Api-Gateway/app/admin/stats.py

"""JSON endpoints backing the monitoring charts.

All endpoints accept:
  hours       — window size (1..720)
  service_id  — restrict to one service
  user_id     — restrict to keys owned by one user
  key_id      — restrict to one API key

Time series use adaptive buckets: 5 min (<=6 h), 1 h (<=48 h), 1 day beyond.
SQLite-specific date functions are used (see README).
"""
import time
from datetime import datetime, timedelta, timezone

from fastapi import APIRouter, Depends, Query
from sqlalchemy import Integer, func
from sqlalchemy.orm import Query as OrmQuery, Session

from app.admin.deps import current_user
from app.database import get_db
from app.models import ApiKey, Endpoint, RequestLog, Service, User

router = APIRouter(prefix="/admin/api", dependencies=[Depends(current_user)])


class Filters:
    def __init__(
        self,
        hours: int = Query(24, ge=1, le=720),
        service_id: int | None = Query(None),
        user_id: int | None = Query(None),
        key_id: int | None = Query(None),
    ):
        self.hours = hours
        self.since = datetime.now(timezone.utc) - timedelta(hours=hours)
        self.service_id = service_id
        self.user_id = user_id
        self.key_id = key_id

    def apply(self, q: OrmQuery, joined_key: bool = False) -> OrmQuery:
        """Apply window + filters. Pass joined_key when the query already
        joins ApiKey so we don't join twice."""
        q = q.filter(RequestLog.timestamp >= self.since)
        if self.service_id:
            q = q.filter(RequestLog.service_id == self.service_id)
        if self.key_id:
            q = q.filter(RequestLog.api_key_id == self.key_id)
        if self.user_id:
            if not joined_key:
                q = q.join(ApiKey, RequestLog.api_key_id == ApiKey.id)
            q = q.filter(ApiKey.user_id == self.user_id)
        return q


def _bucket_seconds(hours: int) -> int:
    if hours <= 6:
        return 300
    if hours <= 48:
        return 3600
    return 86400


def _label_format(hours: int, bucket: int) -> str:
    if bucket < 3600:
        return "%H:%M"
    if bucket == 3600:
        return "%H:00" if hours <= 24 else "%m-%d %H:00"
    return "%Y-%m-%d"


def _bucket_expr(bucket: int):
    # epoch - (epoch % bucket) floors to the bucket start; plain `/` must be
    # avoided because SQLAlchemy renders it as true (float) division.
    epoch = func.cast(func.strftime("%s", RequestLog.timestamp), Integer)
    return epoch - (epoch % bucket)


def _series(f: Filters, db: Session, *aggregates):
    """Group the filtered logs into time buckets and fill gaps with None.
    Returns (labels, [values-per-aggregate])."""
    bucket = _bucket_seconds(f.hours)
    fmt = _label_format(f.hours, bucket)
    expr = _bucket_expr(bucket).label("bucket")
    rows = (
        f.apply(db.query(expr, *aggregates))
        .group_by("bucket").order_by("bucket").all()
    )
    found = {int(r[0]): r for r in rows}
    start = int(f.since.timestamp()) // bucket * bucket
    now = int(time.time())
    labels: list[str] = []
    series: list[list] = [[] for _ in aggregates]
    for t in range(start, now + 1, bucket):
        labels.append(datetime.fromtimestamp(t, timezone.utc).strftime(fmt))
        row = found.get(t)
        for i in range(len(aggregates)):
            series[i].append(row[i + 1] if row else None)
    return labels, series


@router.get("/stats/summary")
def summary(f: Filters = Depends(), db: Session = Depends(get_db)):
    total = f.apply(db.query(func.count(RequestLog.id))).scalar() or 0
    errors = f.apply(db.query(func.count(RequestLog.id))).filter(
        RequestLog.status_code >= 500).scalar() or 0
    avg_latency = f.apply(db.query(func.avg(RequestLog.latency_ms))).scalar() or 0
    return {
        "total_requests": total,
        "error_count": errors,
        "error_rate": round(errors / total * 100, 2) if total else 0,
        "avg_latency_ms": round(avg_latency, 1),
        "active_services": db.query(Service).filter(Service.is_active).count(),
        "active_keys": db.query(ApiKey).filter(ApiKey.is_active).count(),
    }


@router.get("/stats/timeseries")
def timeseries(f: Filters = Depends(), db: Session = Depends(get_db)):
    labels, (ok, errors) = _series(
        f, db,
        func.sum(func.iif(RequestLog.status_code < 500, 1, 0)),
        func.sum(func.iif(RequestLog.status_code >= 500, 1, 0)),
    )
    return {
        "labels": labels,
        "ok": [int(v) if v is not None else 0 for v in ok],
        "errors": [int(v) if v is not None else 0 for v in errors],
    }


@router.get("/stats/latency-timeseries")
def latency_timeseries(f: Filters = Depends(), db: Session = Depends(get_db)):
    labels, (avg_ms,) = _series(f, db, func.avg(RequestLog.latency_ms))
    return {
        "labels": labels,
        "avg_ms": [round(v, 1) if v is not None else None for v in avg_ms],
    }


@router.get("/stats/by-service")
def by_service(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(Service.name, func.count(RequestLog.id).label("count"))
        .join(RequestLog, RequestLog.service_id == Service.id)
    )
    rows = (
        f.apply(q)
        .group_by(Service.id)
        .order_by(func.count(RequestLog.id).desc())
        .all()
    )
    return {"labels": [r.name for r in rows], "counts": [r.count for r in rows]}


@router.get("/stats/by-endpoint")
def by_endpoint(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(Service.slug, Endpoint.method, Endpoint.path,
                 func.count(RequestLog.id).label("count"))
        .select_from(RequestLog)
        .join(Endpoint, RequestLog.endpoint_id == Endpoint.id)
        .join(Service, Endpoint.service_id == Service.id)
    )
    rows = (
        f.apply(q)
        .group_by(Endpoint.id)
        .order_by(func.count(RequestLog.id).desc())
        .limit(15)
        .all()
    )
    return {
        "labels": [f"{r.slug}: {r.method} {r.path}" for r in rows],
        "counts": [r.count for r in rows],
    }


@router.get("/stats/by-status")
def by_status(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(RequestLog.status_code, func.count(RequestLog.id))
    )
    rows = (
        f.apply(q)
        .group_by(RequestLog.status_code)
        .order_by(RequestLog.status_code)
        .all()
    )
    return {"labels": [str(r[0]) for r in rows], "counts": [r[1] for r in rows]}


@router.get("/stats/latency-by-service")
def latency_by_service(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(Service.name, func.avg(RequestLog.latency_ms).label("avg_ms"))
        .join(RequestLog, RequestLog.service_id == Service.id)
    )
    rows = (
        f.apply(q)
        .group_by(Service.id)
        .order_by(func.avg(RequestLog.latency_ms).desc())
        .all()
    )
    return {"labels": [r.name for r in rows], "avg_ms": [round(r.avg_ms, 1) for r in rows]}


@router.get("/stats/top-keys")
def top_keys(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(ApiKey.name, func.count(RequestLog.id).label("count"))
        .join(RequestLog, RequestLog.api_key_id == ApiKey.id)
    )
    rows = (
        f.apply(q, joined_key=True)
        .group_by(ApiKey.id)
        .order_by(func.count(RequestLog.id).desc())
        .limit(10)
        .all()
    )
    return {"labels": [r.name for r in rows], "counts": [r.count for r in rows]}


@router.get("/stats/by-user")
def by_user(f: Filters = Depends(), db: Session = Depends(get_db)):
    q = (
        db.query(User.username, func.count(RequestLog.id).label("count"))
        .select_from(RequestLog)
        .join(ApiKey, RequestLog.api_key_id == ApiKey.id)
        .join(User, ApiKey.user_id == User.id)
    )
    rows = (
        f.apply(q, joined_key=True)
        .group_by(User.id)
        .order_by(func.count(RequestLog.id).desc())
        .limit(15)
        .all()
    )
    return {"labels": [r.username for r in rows], "counts": [r.count for r in rows]}