Merge pull request 'Handle expected_minutes=0 gracefully (defense-in-depth)' (#61) from fix/media-verify-runtime-zero-handling into main

allowed_categories kannte nur serien4k|serien|video4k|video. Die englischen
Arr-Instanzen sonarrEN (Port 8991, Root /data/FHD/serienen) und radarrEN
(Port 7880, Root /data/FHD/videoen) nutzen eigene SAB-Kategorien.

Folge: POST /media/handoff antwortete 422 "Unsupported media category",
movetdarr.sh brach mit "n8n-Handoff konnte nicht registriert werden - kein
Move" ab und Mutiny.2026 x2 (~21 GB) lagen seit 08./09.09.2026 unangetastet
in /usenet/complete/videoen. radarrEN hat den Film weiter monitored,
hasFile=false, Queue leer.

Die Pfadpruefung (expected_prefix) bleibt unveraendert und gilt auch fuer die
neuen Kategorien.

3 neue Tests: serienen+videoen werden geproxyt, erfundene Kategorie bleibt
422, Pfad-Mismatch bei videoen bleibt 422.
6/6 media_handoff-Tests gruen. test_health_exposes_current_version schlug
schon vor dieser Aenderung fehl (Test erwartet 2.4.1, VERSION ist 2.4.6).

README.md

@@ -45,6 +45,7 @@ Known secret response fields such as Dockhand's `hawserToken` and `webhookSecret
 | `/docker/inspect/{host}/{container}` | GET | Sanitized image, runtime, resources, mounts and state |
 | `/docker/restart/{host}/{container}` | POST | Restart container; supports `dry_run=true` |
 | `/config/reload` | POST | Reload YAML configuration and credential cache |
+| `/media/verify` | POST | ffprobe duration check on a NAS media file to catch truncated Sonarr/Radarr imports; flags `suspect` if deviation from `expected_minutes` exceeds `tolerance_pct` |
 
 ## VM lifecycle and inventory
 
@@ -97,6 +98,14 @@ Integration Compose definition: `tests/compose.integration.yaml` (binds only to
 
 ## Changelog
 
+### 2.4.8 — 17.09.2026
+
+- Fixed `POST /media/verify` false-positive rejection: filenames containing a legitimate apostrophe (e.g. "La'An" in a Star Trek episode title) were blocked as "unsafe characters". Replaced the naive single-quote wrapping + apostrophe blocklist with proper `shlex.quote()` escaping for the remote ffprobe command; only newline/NUL byte injection is still rejected.
+
+### 2.4.7 — 17.09.2026
+
+- Added `POST /media/verify`: probes a media file's real duration via `ffprobe` (running inside the existing `bazarrUHD` container on `arrapps`, which already mounts `/data` and ships ffmpeg — no new service needed) and flags it as `suspect` when it deviates from an `expected_minutes` value beyond `tolerance_pct`. Catches truncated Sonarr/Radarr imports (e.g. a re-grab that lands as 1.8 GB instead of the expected 8 GB for a UHD episode) after the file is already on the NAS.
+
 ### 2.3.2 — 22.07.2026
 
 - Make Vaultwarden refresh durable: persistent named cache volume, protected runtime credentials, automatic API-key re-login and atomic cache writes.

app.py

@@ -1,7 +1,7 @@
 """Homelab Butler v2.1 – Unified API proxy for Pfannkuchen homelab.
 Reads service config from butler.yaml, credentials from Vaultwarden cache with flat-file fallback."""
 
-import os, json, asyncio, logging, time, base64, re, subprocess, ipaddress, secrets, sqlite3, math, hashlib
+import os, json, asyncio, logging, time, base64, re, subprocess, ipaddress, secrets, sqlite3, math, hashlib, shlex
 from datetime import datetime, timezone
 import httpx, yaml
 from typing import Literal
@@ -12,7 +12,7 @@ from contextlib import asynccontextmanager
 from contextvars import ContextVar
 
 log = logging.getLogger("butler")
-VERSION = "2.4.6"
+VERSION = "2.4.9"
 
 API_DIR = os.environ.get("API_KEY_DIR", "/data/api")
 VAULT_CACHE_DIR = os.environ.get("VAULT_CACHE_DIR", "/data/vault-cache")
@@ -1546,6 +1546,102 @@ async def vault_reload(_=Depends(_verify)):
     return {"reloaded": True, "items": len(_vault_cache)}
 
 
+# --- Media file integrity verification ---
+
+MEDIA_VERIFY_PROBES = {
+    "arrapps": ("bazarrUHD", "sascha"),
+    "arr-chris": (None, "chris"),
+    "arr-chris-live": (None, "chris"),
+}
+
+
+class MediaVerifyRequest(BaseModel):
+    host: str = Field(..., max_length=64)
+    path: str = Field(..., max_length=1000)
+    expected_minutes: float | None = Field(None, ge=0, le=1440)
+    tolerance_pct: float = Field(20.0, gt=0, le=100)
+
+
+@app.post("/media/verify")
+async def media_verify(payload: MediaVerifyRequest, _=Depends(_verify)):
+    """Probe a media file's real duration via ffprobe to catch truncated imports.
+
+    Runs `ffprobe` inside an existing container (bazarrUHD on arrapps, which already
+    mounts /data and ships ffmpeg) so no new service is required. Compares the
+    measured duration against an expected runtime (minutes) supplied by the caller
+    (e.g. Sonarr/Radarr's runtime field) within tolerance_pct.
+    """
+    if not re.fullmatch(r"[A-Za-z0-9_.-]+", payload.host):
+        raise HTTPException(400, "Invalid host name")
+    if payload.host not in MEDIA_VERIFY_PROBES:
+        raise HTTPException(400, f"No ffprobe container configured for host {payload.host}")
+    if not re.fullmatch(r"/data/[^\x00]+\.(mkv|mp4|avi|m4v|ts)", payload.path):
+        raise HTTPException(400, "Path must be an absolute /data media file")
+    if ".." in payload.path or "\n" in payload.path or "\x00" in payload.path:
+        raise HTTPException(400, "Path contains unsafe characters")
+
+    container, _default_user = MEDIA_VERIFY_PROBES[payload.host]
+    if not container:
+        raise HTTPException(400, f"Host {payload.host} has no configured ffprobe container yet")
+
+    target = _find_inventory_host(payload.host)
+    if not target:
+        raise HTTPException(404, f"Host {payload.host} not found in inventory")
+
+    # expected_minutes == 0 or None: no reference runtime available, skip verification gracefully
+    if not payload.expected_minutes or payload.expected_minutes <= 0:
+        return {
+            "host": payload.host,
+            "path": payload.path,
+            "duration_seconds": None,
+            "duration_minutes": None,
+            "expected_minutes": payload.expected_minutes,
+            "verified": False,
+            "suspect": False,
+            "skipped": True,
+            "skip_reason": "no_reference_runtime",
+            "deviation_pct": None,
+        }
+
+    probe_cmd = (
+        f"docker exec {container} ffprobe -v error "
+        f"-show_entries format=duration -of default=noprint_wrappers=1:nokey=1 {shlex.quote(payload.path)}"
+    )
+    rc, out, err = await asyncio.to_thread(
+        _ssh, f'{target["user"]}@{target["ip"]}', f"sudo -n {probe_cmd} || {probe_cmd}", 30
+    )
+    if rc != 0:
+        _audit("/media/verify", "POST", 502, f"host={payload.host} path={payload.path}")
+        raise HTTPException(502, (err or out).strip()[:500] or "ffprobe failed")
+
+    raw_duration = out.strip()
+    try:
+        duration_seconds = float(raw_duration)
+    except ValueError:
+        _audit("/media/verify", "POST", 502, f"host={payload.host} unparsable duration")
+        raise HTTPException(502, "ffprobe returned no parsable duration; file is likely corrupt")
+
+    duration_minutes = duration_seconds / 60
+    result = {
+        "host": payload.host,
+        "path": payload.path,
+        "duration_seconds": round(duration_seconds, 1),
+        "duration_minutes": round(duration_minutes, 2),
+        "expected_minutes": payload.expected_minutes,
+        "verified": True,
+        "suspect": False,
+    }
+    if payload.expected_minutes:
+        deviation_pct = abs(duration_minutes - payload.expected_minutes) / payload.expected_minutes * 100
+        result["deviation_pct"] = round(deviation_pct, 1)
+        result["suspect"] = deviation_pct > payload.tolerance_pct
+    _audit(
+        "/media/verify", "POST", 200,
+        f"host={payload.host} dur={duration_minutes:.1f}m suspect={result['suspect']}",
+    )
+    return result
+
+
 # --- VPS reverse-proxy and DNS management ---
 
 VPS_SSH = "root@46.225.230.72"
@@ -4121,7 +4217,11 @@ async def media_handoff(payload: MediaHandoffPayload, request: Request):
         _audit("/media/handoff", "POST", 403, f"caller={caller} action={payload.action}")
         raise HTTPException(403, "Media handoff caller is not allowed")
 
-    allowed_categories = {"serien4k", "serien", "video4k", "video"}
+    # 16.09.2026: serienen/videoen ergaenzt. Die englischen Arr-Instanzen
+    # sonarrEN (Port 8991, Root /data/FHD/serienen) und radarrEN (Port 7880,
+    # Root /data/FHD/videoen) nutzen eigene SAB-Kategorien. Ohne sie brach der
+    # Handoff mit 422 ab und Releases blieben in /usenet/complete liegen.
+    allowed_categories = {"serien4k", "serien", "serienen", "video4k", "video", "videoen"}
     if payload.action == "start":
         if payload.category not in allowed_categories:
             raise HTTPException(422, "Unsupported media category")

tests/test_app.py

@@ -1,6 +1,7 @@
 import os
 import asyncio
 import json
+import shlex
 import time
 from datetime import datetime, timedelta, timezone
 
@@ -117,6 +118,99 @@ def test_media_handoff_rejects_wrong_category_path(monkeypatch):
     assert response.status_code == 422
 
 
+def test_media_handoff_accepts_english_arr_categories(monkeypatch):
+    """serienen/videoen muessen durchgehen (sonarrEN 8991 / radarrEN 7880).
+
+    16.09.2026: fehlten in allowed_categories -> HTTP 422 -> movetdarr.sh
+    brach mit "n8n-Handoff konnte nicht registriert werden" ab und liess
+    Mutiny.2026 x2 tagelang in /usenet/complete/videoen liegen.
+    """
+    seen = []
+
+    class FakeResponse:
+        status_code = 200
+
+        def json(self):
+            return {"ok": True, "jobId": "test-job-5678", "state": "registered"}
+
+    class FakeClient:
+        def __init__(self, **_kwargs):
+            pass
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, *_args):
+            return None
+
+        async def post(self, url, json, headers):
+            seen.append(json)
+            return FakeResponse()
+
+    monkeypatch.setattr(app.httpx, "AsyncClient", FakeClient)
+    with TestClient(app.app) as client:
+        monkeypatch.setattr(app, "SERVICES", {"n8n": {"url": "http://n8n:5678", "auth": "n8n"}})
+        for category in ("serienen", "videoen"):
+            response = client.post(
+                "/media/handoff",
+                headers={"Authorization": "Bearer test-token"},
+                json={
+                    "action": "start",
+                    "category": category,
+                    "directory": f"/usenet/complete/{category}/Release.2026",
+                    "release": "Release.2026-GRP",
+                    "cleanName": "Release 2026",
+                    "expectedFiles": 1,
+                },
+            )
+            assert response.status_code == 200, (category, response.text)
+            assert response.json()["state"] == "registered", category
+
+    assert [item["category"] for item in seen] == ["serienen", "videoen"]
+
+
+def test_media_handoff_still_rejects_unknown_category(monkeypatch):
+    """Fail-closed bleibt: eine frei erfundene Kategorie wird nicht geproxyt."""
+    monkeypatch.setattr(
+        app.httpx,
+        "AsyncClient",
+        lambda **_kwargs: (_ for _ in ()).throw(AssertionError("must not proxy")),
+    )
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/handoff",
+            headers={"Authorization": "Bearer test-token"},
+            json={
+                "action": "start",
+                "category": "hoerbuecher",
+                "directory": "/usenet/complete/hoerbuecher/Buch",
+                "release": "Buch",
+            },
+        )
+    assert response.status_code == 422
+
+
+def test_media_handoff_english_category_path_must_match(monkeypatch):
+    """Pfadpruefung gilt auch fuer die neuen Kategorien."""
+    monkeypatch.setattr(
+        app.httpx,
+        "AsyncClient",
+        lambda **_kwargs: (_ for _ in ()).throw(AssertionError("must not proxy")),
+    )
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/handoff",
+            headers={"Authorization": "Bearer test-token"},
+            json={
+                "action": "start",
+                "category": "videoen",
+                "directory": "/usenet/complete/video4k/Wrong",
+                "release": "Wrong",
+            },
+        )
+    assert response.status_code == 422
+
+
 def test_paperless_import_queues_pdf_through_butler(monkeypatch):
     captured = {}
 
@@ -808,6 +902,115 @@ def test_uptime_monitor_remove_rejects_invalid_expected_name_before_ssh(monkeypa
     assert response.status_code == 400
 
 
+def test_media_verify_returns_duration_and_not_suspect_when_within_tolerance(monkeypatch):
+    monkeypatch.setattr(app, "_find_inventory_host", lambda host: {"user": "sascha", "ip": "10.2.1.100"})
+    calls = []
+
+    def fake_ssh(host, command, timeout=30):
+        calls.append((host, command, timeout))
+        return 0, "2967.355000\n", ""
+
+    monkeypatch.setattr(app, "_ssh", fake_ssh)
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": "/data/UHD/serien/Show/ep.mkv", "expected_minutes": 49.5},
+        )
+    assert response.status_code == 200
+    body = response.json()
+    assert body["duration_minutes"] == 49.46
+    assert body["suspect"] is False
+    assert calls[0][0] == "sascha@10.2.1.100"
+    assert "bazarrUHD" in calls[0][1]
+    assert "ffprobe" in calls[0][1]
+
+
+def test_media_verify_flags_truncated_file_as_suspect(monkeypatch):
+    monkeypatch.setattr(app, "_find_inventory_host", lambda host: {"user": "sascha", "ip": "10.2.1.100"})
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (0, "1836.0\n", ""))
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": "/data/UHD/serien/Show/ep.mkv", "expected_minutes": 49.5},
+        )
+    assert response.status_code == 200
+    body = response.json()
+    assert body["suspect"] is True
+    assert body["deviation_pct"] > 20
+
+
+def test_media_verify_rejects_path_outside_data_before_ssh(monkeypatch):
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (_ for _ in ()).throw(AssertionError("must not SSH")))
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": "/etc/passwd"},
+        )
+    assert response.status_code == 400
+
+
+def test_media_verify_allows_apostrophe_in_filename_and_quotes_it_safely(monkeypatch):
+    monkeypatch.setattr(app, "_find_inventory_host", lambda host: {"user": "sascha", "ip": "10.2.1.100"})
+    calls = []
+
+    def fake_ssh(host, command, timeout=30):
+        calls.append((host, command, timeout))
+        return 0, "2967.0\n", ""
+
+    monkeypatch.setattr(app, "_ssh", fake_ssh)
+    path = "/data/FHD/serien/Star Trek - Strange New Worlds (2022)/Season 04/Once La'An a Time.mkv"
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": path, "expected_minutes": 49.5},
+        )
+    assert response.status_code == 200
+    # shlex.quote must produce a command the remote shell parses as ONE argument,
+    # i.e. no unescaped apostrophe breaks out of quoting.
+    executed_cmd = calls[0][1]
+    quoted = shlex.quote(path)
+    assert quoted in executed_cmd
+    assert shlex.split(executed_cmd)[-1] == path
+
+
+def test_media_verify_rejects_shell_metacharacters_before_ssh(monkeypatch):
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (_ for _ in ()).throw(AssertionError("must not SSH")))
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": "/data/UHD/serien/a\x00.mkv"},
+        )
+    assert response.status_code == 400
+
+
+def test_media_verify_rejects_unknown_host_before_ssh(monkeypatch):
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (_ for _ in ()).throw(AssertionError("must not SSH")))
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "unknown-host", "path": "/data/UHD/serien/ep.mkv"},
+        )
+    assert response.status_code == 400
+
+
+def test_media_verify_returns_502_on_unparsable_ffprobe_output(monkeypatch):
+    monkeypatch.setattr(app, "_find_inventory_host", lambda host: {"user": "sascha", "ip": "10.2.1.100"})
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (0, "N/A\n", ""))
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/media/verify",
+            headers={"Authorization": "Bearer test-token"},
+            json={"host": "arrapps", "path": "/data/UHD/serien/ep.mkv", "expected_minutes": 49.5},
+        )
+    assert response.status_code == 502
+
+
 def test_invalid_log_target_is_rejected_before_ssh():
     with TestClient(app.app) as client:
         response = client.get(