvInput,
    MetricBySystemInput,
    MetricTimeseriesInput,
    PrometheusExportInput,
)
from yuxi.storage.postgres.models_business import User
from yuxi.utils.datetime_utils import utc_now_naive

from server.routers.external_systems import get_use_cases, parse_and_naive, validate_time_range
from server.utils.auth_middleware import get_admin_user, get_required_user

metric_router = APIRouter(prefix="/metrics", tags=["external-systems-metric"])

# 最大查询范围约束
_MAX_RANGE_DAYS = 366
_TIMESERIES_MAX_RANGE: dict[str, timedelta] = {
    "hour": timedelta(days=31),
    "day": timedelta(days=366),
}


def _validate_max_range(start: datetime, end: datetime, max_delta: timedelta) -> None:
    """校验时间范围不超过 max_delta（时区转换与 start<=end 校验由共享函数完成）。"""
    if end - start > max_delta:
        raise DomainValidationError(f"时间范围不得超过 {max_delta.days} 天")


# ---------------- Endpoints ----------------


# ---- 静态路径端点（必须在根路径 "" 之前声明） ----


@metric_router.get("/aggregate", response_model=dict)
async def aggregate_metrics(
    system_id: int | None = Query(None),
    env_key: str | None = Query(None, max_length=32),
    adapter_type: str | None = Query(None, max_length=32),
    start_time: str = Query(..., description="起始时间（ISO 8601，必填，避免全表扫描）"),
    end_time: str = Query(..., description="结束时间（ISO 8601，必填，避免全表扫描）"),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """指标聚合（可按系统/环境/适配器类型过滤）。响应字段对齐 ORM aggregate 返回。"""
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    validate_time_range(start_dt, end_dt)
    _validate_max_range(start_dt, end_dt, timedelta(days=_MAX_RANGE_DAYS))
    input_dto = MetricAggregateInput(
        system_id=system_id,
        env_key=env_key,
        adapter_type=adapter_type,
        start=start_dt,
        end=end_dt,
    )
    output = await use_cases.metric_service.aggregate(input_dto)
    return {"success": True, "data": output.model_dump()}


@metric_router.get("/timeseries", response_model=dict)
async def get_metrics_timeseries(
    system_id: int | None = Query(None),
    env_key: str | None = Query(None, max_length=32),
    adapter_type: str | None = Query(None, max_length=32),
    start_time: str = Query(..., description="起始时间（ISO 8601，必填，避免全表扫描）"),
    end_time: str = Query(..., description="结束时间（ISO 8601，必填，避免全表扫描）"),
    interval: Literal["hour", "day"] = Query("hour", description="聚合间隔：hour / day"),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """时间序列聚合。

    最大范围约束：hour 间隔 ≤ 31 天，day 间隔 ≤ 366 天。
    """
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    validate_time_range(start_dt, end_dt)
    _validate_max_range(start_dt, end_dt, _TIMESERIES_MAX_RANGE[interval])
    input_dto = MetricTimeseriesInput(
        system_id=system_id,
        env_key=env_key,
        adapter_type=adapter_type,
        start=start_dt,
        end=end_dt,
        interval=interval,
    )
    series = await use_cases.metric_service.timeseries(input_dto)
    return {"success": True, "data": {"interval": interval, "series": series}}


@metric_router.get("/by-system", response_model=dict)
async def get_metrics_by_system(
    system_id: int | None = Query(None, description="按系统 ID 过滤（可选）"),
    env_key: str | None = Query(None, max_length=32),
    adapter_type: str | None = Query(None, max_length=32),
    start_time: str = Query(..., description="起始时间（ISO 8601，必填，避免全表扫描）"),
    end_time: str = Query(..., description="结束时间（ISO 8601，必填，避免全表扫描）"),
    limit: int = Query(20, ge=1, le=100),
    offset: int = Query(0, ge=0),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """跨系统聚合排名。按 total_calls 降序，支持 offset 分页。"""
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    validate_time_range(start_dt, end_dt)
    _validate_max_range(start_dt, end_dt, timedelta(days=_MAX_RANGE_DAYS))
    input_dto = MetricBySystemInput(
        system_id=system_id,
        env_key=env_key,
        adapter_type=adapter_type,
        start=start_dt,
        end=end_dt,
        limit=limit,
        offset=offset,
    )
    items = await use_cases.metric_service.aggregate_by_system(input_dto)
    total = await use_cases.metric_service.count_systems(input_dto)
    return {"success": True, "data": {"items": items, "total": total}}


@metric_router.get("/by-env", response_model=dict)
async def get_metrics_by_env(
    system_id: int = Query(..., description="系统 ID（必填）"),
    start_time: str = Query(..., description="起始时间（ISO 8601，必填，避免全表扫描）"),
    end_time: str = Query(..., description="结束时间（ISO 8601，必填，避免全表扫描）"),
    limit: int = Query(20, ge=1, le=100),
    offset: int = Query(0, ge=0),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """单系统按环境分解指标。支持 offset 分页。"""
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    validate_time_range(start_dt, end_dt)
    _validate_max_range(start_dt, end_dt, timedelta(days=_MAX_RANGE_DAYS))
    input_dto = MetricByEnvInput(
        system_id=system_id,
        start=start_dt,
        end=end_dt,
        limit=limit,
        offset=offset,
    )
    items = await use_cases.metric_service.aggregate_by_env(input_dto)
    total = await use_cases.metric_service.count_envs(input_dto)
    return {"success": True, "data": {"items": items, "total": total}}


@metric_router.get("/latest", response_model=dict)
async def get_latest_metric_bucket(
    system_id: int = Query(..., description="系统 ID（必填）"),
    env_key: str = Query("default", max_length=32, description="环境标识"),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """最新桶快照。无数据时 data 为 null。"""
    input_dto = GetLatestBucketInput(system_id=system_id, env_key=env_key)
    data = await use_cases.metric_service.get_latest_bucket(input_dto)
    return {"success": True, "data": data}


@metric_router.get("/prometheus")
async def export_metrics_prometheus(
    start_time: str | None = Query(None, description="起始时间（ISO 8601，不传时默认最近 1 小时）"),
    end_time: str | None = Query(None, description="结束时间（ISO 8601，不传时默认当前）"),
    env_key: str | None = Query(None, max_length=32),
    adapter_type: str | None = Query(None, max_length=32),
    system_id: int | None = Query(None, description="按系统 ID 过滤（可选）"),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> Response:
    """Prometheus 格式导出。支持 API Key 认证（Authorization: Bearer yxkey_<key>）。

    不传时间范围时默认导出最近 1 小时数据。支持 adapter_type / system_id 过滤。
    传入 start_time 时校验 start ≤ end 且不超过 366 天（end_time 未传时以当前时间补全）。
    """
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    if start_dt is not None:
        effective_end = end_dt if end_dt is not None else utc_now_naive()
        validate_time_range(start_dt, effective_end)
        _validate_max_range(start_dt, effective_end, timedelta(days=_MAX_RANGE_DAYS))
        if end_dt is not None:
            end_dt = effective_end
    input_dto = PrometheusExportInput(
        start=start_dt,
        end=end_dt,
        env_key=env_key,
        adapter_type=adapter_type,
        system_id=system_id,
    )
    output = await use_cases.metric_service.export_prometheus(input_dto)
    return Response(content=output.content, media_type=output.content_type)


@metric_router.delete("/buckets", response_model=dict)
async def delete_old_metric_buckets(
    before: str = Query(..., description="清理此时间之前的桶（ISO 8601，必填）"),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_admin_user),
) -> dict[str, Any]:
    """清理过期指标桶（软删除）。before 必填，避免误删全表。

    审计字段 updated_by 记录执行清理的管理员 uid。
    """
    before_dt = parse_and_naive("before", before)
    input_dto = DeleteOldBucketsInput(before=before_dt, updated_by=current_user.uid)
    output = await use_cases.metric_service.delete_old_buckets(input_dto)
    return {"success": True, "data": output.model_dump()}


# ---- 根路径端点（最后声明） ----


@metric_router.get("", response_model=dict)
async def list_metric_buckets(
    system_id: int | None = Query(None),
    env_key: str | None = Query(None, max_length=32),
    start_time: str | None = Query(None, description="开始时间（ISO 8601）"),
    end_time: str | None = Query(None, description="结束时间（ISO 8601）"),
    limit: int = Query(20, ge=1, le=100),
    offset: int = Query(0, ge=0),
    use_cases: UseCases = Depends(get_use_cases),
    current_user: User = Depends(get_required_user),
) -> dict[str, Any]:
    """桶列表分页查询。返回 Metric dataclass 字段 + avg_latency_ms。"""
    start_dt = parse_and_naive("start_time", start_time)
    end_dt = parse_and_naive("end_time", end_time)
    validate_time_range(start_dt, end_dt)
    input_dto = ListMetricBucketsInput(
        system_id=system_id,
        env_key=env_key,
        start=start_dt,
        end=end_dt,
        limit=limit,
        offset=offset,
    )
    output = await use_cases.metric_service.list_buckets(input_dto)
    return {"success": True, "data": output.model_dump()}