feat: support orientation-aware recording and device wakefulness

Roll scrcpy recordings by device rotation and duration, and publish timeline manifests for playback and analysis.

Add seamless segmented playback in the showcase UI and configurable stay-awake strategies for Android automation.
This commit is contained in:
somew1nd committed 2026-08-22 19:45:02 -07:00
1 parent 160f43a74a
commit 474e56ef8e
14 files changed
+1150 -310

No files matched your search

+3
View File
@@ -19,6 +19,9 @@ OCR_API_KEY=
ADB_HOST=127.0.0.1 ADB_HOST=127.0.0.1
ADB_PORT=5037 ADB_PORT=5037
# ADB_DEVICE_SERIAL= # ADB_DEVICE_SERIAL=
# Keep the display awake, falling back to a wake lock or safe heartbeat when needed.
# Set to false if Artemis is used with a personal device whose screen should time out.
ARTEMIS_KEEP_DEVICE_AWAKE=true
# Artemis Execution Defaults (flash or pro) # Artemis Execution Defaults (flash or pro)
ARTEMIS_DEFAULT_PROFILE=flash ARTEMIS_DEFAULT_PROFILE=flash
+2
View File
@@ -92,10 +92,12 @@ async def get_session_video(session_id: str):
row = session_repo.get_session_by_id(session_id) row = session_repo.get_session_by_id(session_id)
row_dict = dict(row) if row else {"session_id": session_id} row_dict = dict(row) if row else {"session_id": session_id}
v_url = media_service.resolve_video_url(row_dict, video_rec_map, video_idx) v_url = media_service.resolve_video_url(row_dict, video_rec_map, video_idx)
video_segments = media_service.resolve_video_segments(v_url)
return { return {
"session_id": session_id, "session_id": session_id,
"has_video": bool(v_url), "has_video": bool(v_url),
"video_url": v_url, "video_url": v_url,
"video_segments": video_segments,
} }
@@ -127,6 +127,36 @@ class MediaService:
return v_url return v_url
@classmethod
def resolve_video_segments(cls, video_url: str | None) -> list[dict[str, Any]]:
"""Resolve a recording sidecar manifest into browser-safe segment URLs."""
if not video_url or not video_url.startswith("/videos/"):
return []
relative = urllib.parse.unquote(video_url.removeprefix("/videos/"))
video_path = (WORKSPACE_ROOT / relative).resolve()
manifest_path = video_path.parent / "recording.json"
if not manifest_path.is_file():
return []
try:
payload = json.loads(manifest_path.read_text(encoding="utf-8"))
segments = []
for item in payload.get("segments", []):
segment_path = (manifest_path.parent / str(item["file"])).resolve()
if segment_path.parent != manifest_path.parent.resolve() or not segment_path.is_file():
continue
segments.append(
{
"url": cls.path_to_video_url(segment_path),
"start": float(item.get("start", 0)),
"duration": float(item.get("duration", 0)),
"width": int(item.get("width", 0)),
"height": int(item.get("height", 0)),
}
)
return segments
except (OSError, KeyError, TypeError, ValueError, json.JSONDecodeError):
return []
@staticmethod @staticmethod
def unwrap_payload(obj: Any) -> Any: def unwrap_payload(obj: Any) -> Any:
"""Unwraps trace payloads, storing base64 images as disk files and """Unwraps trace payloads, storing base64 images as disk files and
@@ -134,7 +134,7 @@
<div class="video-container" *ngIf="!videoLoadError()"> <div class="video-container" *ngIf="!videoLoadError()">
<video <video
#videoRef #videoRef
[src]="videoUrl" [src]="currentVideoUrl()"
playsinline playsinline
(loadedmetadata)="onLoadedMetadata()" (loadedmetadata)="onLoadedMetadata()"
(timeupdate)="onTimeUpdate()" (timeupdate)="onTimeUpdate()"
@@ -373,16 +373,15 @@
.video-container { .video-container {
position: relative; position: relative;
width: 100%; width: 100%;
height: 100%;
display: flex; display: flex;
flex-direction: column; flex-direction: column;
align-items: center; align-items: center;
justify-content: center; justify-content: center;
background: #ffffff; background: #ffffff;
min-height: 220px;
video { video {
width: 100%; width: 100%;
height: auto;
max-height: 520px; max-height: 520px;
display: block; display: block;
object-fit: contain; object-fit: contain;
@@ -51,6 +51,13 @@ export class FloatingVideoPlayerComponent {
public videoLoadError = signal<boolean>(false); public videoLoadError = signal<boolean>(false);
public liveStreamUrl = signal<string>('/api/stream/device-live'); public liveStreamUrl = signal<string>('/api/stream/device-live');
public liveStreamError = signal<boolean>(false); public liveStreamError = signal<boolean>(false);
public activeSegmentIndex = signal<number>(0);
public currentVideoUrl = computed(() => {
const segments = this.agentService.activeVideoSegments();
return segments[this.activeSegmentIndex()]?.url || this.agentService.activeVideoUrl();
});
private pendingLocalTime: number | null = null;
private pendingAutoplay = false;
// Available speed options // Available speed options
public speedOptions = [0.5, 1.0, 1.5, 2.0, 4.0]; public speedOptions = [0.5, 1.0, 1.5, 2.0, 4.0];
@@ -74,9 +81,11 @@ export class FloatingVideoPlayerComponent {
effect( effect(
() => { () => {
const url = this.agentService.activeVideoUrl(); const url = this.agentService.activeVideoUrl();
const segments = this.agentService.activeVideoSegments();
this.videoLoadError.set(false); this.videoLoadError.set(false);
this.activeSegmentIndex.set(0);
this.currentTime.set(0); this.currentTime.set(0);
this.duration.set(0); this.duration.set(segments.reduce((sum, segment) => sum + segment.duration, 0));
this.isPlaying.set(false); this.isPlaying.set(false);
}, },
{ allowSignalWrites: true } { allowSignalWrites: true }
@@ -177,18 +186,32 @@ export class FloatingVideoPlayerComponent {
public onLoadedMetadata(): void { public onLoadedMetadata(): void {
const v = this.videoRef?.nativeElement; const v = this.videoRef?.nativeElement;
if (v) { if (v) {
this.duration.set(v.duration || 0); const segments = this.agentService.activeVideoSegments();
if (!segments.length) {
this.duration.set(v.duration || 0);
}
this.videoLoadError.set(false); this.videoLoadError.set(false);
v.playbackRate = this.playbackRate(); v.playbackRate = this.playbackRate();
v.loop = this.isLooping(); v.loop = this.isLooping() && !segments.length;
v.muted = this.isMuted();
if (this.pendingLocalTime !== null) {
v.currentTime = Math.max(0, Math.min(v.duration || this.pendingLocalTime, this.pendingLocalTime));
this.pendingLocalTime = null;
}
if (this.pendingAutoplay) {
this.pendingAutoplay = false;
v.play().catch(() => {});
}
} }
} }
public onTimeUpdate(): void { public onTimeUpdate(): void {
const v = this.videoRef?.nativeElement; const v = this.videoRef?.nativeElement;
if (v) { if (v) {
this.currentTime.set(v.currentTime); const segments = this.agentService.activeVideoSegments();
if (v.duration && v.duration !== this.duration()) { const segment = segments[this.activeSegmentIndex()];
this.currentTime.set((segment?.start || 0) + v.currentTime);
if (!segments.length && v.duration && v.duration !== this.duration()) {
this.duration.set(v.duration); this.duration.set(v.duration);
} }
} }
@@ -203,6 +226,17 @@ export class FloatingVideoPlayerComponent {
} }
public onEnded(): void { public onEnded(): void {
const segments = this.agentService.activeVideoSegments();
if (segments.length && this.activeSegmentIndex() < segments.length - 1) {
this.pendingLocalTime = 0;
this.pendingAutoplay = true;
this.activeSegmentIndex.update(index => index + 1);
return;
}
if (segments.length && this.isLooping()) {
this.seek(0, true);
return;
}
this.isPlaying.set(false); this.isPlaying.set(false);
} }
@@ -225,16 +259,31 @@ export class FloatingVideoPlayerComponent {
} }
public restart(): void { public restart(): void {
const v = this.videoRef?.nativeElement; this.seek(0, true);
if (!v) return;
v.currentTime = 0;
v.play().catch(() => {});
} }
public seek(seconds: number): void { public seek(seconds: number, autoplay = false): void {
const v = this.videoRef?.nativeElement; const v = this.videoRef?.nativeElement;
if (!v) return; if (!v) return;
v.currentTime = Math.max(0, Math.min(this.duration(), seconds)); const target = Math.max(0, Math.min(this.duration(), seconds));
const segments = this.agentService.activeVideoSegments();
if (!segments.length) {
v.currentTime = target;
if (autoplay) v.play().catch(() => {});
return;
}
let index = segments.findIndex(segment => target < segment.start + segment.duration);
if (index < 0) index = segments.length - 1;
const localTime = Math.max(0, target - segments[index].start);
if (index === this.activeSegmentIndex()) {
v.currentTime = localTime;
if (autoplay) v.play().catch(() => {});
} else {
this.pendingLocalTime = localTime;
this.pendingAutoplay = autoplay || !v.paused;
this.activeSegmentIndex.set(index);
}
this.currentTime.set(target);
} }
public onScrub(event: Event): void { public onScrub(event: Event): void {
@@ -256,7 +305,7 @@ export class FloatingVideoPlayerComponent {
this.isLooping.set(next); this.isLooping.set(next);
const v = this.videoRef?.nativeElement; const v = this.videoRef?.nativeElement;
if (v) { if (v) {
v.loop = next; v.loop = next && !this.agentService.activeVideoSegments().length;
} }
} }
@@ -22,6 +22,14 @@ import { Session, ModelInfo, TaskQueueItem, AgentStatusResponse } from '../core/
import { StepItemData } from '../core/models/stream.model'; import { StepItemData } from '../core/models/stream.model';
export type { Session, ModelInfo, TaskQueueItem, AgentStatusResponse, StepItemData }; export type { Session, ModelInfo, TaskQueueItem, AgentStatusResponse, StepItemData };
export interface VideoSegment {
url: string;
start: number;
duration: number;
width: number;
height: number;
}
@Injectable({ @Injectable({
providedIn: 'root' providedIn: 'root'
}) })
@@ -104,6 +112,7 @@ export class AgentService {
public isVideoWindowOpen = signal<boolean>(false); public isVideoWindowOpen = signal<boolean>(false);
public isVideoMinimized = signal<boolean>(false); public isVideoMinimized = signal<boolean>(false);
public activeVideoUrl = signal<string | null>(null); public activeVideoUrl = signal<string | null>(null);
public activeVideoSegments = signal<VideoSegment[]>([]);
public activeVideoTitle = signal<string>(''); public activeVideoTitle = signal<string>('');
public isVideoLoading = signal<boolean>(false); public isVideoLoading = signal<boolean>(false);
@@ -1000,29 +1009,33 @@ export class AgentService {
this.isVideoWindowOpen.set(true); this.isVideoWindowOpen.set(true);
this.isVideoMinimized.set(false); this.isVideoMinimized.set(false);
if (targetUrl) { this.activeVideoUrl.set(targetUrl);
this.activeVideoUrl.set(targetUrl); this.activeVideoSegments.set([]);
} else if (targetSessionId) { if (targetSessionId) {
this.isVideoLoading.set(true); this.isVideoLoading.set(true);
this.http.get<{ session_id: string; has_video: boolean; video_url: string | null }>(`/api/sessions/${targetSessionId}/video`).subscribe({ this.http.get<{ session_id: string; has_video: boolean; video_url: string | null; video_segments?: VideoSegment[] }>(`/api/sessions/${targetSessionId}/video`).subscribe({
next: (res) => { next: (res) => {
this.isVideoLoading.set(false); this.isVideoLoading.set(false);
if (res && res.has_video && res.video_url) { if (res && res.has_video && res.video_url) {
this.activeVideoUrl.set(res.video_url); this.activeVideoUrl.set(res.video_url);
this.activeVideoSegments.set(res.video_segments || []);
this.rawSessions.update((list) => this.rawSessions.update((list) =>
list.map((s) => s.session_id === targetSessionId ? { ...s, video_url: res.video_url || undefined } : s) list.map((s) => s.session_id === targetSessionId ? { ...s, video_url: res.video_url || undefined } : s)
); );
} else { } else {
this.activeVideoUrl.set(null); this.activeVideoUrl.set(null);
this.activeVideoSegments.set([]);
} }
}, },
error: () => { error: () => {
this.isVideoLoading.set(false); this.isVideoLoading.set(false);
this.activeVideoUrl.set(null); this.activeVideoUrl.set(null);
this.activeVideoSegments.set([]);
} }
}); });
} else { } else {
this.activeVideoUrl.set(null); this.activeVideoUrl.set(null);
this.activeVideoSegments.set([]);
} }
} }
+170
View File
@@ -21,8 +21,10 @@ Handles Maestro blocker detection and removal before connecting.
import base64 import base64
from io import BytesIO from io import BytesIO
import os
import re as _re import re as _re
import subprocess import subprocess
import threading
import time import time
from typing import TYPE_CHECKING from typing import TYPE_CHECKING
import xml.etree.ElementTree as ET import xml.etree.ElementTree as ET
@@ -40,6 +42,129 @@ logger = get_logger(__name__)
MAESTRO_PACKAGE = "dev.mobile.maestro" MAESTRO_PACKAGE = "dev.mobile.maestro"
AWAKE_STRATEGY_STAY_ON = "stay_on"
AWAKE_STRATEGY_WAKE_LOCK = "wake_lock"
AWAKE_STRATEGY_HEARTBEAT = "heartbeat"
AWAKE_HEARTBEAT_INTERVAL_SECONDS = 5.0
def _run_awake_adb_command(
device_id: str, args: list[str], description: str
) -> subprocess.CompletedProcess[str] | None:
"""Run one non-fatal ADB command used by the awake strategy."""
try:
result = subprocess.run(
["adb", "-s", device_id, *args],
capture_output=True,
text=True,
timeout=10,
check=False,
)
if result.returncode != 0:
detail = (result.stderr or result.stdout).strip()
logger.warning(
f"Could not {description} on {device_id}: {detail or 'ADB command failed'}"
)
return result
except Exception as e:
logger.warning(f"Could not {description} on {device_id}: {e}")
return None
def _keep_device_awake(device_id: str) -> str | None:
"""Best-effort preparation of an Android device for UI automation.
``svc power stayon true`` asks Android to keep the display on while the
device has any external power source. It avoids changing the user's
normal battery-powered screen timeout. The wake and dismiss commands
also recover a device that went to sleep before Artemis connected;
``wm dismiss-keyguard`` cannot bypass a secure PIN/password keyguard.
The primary strategy is verified through PowerManager. If it is not
actually active, Artemis tries a shell-owned bright-screen wake lock. If
that command is unavailable, the caller starts a periodic KEYCODE_WAKEUP
heartbeat. Neither fallback changes the user's screen-timeout setting.
This must never make device connection fail. Some vendor Android builds
restrict one or more shell commands, so failures are logged only.
"""
enabled = os.environ.get("ARTEMIS_KEEP_DEVICE_AWAKE", "true").strip().lower()
if enabled in {"0", "false", "no", "off"}:
return None
commands = (
("enable stay-awake while powered", ["shell", "svc", "power", "stayon", "true"]),
("wake display", ["shell", "input", "keyevent", "KEYCODE_WAKEUP"]),
("dismiss non-secure keyguard", ["shell", "wm", "dismiss-keyguard"]),
)
for description, args in commands:
_run_awake_adb_command(device_id, args, description)
power_state = _run_awake_adb_command(
device_id, ["shell", "dumpsys", "power"], "verify stay-awake state"
)
if (
power_state is not None
and power_state.returncode == 0
and _re.search(r"\bmStayOn=true\b", power_state.stdout)
):
logger.info(f"Stay-awake primary strategy verified on {device_id}")
return AWAKE_STRATEGY_STAY_ON
logger.warning(
f"Stay-awake primary strategy is not active on {device_id}; "
"trying a screen-bright wake lock"
)
_run_awake_adb_command(
device_id,
[
"shell",
"cmd",
"power",
"set-wakelock",
"acquire",
"-d",
"0",
"SCREEN_BRIGHT_WAKE_LOCK",
],
"acquire a screen-bright wake lock",
)
wake_locks = _run_awake_adb_command(
device_id,
["shell", "cmd", "power", "set-wakelock", "list"],
"verify the screen-bright wake lock",
)
if (
wake_locks is not None
and wake_locks.returncode == 0
and _re.search(r"SCREEN_BRIGHT_WAKE_LOCK:.*\bheld=true\b", wake_locks.stdout)
):
logger.info(f"Screen-bright wake lock verified on {device_id}")
return AWAKE_STRATEGY_WAKE_LOCK
logger.warning(
f"Screen-bright wake lock is unavailable on {device_id}; "
f"using a {AWAKE_HEARTBEAT_INTERVAL_SECONDS:g}-second wakeup heartbeat"
)
return AWAKE_STRATEGY_HEARTBEAT
def _release_screen_wake_lock(device_id: str) -> None:
"""Release the shell-owned screen wake lock acquired for this session."""
_run_awake_adb_command(
device_id,
[
"shell",
"cmd",
"power",
"set-wakelock",
"release",
"-d",
"0",
"SCREEN_BRIGHT_WAKE_LOCK",
],
"release the screen-bright wake lock",
)
class _CyFunctionDetectorMeta(type): class _CyFunctionDetectorMeta(type):
@@ -257,6 +382,41 @@ class UIAutomatorClient:
""" """
self._device_id = device_id self._device_id = device_id
self._device: Device | None = None self._device: Device | None = None
self._awake_strategy: str | None = None
self._heartbeat_stop_event: threading.Event | None = None
self._heartbeat_thread: threading.Thread | None = None
def _start_awake_heartbeat(self) -> None:
"""Start a daemon heartbeat that safely refreshes Android user activity."""
if self._heartbeat_thread is not None and self._heartbeat_thread.is_alive():
return
stop_event = threading.Event()
self._heartbeat_stop_event = stop_event
def heartbeat() -> None:
while not stop_event.wait(AWAKE_HEARTBEAT_INTERVAL_SECONDS):
_run_awake_adb_command(
self._device_id,
["shell", "input", "keyevent", "KEYCODE_WAKEUP"],
"send the stay-awake heartbeat",
)
self._heartbeat_thread = threading.Thread(
target=heartbeat,
name=f"artemis-awake-{self._device_id}",
daemon=True,
)
self._heartbeat_thread.start()
def _stop_awake_heartbeat(self) -> None:
"""Stop the session heartbeat without delaying shutdown."""
if self._heartbeat_stop_event is not None:
self._heartbeat_stop_event.set()
if self._heartbeat_thread is not None:
self._heartbeat_thread.join(timeout=2.0)
self._heartbeat_stop_event = None
self._heartbeat_thread = None
def _ensure_connected(self) -> "Device": def _ensure_connected(self) -> "Device":
"""Ensure connection to the device, handling Maestro blocker. """Ensure connection to the device, handling Maestro blocker.
@@ -276,6 +436,12 @@ class UIAutomatorClient:
# Ensure Maestro is not blocking us # Ensure Maestro is not blocking us
_ensure_maestro_not_installed(self._device_id) _ensure_maestro_not_installed(self._device_id)
# Configure this automation session before UIAutomator2 reads the screen.
if self._awake_strategy is None:
self._awake_strategy = _keep_device_awake(self._device_id)
if self._awake_strategy == AWAKE_STRATEGY_HEARTBEAT:
self._start_awake_heartbeat()
# Connect to device # Connect to device
logger.info(f"Connecting UIAutomator2 to device: {self._device_id}") logger.info(f"Connecting UIAutomator2 to device: {self._device_id}")
self._device = u2.connect(self._device_id) self._device = u2.connect(self._device_id)
@@ -400,6 +566,10 @@ class UIAutomatorClient:
def disconnect(self) -> None: def disconnect(self) -> None:
"""Disconnect from the device.""" """Disconnect from the device."""
self._stop_awake_heartbeat()
if self._awake_strategy == AWAKE_STRATEGY_WAKE_LOCK:
_release_screen_wake_lock(self._device_id)
self._awake_strategy = None
self._device = None self._device = None
logger.info("UIAutomator2 client disconnected") logger.info("UIAutomator2 client disconnected")
+18 -42
View File
@@ -45,10 +45,12 @@ from artemis.utils.video import (
DEFAULT_MAX_DURATION_SECONDS, DEFAULT_MAX_DURATION_SECONDS,
RecordingSession, RecordingSession,
VideoRecordingResult, VideoRecordingResult,
build_scrcpy_record_command,
cleanup_video_segments, cleanup_video_segments,
concatenate_videos, concatenate_videos,
get_active_session, get_active_session,
has_active_session, has_active_session,
normalize_recording_to_mp4,
remove_active_session, remove_active_session,
set_active_session, set_active_session,
trim_video, trim_video,
@@ -561,6 +563,8 @@ class AndroidDeviceController(MobileDeviceController):
start_time=time.time(), start_time=time.time(),
data_engine_start_time=self.data_engine_start_time, data_engine_start_time=self.data_engine_start_time,
local_video_path=local_video_path, local_video_path=local_video_path,
capture_width=self.device_width,
capture_height=self.device_height,
) )
# Persist to local database if Data Engine is active # Persist to local database if Data Engine is active
@@ -578,18 +582,9 @@ class AndroidDeviceController(MobileDeviceController):
logger.error(f"Failed to persist video recording start to DB: {db_err}") logger.error(f"Failed to persist video recording start to DB: {db_err}")
# Start scrcpy in background # Start scrcpy in background
cmd = [ cmd = build_scrcpy_record_command(
"scrcpy", "scrcpy", self.device_id, local_video_path, lock_capture_orientation=False
"--serial", )
self.device_id,
"--no-window",
"--record",
str(local_video_path),
"--record-format",
"mkv",
"--video-bit-rate",
"2M",
]
process = await asyncio.create_subprocess_exec( process = await asyncio.create_subprocess_exec(
*cmd, *cmd,
@@ -760,18 +755,9 @@ class AndroidDeviceController(MobileDeviceController):
output_dir = mkv_path.parent output_dir = mkv_path.parent
new_video_path = output_dir / f"recording_{session.android_segment_index}.mkv" new_video_path = output_dir / f"recording_{session.android_segment_index}.mkv"
cmd = [ cmd = build_scrcpy_record_command(
"scrcpy", "scrcpy", self.device_id, new_video_path, lock_capture_orientation=False
"--serial", )
self.device_id,
"--no-window",
"--record",
str(new_video_path),
"--record-format",
"mkv",
"--video-bit-rate",
"2M",
]
process = await asyncio.create_subprocess_exec( process = await asyncio.create_subprocess_exec(
*cmd, *cmd,
@@ -892,21 +878,11 @@ class AndroidDeviceController(MobileDeviceController):
) )
async def _convert_mkv_to_mp4(self, mkv_path: Path, mp4_path: Path) -> bool: async def _convert_mkv_to_mp4(self, mkv_path: Path, mp4_path: Path) -> bool:
"""Convert MKV to MP4 using ffmpeg.""" """Normalize MKV into a fixed-size, browser-safe MP4."""
try: session = get_active_session(self.device_id)
process = await asyncio.create_subprocess_exec( return await normalize_recording_to_mp4(
"ffmpeg", mkv_path,
"-y", mp4_path,
"-i", session.capture_width if session else self.device_width,
str(mkv_path), session.capture_height if session else self.device_height,
"-c", )
"copy",
str(mp4_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
await process.wait()
return mp4_path.exists()
except Exception as e:
logger.error(f"Failed to convert MKV to MP4: {e}")
return False
+176 -219
View File
@@ -15,6 +15,8 @@
import asyncio import asyncio
import os import os
from pathlib import Path from pathlib import Path
import signal
import subprocess
import tempfile import tempfile
import time import time
from typing import Any from typing import Any
@@ -33,17 +35,23 @@ from artemis.controllers.types import (
) )
from artemis.utils.logger import get_logger from artemis.utils.logger import get_logger
from artemis.utils.video import ( from artemis.utils.video import (
ANDROID_RECORDING_SEGMENT_SECONDS,
DEFAULT_MAX_DURATION_SECONDS, DEFAULT_MAX_DURATION_SECONDS,
RecordingSession, RecordingSession,
VideoRecordingResult, VideoRecordingResult,
build_scrcpy_record_command,
cleanup_video_segments, cleanup_video_segments,
concatenate_videos, concatenate_videos,
get_android_display_state,
get_active_session, get_active_session,
get_ffmpeg_path,
has_active_session, has_active_session,
normalize_recording_to_mp4,
remux_recording_to_mp4,
render_timeline_clip,
remove_active_session, remove_active_session,
set_active_session, set_active_session,
trim_video, trim_video,
write_recording_manifest,
) )
logger = get_logger(__name__) logger = get_logger(__name__)
@@ -63,6 +71,29 @@ class UnifiedMobileController:
def controller(self) -> Any: def controller(self) -> Any:
return self._driver return self._driver
@staticmethod
async def _spawn_scrcpy(command: list[str]) -> asyncio.subprocess.Process:
kwargs: dict[str, Any] = {
"stdout": asyncio.subprocess.PIPE,
"stderr": asyncio.subprocess.PIPE,
}
if os.name == "nt":
kwargs["creationflags"] = subprocess.CREATE_NEW_PROCESS_GROUP
return await asyncio.create_subprocess_exec(*command, **kwargs)
@staticmethod
async def _stop_scrcpy(process: asyncio.subprocess.Process) -> None:
"""Ask scrcpy to flush its recorder before falling back to termination."""
if process.returncode is not None:
return
try:
process.send_signal(signal.CTRL_BREAK_EVENT if os.name == "nt" else signal.SIGINT)
await asyncio.wait_for(process.wait(), timeout=8.0)
except (ProcessLookupError, TimeoutError):
if process.returncode is None:
process.terminate()
await asyncio.wait_for(process.wait(), timeout=5.0)
async def tap_at( async def tap_at(
self, self,
x: int, x: int,
@@ -301,68 +332,6 @@ class UnifiedMobileController:
message="Recording file not found", message="Recording file not found",
) )
# Check if scrcpy process terminated and restart if needed (synchronous fallback)
crashed = False
restart_message = ""
if session.process and session.process.returncode is not None:
logger.warning(f"scrcpy process for {device_id} has terminated unexpectedly")
crashed = True
if mkv_path.exists() and mkv_path not in session.android_video_segments:
session.android_video_segments.append(mkv_path)
try:
session.android_segment_index += 1
output_dir = mkv_path.parent
new_video_path = output_dir / f"recording_{session.android_segment_index}.mkv"
cmd = [
"scrcpy",
"--serial",
device_id,
"--no-window",
"--record",
str(new_video_path),
"--record-format",
"mkv",
"--video-bit-rate",
"2M",
]
process = await asyncio.create_subprocess_exec(
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
await asyncio.sleep(0.8)
if process.returncode is None:
session.process = process
session.local_video_path = new_video_path
mkv_path = new_video_path
restart_message = "Successfully restarted scrcpy."
else:
stderr = await process.stderr.read()
restart_message = f"Failed to restart scrcpy: {stderr.decode()}"
except Exception as e:
restart_message = f"Failed to restart scrcpy: {e}"
# Determine source file to trim from (support multi-segment seamless trimming)
all_segments = [p for p in session.android_video_segments if p.exists()]
if mkv_path.exists() and mkv_path not in all_segments:
all_segments.append(mkv_path)
if not all_segments:
return VideoRecordingResult(
success=False,
message="No recording segments found on disk",
)
temp_combined_file = None
if len(all_segments) > 1:
temp_combined_file = mkv_path.parent / "temp_combined_active.mkv"
concat_ok = await concatenate_videos(all_segments, temp_combined_file)
trim_source = (
temp_combined_file if (concat_ok and temp_combined_file.exists()) else mkv_path
)
else:
trim_source = all_segments[0]
current_time = time.time() current_time = time.time()
video_duration = current_time - session.start_time video_duration = current_time - session.start_time
@@ -415,8 +384,18 @@ class UnifiedMobileController:
success = False success = False
for attempt in range(3): for attempt in range(3):
success = await trim_video( timeline_segments = [dict(record) for record in session.android_segment_records]
trim_source, active_start = session.android_segment_started_at or session.start_time
if mkv_path.exists():
timeline_segments.append(
{
"path": mkv_path,
"start": max(0.0, active_start - session.start_time),
"end": max(0.0, current_time - session.start_time),
}
)
success = await render_timeline_clip(
timeline_segments,
video_start_relative_time, video_start_relative_time,
video_end_relative_time, video_end_relative_time,
trim_output_path, trim_output_path,
@@ -426,13 +405,6 @@ class UnifiedMobileController:
logger.warning(f"ffmpeg trim attempt {attempt + 1} failed, retrying in 0.5s...") logger.warning(f"ffmpeg trim attempt {attempt + 1} failed, retrying in 0.5s...")
await asyncio.sleep(0.5) await asyncio.sleep(0.5)
# Cleanup temp combined MKV if created
if temp_combined_file and temp_combined_file.exists():
try:
temp_combined_file.unlink()
except Exception:
pass
if not success or not trim_output_path.exists(): if not success or not trim_output_path.exists():
return VideoRecordingResult( return VideoRecordingResult(
success=False, success=False,
@@ -445,11 +417,6 @@ class UnifiedMobileController:
duration = video_end_relative_time - video_start_relative_time duration = video_end_relative_time - video_start_relative_time
message = f"Video segment retrieved for range {actual_start:.1f}s to {actual_end:.1f}s" message = f"Video segment retrieved for range {actual_start:.1f}s to {actual_end:.1f}s"
if crashed:
message += (
f". Warning: scrcpy crashed. Data loss may have occurred. {restart_message}"
)
res = VideoRecordingResult( res = VideoRecordingResult(
success=True, success=True,
message=message, message=message,
@@ -469,61 +436,98 @@ class UnifiedMobileController:
message=f"Failed to get video segment: {e}", message=f"Failed to get video segment: {e}",
) )
async def _recording_watchdog(self, device_id: str) -> None: @staticmethod
"""Background watchdog to automatically detect crashes and recover scrcpy recording.""" def _segment_mp4_path(source_path: Path, index: int) -> Path:
while True: return source_path.parent / ("recording.mp4" if index == 0 else f"recording_{index:03d}.mp4")
session = get_active_session(device_id)
if not session or not session.is_active: @staticmethod
break async def _remux_segment_record(record: dict[str, Any]) -> bool:
proc = session.process source_path = Path(record["path"])
if proc is None: output_path = Path(record["output_path"])
break success = await remux_recording_to_mp4(source_path, output_path)
if success:
record["path"] = output_path
try: try:
ret = await proc.wait() source_path.unlink()
except OSError:
pass
return success
def _finalize_current_segment(self, session: RecordingSession, end_time: float) -> None:
source_path = session.local_video_path
if not source_path or not source_path.exists():
return
if any(Path(record["path"]) == source_path for record in session.android_segment_records):
return
start_time = session.android_segment_started_at or session.start_time
output_path = self._segment_mp4_path(source_path, session.android_segment_index)
session.android_video_segments.append(source_path)
session.android_segment_records.append(
{
"path": source_path,
"output_path": output_path,
"start": max(0.0, start_time - session.start_time),
"end": max(0.0, end_time - session.start_time),
"rotation": session.android_rotation,
"conversion_scheduled": True,
}
)
session.android_conversion_tasks.append(
asyncio.create_task(self._remux_segment_record(session.android_segment_records[-1]))
)
async def _start_next_recording_segment(
self, session: RecordingSession, display_state: tuple[int, int, int] | None
) -> bool:
session.android_segment_index += 1
output_dir = session.local_video_path.parent
new_video_path = output_dir / f"recording_{session.android_segment_index:03d}.mkv"
process = await self._spawn_scrcpy(
build_scrcpy_record_command("scrcpy", session.device_id, new_video_path)
)
await asyncio.sleep(0.8)
if process.returncode is not None:
stderr = await process.stderr.read()
logger.error(f"Failed to start scrcpy segment: {stderr.decode(errors='replace')}")
return False
session.process = process
session.local_video_path = new_video_path
session.android_segment_started_at = time.time()
if display_state:
session.android_rotation, session.capture_width, session.capture_height = display_state
return True
async def _recording_watchdog(self, device_id: str) -> None:
"""Roll fixed-orientation segments and recover scrcpy crashes."""
try:
while True:
await asyncio.sleep(0.5)
session = get_active_session(device_id) session = get_active_session(device_id)
if not session or not session.is_active: if not session or not session.is_active or not session.process:
break return
now = time.time()
logger.warning( display_state = await get_android_display_state(device_id)
f"scrcpy recording process for {device_id} terminated unexpectedly (code {ret}). " rotated = bool(
f"Auto-restarting recording segment {session.android_segment_index + 1}..." display_state
and session.android_rotation is not None
and display_state[0] != session.android_rotation
) )
if session.local_video_path and session.local_video_path.exists(): segment_age = now - (session.android_segment_started_at or session.start_time)
if session.local_video_path not in session.android_video_segments: crashed = session.process.returncode is not None
session.android_video_segments.append(session.local_video_path) if not crashed and not rotated and segment_age < ANDROID_RECORDING_SEGMENT_SECONDS:
continue
session.android_segment_index += 1 if not crashed:
output_dir = session.local_video_path.parent await self._stop_scrcpy(session.process)
new_video_path = output_dir / f"recording_{session.android_segment_index}.mkv" self._finalize_current_segment(session, time.time())
cmd = [ reason = "rotation" if rotated else "time limit" if not crashed else "recorder crash"
"scrcpy", logger.info(f"Rolling screen recording segment after {reason}")
"--serial", if not await self._start_next_recording_segment(session, display_state):
device_id, return
"--no-window", except asyncio.CancelledError:
"--record", return
str(new_video_path), except Exception as e:
"--record-format", logger.error(f"Error in recording supervisor for {device_id}: {e}")
"mkv",
"--video-bit-rate",
"2M",
]
new_proc = await asyncio.create_subprocess_exec(
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
session.process = new_proc
session.local_video_path = new_video_path
await asyncio.sleep(0.8)
if new_proc.returncode is not None:
stderr = await new_proc.stderr.read()
logger.error(f"Failed to auto-restart scrcpy recording: {stderr.decode()}")
break
except asyncio.CancelledError:
break
except Exception as e:
logger.error(f"Error in recording watchdog for {device_id}: {e}")
break
async def start_video_recording( async def start_video_recording(
self, self,
@@ -561,6 +565,7 @@ class UnifiedMobileController:
local_video_path = output_dir / "recording.mkv" local_video_path = output_dir / "recording.mkv"
video_id = uuid4() video_id = uuid4()
start_time = time.time() start_time = time.time()
display_state = await get_android_display_state(device_id)
data_engine_start_time = ( data_engine_start_time = (
self.ctx.data_engine.session_start_time self.ctx.data_engine.session_start_time
if (self.ctx and self.ctx.data_engine) if (self.ctx and self.ctx.data_engine)
@@ -573,8 +578,14 @@ class UnifiedMobileController:
start_time=start_time, start_time=start_time,
data_engine_start_time=data_engine_start_time, data_engine_start_time=data_engine_start_time,
local_video_path=local_video_path, local_video_path=local_video_path,
capture_width=getattr(getattr(self.ctx, "device", None), "device_width", None),
capture_height=getattr(getattr(self.ctx, "device", None), "device_height", None),
android_rotation=display_state[0] if display_state else None,
android_segment_started_at=start_time,
is_active=True, is_active=True,
) )
if display_state:
session.capture_width, session.capture_height = display_state[1:]
# Persist to local database if Data Engine is active # Persist to local database if Data Engine is active
if self.ctx and self.ctx.data_engine: if self.ctx and self.ctx.data_engine:
@@ -586,24 +597,9 @@ class UnifiedMobileController:
) )
# Start scrcpy in background # Start scrcpy in background
cmd = [ cmd = build_scrcpy_record_command("scrcpy", device_id, local_video_path)
"scrcpy",
"--serial",
device_id,
"--no-window",
"--record",
str(local_video_path),
"--record-format",
"mkv",
"--video-bit-rate",
"2M",
]
process = await asyncio.create_subprocess_exec( process = await self._spawn_scrcpy(cmd)
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
session.process = process session.process = process
set_active_session(device_id, session) set_active_session(device_id, session)
@@ -667,13 +663,13 @@ class UnifiedMobileController:
session.is_active = False session.is_active = False
if session.watchdog_task and not session.watchdog_task.done(): if session.watchdog_task and not session.watchdog_task.done():
session.watchdog_task.cancel() session.watchdog_task.cancel()
await session.watchdog_task
try: try:
process = session.process process = session.process
if process is not None: if process is not None:
try: try:
process.terminate() await self._stop_scrcpy(process)
await asyncio.wait_for(process.wait(), timeout=5.0)
except Exception as proc_e: except Exception as proc_e:
logger.warning(f"Error terminating scrcpy process: {proc_e}") logger.warning(f"Error terminating scrcpy process: {proc_e}")
@@ -685,27 +681,26 @@ class UnifiedMobileController:
message="Recording file not found on disk", message="Recording file not found on disk",
) )
# Concatenate previous segments if scrcpy crashed and restarted self._finalize_current_segment(session, time.time())
if session.android_video_segments: for record in session.android_segment_records:
all_segments = session.android_video_segments if not record.get("conversion_scheduled"):
if output_path.exists() and output_path not in all_segments: record["conversion_scheduled"] = True
all_segments.append(output_path) session.android_conversion_tasks.append(
concatenated_mkv = output_path.parent / "full_recording.mkv" asyncio.create_task(
if await concatenate_videos(all_segments, concatenated_mkv): self._remux_segment_record(record)
cleanup_video_segments(all_segments) )
output_path = concatenated_mkv )
if session.android_conversion_tasks:
# Convert MKV to web-streamable MP4 await asyncio.gather(*session.android_conversion_tasks, return_exceptions=True)
mp4_path = ( mp4_paths = [
output_path.parent / "recording.mp4" Path(record["output_path"])
if output_path.name == "full_recording.mkv" for record in session.android_segment_records
else output_path.with_suffix(".mp4") if Path(record["output_path"]).exists()
) ]
success = await self._convert_mkv_to_mp4(output_path, mp4_path) await write_recording_manifest(output_path.parent, mp4_paths)
remove_active_session(device_id) remove_active_session(device_id)
final_video_path = mp4_paths[0] if mp4_paths else output_path
final_video_path = mp4_path if (success and mp4_path.exists()) else output_path
# Persist update to local database if Data Engine is active # Persist update to local database if Data Engine is active
if self.ctx and self.ctx.data_engine: if self.ctx and self.ctx.data_engine:
@@ -717,16 +712,11 @@ class UnifiedMobileController:
end_time=time.time(), end_time=time.time(),
) )
if success and mp4_path.exists(): if mp4_paths:
try:
if output_path != mp4_path and output_path.exists():
output_path.unlink()
except Exception:
pass
return VideoRecordingResult( return VideoRecordingResult(
success=True, success=True,
message=f"Recording stopped, saved to {mp4_path}", message=f"Recording stopped, saved {len(mp4_paths)} orientation-safe segments",
video_path=mp4_path, video_path=mp4_paths[0],
) )
else: else:
return VideoRecordingResult( return VideoRecordingResult(
@@ -744,52 +734,19 @@ class UnifiedMobileController:
) )
async def _convert_mkv_to_mp4(self, mkv_path: Path, mp4_path: Path) -> bool: async def _convert_mkv_to_mp4(self, mkv_path: Path, mp4_path: Path) -> bool:
"""Convert MKV to MP4 using ffmpeg with faststart for web streaming.""" """Normalize MKV into a fixed-size, browser-safe MP4.
if not mkv_path.exists():
return False
try:
# 1. Try fast stream copy with +faststart
process = await asyncio.create_subprocess_exec(
get_ffmpeg_path(),
"-y",
"-i",
str(mkv_path),
"-c",
"copy",
"-movflags",
"+faststart",
str(mp4_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
await process.wait()
if mp4_path.exists() and mp4_path.stat().st_size > 0:
return True
# 2. Fallback to re-encoding if stream copy fails Stream-copying a scrcpy H.264 track is unsafe because historical or
logger.warning("ffmpeg fast copy failed, re-encoding MKV to MP4...") recovered recordings may contain resolution changes. Re-encoding onto
process = await asyncio.create_subprocess_exec( the initial capture canvas guarantees one coded size for the full MP4.
get_ffmpeg_path(), """
"-y", session = get_active_session(self._get_device_id())
"-i", return await normalize_recording_to_mp4(
str(mkv_path), mkv_path,
"-c:v", mp4_path,
"libx264", session.capture_width if session else None,
"-preset", session.capture_height if session else None,
"fast", )
"-pix_fmt",
"yuv420p",
"-movflags",
"+faststart",
str(mp4_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
await process.wait()
return mp4_path.exists() and mp4_path.stat().st_size > 0
except Exception as e:
logger.error(f"Failed to convert MKV to MP4: {e}")
return False
async def cleanup(self) -> None: async def cleanup(self) -> None:
await self._driver.disconnect() await self._driver.disconnect()
+7 -10
View File
@@ -28,6 +28,7 @@ from artemis.clients.ui_automator_client import (
from artemis.config.paths import get_temp_dir from artemis.config.paths import get_temp_dir
from artemis.drivers.base import BaseDeviceDriver, KeyCode, ScreenData, SwipeDirection from artemis.drivers.base import BaseDeviceDriver, KeyCode, ScreenData, SwipeDirection
from artemis.toolchain import find_ffmpeg, find_scrcpy from artemis.toolchain import find_ffmpeg, find_scrcpy
from artemis.utils.video import build_scrcpy_record_command
from artemis.utils.ui_filter import filter_ui_hierarchy from artemis.utils.ui_filter import filter_ui_hierarchy
from artemis.utils.logger import get_logger from artemis.utils.logger import get_logger
@@ -114,6 +115,8 @@ class AndroidAdbDriver(BaseDeviceDriver):
"""Clean up ongoing recordings and tunnels.""" """Clean up ongoing recordings and tunnels."""
if self._recording_process: if self._recording_process:
await self.stop_video_recording() await self.stop_video_recording()
if self._ui_adb_client and hasattr(self._ui_adb_client, "disconnect"):
await asyncio.to_thread(self._ui_adb_client.disconnect)
async def get_screen_data(self, skip_settling: bool = False) -> ScreenData: async def get_screen_data(self, skip_settling: bool = False) -> ScreenData:
"""Captures screenshot and XML hierarchy concurrently.""" """Captures screenshot and XML hierarchy concurrently."""
@@ -442,18 +445,12 @@ class AndroidAdbDriver(BaseDeviceDriver):
logger.info(f"Starting scrcpy video recording to {self._recording_mkv_path}...") logger.info(f"Starting scrcpy video recording to {self._recording_mkv_path}...")
try: try:
scrcpy_bin = find_scrcpy() scrcpy_bin = find_scrcpy()
cmd = [ cmd = build_scrcpy_record_command(
scrcpy_bin, scrcpy_bin,
"--serial",
self.device_id, self.device_id,
"--no-window", self._recording_mkv_path,
"--record", lock_capture_orientation=False,
str(self._recording_mkv_path), )
"--record-format",
"mkv",
"--video-bit-rate",
"2M",
]
self._scrcpy_process = await asyncio.create_subprocess_exec( self._scrcpy_process = await asyncio.create_subprocess_exec(
*cmd, *cmd,
stdout=asyncio.subprocess.PIPE, stdout=asyncio.subprocess.PIPE,
+331
View File
@@ -19,8 +19,10 @@ Provides shared types and utilities for video recording across platforms.
import asyncio import asyncio
import importlib.util import importlib.util
import json
from pathlib import Path from pathlib import Path
import platform import platform
import re
import shutil import shutil
import subprocess import subprocess
from typing import Any from typing import Any
@@ -34,6 +36,7 @@ from artemis.utils.logger import get_logger
logger = get_logger(__name__) logger = get_logger(__name__)
DEFAULT_MAX_DURATION_SECONDS = 900 # 15 minutes DEFAULT_MAX_DURATION_SECONDS = 900 # 15 minutes
ANDROID_RECORDING_SEGMENT_SECONDS = 1800
VIDEO_READY_DELAY_SECONDS = 1 VIDEO_READY_DELAY_SECONDS = 1
ANDROID_DEVICE_VIDEO_PATH = "/sdcard/screen_recording.mp4" ANDROID_DEVICE_VIDEO_PATH = "/sdcard/screen_recording.mp4"
ANDROID_MAX_RECORDING_DURATION_SECONDS = 180 # Android screenrecord limit ANDROID_MAX_RECORDING_DURATION_SECONDS = 180 # Android screenrecord limit
@@ -44,6 +47,37 @@ MAX_VIDEO_SIZE_MB = 500
MAX_VIDEO_SIZE_BYTES = MAX_VIDEO_SIZE_MB * 1024 * 1024 MAX_VIDEO_SIZE_BYTES = MAX_VIDEO_SIZE_MB * 1024 * 1024
def build_scrcpy_record_command(
scrcpy_executable: str,
device_id: str,
output_path: Path,
video_bit_rate: str = "2M",
lock_capture_orientation: bool = True,
) -> list[str]:
"""Build the shared scrcpy command used by all recording paths.
Each recording segment locks the orientation present when scrcpy starts.
The recording supervisor starts a new segment when Android rotates, so a
segment never contains multiple coded sizes while still displaying the app
in its natural orientation.
"""
command = [
scrcpy_executable,
"--serial",
device_id,
"--no-window",
"--record",
str(output_path),
"--record-format",
"mkv",
"--video-bit-rate",
video_bit_rate,
]
if lock_capture_orientation:
command.append("--capture-orientation=@")
return command
class _CyFunctionDetectorMeta(type): class _CyFunctionDetectorMeta(type):
def __instancecheck__(self, instance): def __instancecheck__(self, instance):
name = type(instance).__name__ name = type(instance).__name__
@@ -73,11 +107,17 @@ class RecordingSession(BaseModel):
process: Any = None process: Any = None
data_engine_start_time: float | None = None data_engine_start_time: float | None = None
local_video_path: Path | None = None local_video_path: Path | None = None
capture_width: int | None = None
capture_height: int | None = None
android_device_path: str = ANDROID_DEVICE_VIDEO_PATH android_device_path: str = ANDROID_DEVICE_VIDEO_PATH
android_video_segments: list[Path] = [] android_video_segments: list[Path] = []
android_segment_index: int = 0 android_segment_index: int = 0
android_restart_task: asyncio.Task | None = None android_restart_task: asyncio.Task | None = None
watchdog_task: asyncio.Task | None = None watchdog_task: asyncio.Task | None = None
android_rotation: int | None = None
android_segment_started_at: float | None = None
android_segment_records: list[dict[str, Any]] = []
android_conversion_tasks: list[asyncio.Task] = []
is_active: bool = True is_active: bool = True
errors: list[str] = [] errors: list[str] = []
@@ -138,6 +178,297 @@ def get_ffmpeg_path() -> str:
return "ffmpeg" return "ffmpeg"
def get_ffprobe_path() -> str:
"""Get ffprobe for lightweight segment metadata inspection."""
return shutil.which("ffprobe") or "ffprobe"
async def normalize_recording_to_mp4(
source_path: Path,
output_path: Path,
capture_width: int | None,
capture_height: int | None,
) -> bool:
"""Re-encode a recording onto one fixed canvas for browser playback."""
if not source_path.exists():
return False
width = max(2, int(capture_width or 1080)) // 2 * 2
height = max(2, int(capture_height or 1920)) // 2 * 2
video_filter = (
f"scale={width}:{height}:force_original_aspect_ratio=decrease:"
f"force_divisible_by=2,pad={width}:{height}:"
"(ow-iw)/2:(oh-ih)/2:color=black,setsar=1"
)
try:
process = await asyncio.create_subprocess_exec(
get_ffmpeg_path(),
"-y",
"-fflags",
"+genpts",
"-i",
str(source_path),
"-map",
"0:v:0",
"-map",
"0:a?",
"-vf",
video_filter,
"-r",
"30",
"-fps_mode",
"cfr",
"-c:v",
"libx264",
"-preset",
"veryfast",
"-crf",
"23",
"-pix_fmt",
"yuv420p",
"-c:a",
"aac",
"-b:a",
"128k",
"-movflags",
"+faststart",
str(output_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
_stdout, stderr = await process.communicate()
if process.returncode == 0 and output_path.exists() and output_path.stat().st_size > 0:
return True
error_detail = stderr.decode(errors="replace")[-4000:]
logger.error(
f"Failed to normalize recording to MP4 (code {process.returncode}): {error_detail}"
)
if output_path.exists():
output_path.unlink()
return False
except Exception as e:
logger.error(f"Failed to normalize recording to MP4: {e}")
if output_path.exists():
output_path.unlink()
return False
async def remux_recording_to_mp4(source_path: Path, output_path: Path) -> bool:
"""Move a fixed-dimension recording segment into MP4 without re-encoding."""
if not source_path.exists() or source_path.stat().st_size == 0:
return False
try:
process = await asyncio.create_subprocess_exec(
get_ffmpeg_path(),
"-y",
"-fflags",
"+genpts",
"-i",
str(source_path),
"-map",
"0:v:0",
"-map",
"0:a?",
"-c",
"copy",
"-movflags",
"+faststart",
str(output_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
_stdout, stderr = await process.communicate()
valid = process.returncode == 0 and output_path.exists() and output_path.stat().st_size > 0
if valid:
return True
logger.error(
f"Failed to remux recording segment (code {process.returncode}): "
f"{stderr.decode(errors='replace')[-2000:]}"
)
except Exception as e:
logger.error(f"Failed to remux recording segment: {e}")
if output_path.exists():
output_path.unlink()
return False
async def probe_video_segment(video_path: Path) -> dict[str, float | int]:
"""Read duration and coded dimensions for a finalized segment."""
process = await asyncio.create_subprocess_exec(
get_ffprobe_path(),
"-v",
"error",
"-select_streams",
"v:0",
"-show_entries",
"stream=width,height:format=duration",
"-of",
"json",
str(video_path),
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
stdout, _stderr = await process.communicate()
if process.returncode != 0:
return {}
try:
payload = json.loads(stdout)
stream = (payload.get("streams") or [{}])[0]
return {
"duration": float((payload.get("format") or {}).get("duration") or 0),
"width": int(stream.get("width") or 0),
"height": int(stream.get("height") or 0),
}
except (TypeError, ValueError, json.JSONDecodeError):
return {}
async def get_android_display_state(device_id: str) -> tuple[int, int, int] | None:
"""Return Android's current (rotation, width, height)."""
try:
process = await asyncio.create_subprocess_exec(
"adb",
"-s",
device_id,
"shell",
"dumpsys",
"window",
"displays",
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.DEVNULL,
)
stdout, _stderr = await asyncio.wait_for(process.communicate(), timeout=3.0)
text = stdout.decode(errors="replace")
rotation_match = re.search(r"\bmRotation=(\d+)", text)
size_match = re.search(r"\bcur=(\d+)x(\d+)", text)
if rotation_match and size_match:
return (
int(rotation_match.group(1)),
int(size_match.group(1)),
int(size_match.group(2)),
)
except (OSError, TimeoutError):
pass
return None
async def write_recording_manifest(output_dir: Path, mp4_paths: list[Path]) -> Path | None:
"""Write the browser playlist used for orientation-aware playback."""
segments = []
timeline = 0.0
for path in mp4_paths:
if not path.exists():
continue
metadata = await probe_video_segment(path)
duration = float(metadata.get("duration", 0))
segments.append(
{
"file": path.name,
"start": round(timeline, 3),
"duration": round(duration, 3),
"width": int(metadata.get("width", 0)),
"height": int(metadata.get("height", 0)),
}
)
timeline += duration
if not segments:
return None
manifest_path = output_dir / "recording.json"
manifest_path.write_text(
json.dumps(
{"version": 1, "duration": round(timeline, 3), "segments": segments},
indent=2,
),
encoding="utf-8",
)
return manifest_path
async def render_timeline_clip(
segments: list[dict[str, Any]],
start_time: float,
end_time: float,
output_path: Path,
canvas_width: int = 720,
canvas_height: int = 1280,
) -> bool:
"""Render one continuous analyzer clip from orientation-aware segments.
Only the requested time window is decoded. This keeps the video analyzer's
single-MP4 contract without ever re-encoding the full recording.
"""
overlaps = []
for segment in segments:
path = Path(segment["path"])
seg_start = float(segment["start"])
seg_end = float(segment["end"])
overlap_start = max(start_time, seg_start)
overlap_end = min(end_time, seg_end)
if path.exists() and overlap_end > overlap_start:
overlaps.append((path, overlap_start - seg_start, overlap_end - seg_start))
if not overlaps:
return False
command = [get_ffmpeg_path(), "-y", "-fflags", "+genpts"]
for path, _start, _end in overlaps:
command.extend(["-i", str(path)])
filter_parts = []
labels = []
for index, (_path, local_start, local_end) in enumerate(overlaps):
label = f"v{index}"
filter_parts.append(
f"[{index}:v:0]trim=start={local_start:.6f}:end={local_end:.6f},"
"setpts=PTS-STARTPTS,"
f"scale={canvas_width}:{canvas_height}:force_original_aspect_ratio=decrease:"
"force_divisible_by=2,"
f"pad={canvas_width}:{canvas_height}:(ow-iw)/2:(oh-ih)/2:color=black,"
f"setsar=1,fps=15,format=yuv420p[{label}]"
)
labels.append(f"[{label}]")
if len(labels) == 1:
filter_parts.append(f"{labels[0]}null[outv]")
else:
filter_parts.append(f"{''.join(labels)}concat=n={len(labels)}:v=1:a=0[outv]")
command.extend(
[
"-filter_complex",
";".join(filter_parts),
"-map",
"[outv]",
"-c:v",
"libx264",
"-preset",
"veryfast",
"-crf",
"23",
"-movflags",
"+faststart",
str(output_path),
]
)
try:
process = await asyncio.create_subprocess_exec(
*command,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
_stdout, stderr = await process.communicate()
if process.returncode == 0 and output_path.exists() and output_path.stat().st_size > 0:
return True
logger.error(
f"Failed to render analyzer timeline clip (code {process.returncode}): "
f"{stderr.decode(errors='replace')[-3000:]}"
)
except Exception as e:
logger.error(f"Failed to render analyzer timeline clip: {e}")
if output_path.exists():
output_path.unlink()
return False
def is_scrcpy_installed() -> bool: def is_scrcpy_installed() -> bool:
"""Check if scrcpy is available in the system PATH.""" """Check if scrcpy is available in the system PATH."""
@@ -0,0 +1,165 @@
from unittest.mock import MagicMock, patch
import pytest
from artemis.clients.ui_automator_client import (
AWAKE_STRATEGY_HEARTBEAT,
AWAKE_STRATEGY_STAY_ON,
AWAKE_STRATEGY_WAKE_LOCK,
UIAutomatorClient,
_keep_device_awake,
)
from artemis.drivers.android.adb_driver import AndroidAdbDriver
def completed(stdout="", returncode=0, stderr=""):
return MagicMock(returncode=returncode, stdout=stdout, stderr=stderr)
@patch("artemis.clients.ui_automator_client.subprocess.run")
def test_keep_device_awake_uses_verified_primary_strategy(mock_run):
mock_run.side_effect = [
completed(),
completed(),
completed(),
completed(" mStayOn=true\n"),
]
strategy = _keep_device_awake("device-123")
assert strategy == AWAKE_STRATEGY_STAY_ON
assert mock_run.call_count == 4
@patch("artemis.clients.ui_automator_client.subprocess.run")
def test_keep_device_awake_falls_back_to_verified_screen_wake_lock(mock_run):
mock_run.side_effect = [
completed(),
completed(),
completed(),
completed(" mStayOn=false\n"),
completed("WakeLock{abc held=true, refCount=1}\n"),
completed(
"Display 0, wakelock type: SCREEN_BRIGHT_WAKE_LOCK: "
"WakeLock{abc held=true, refCount=1}\n"
),
]
strategy = _keep_device_awake("device-123")
assert strategy == AWAKE_STRATEGY_WAKE_LOCK
assert mock_run.call_args_list[-2].args[0] == [
"adb",
"-s",
"device-123",
"shell",
"cmd",
"power",
"set-wakelock",
"acquire",
"-d",
"0",
"SCREEN_BRIGHT_WAKE_LOCK",
]
@patch("artemis.clients.ui_automator_client.subprocess.run")
def test_keep_device_awake_uses_heartbeat_when_wake_lock_is_unavailable(mock_run):
mock_run.side_effect = [
completed(),
completed(),
completed(),
completed(" mStayOn=false\n"),
completed(returncode=1, stderr="unknown command"),
completed("Wakelocks:\n"),
]
strategy = _keep_device_awake("device-123")
assert strategy == AWAKE_STRATEGY_HEARTBEAT
@patch.dict("os.environ", {"ARTEMIS_KEEP_DEVICE_AWAKE": "false"})
@patch("artemis.clients.ui_automator_client.subprocess.run")
def test_keep_device_awake_can_be_disabled(mock_run):
assert _keep_device_awake("device-123") is None
mock_run.assert_not_called()
@patch("artemis.clients.ui_automator_client.u2.connect")
@patch(
"artemis.clients.ui_automator_client._keep_device_awake",
return_value=AWAKE_STRATEGY_HEARTBEAT,
)
@patch("artemis.clients.ui_automator_client._ensure_maestro_not_installed")
def test_new_ui_connection_starts_heartbeat_fallback(
mock_remove_maestro, mock_keep_awake, mock_connect
):
client = UIAutomatorClient("device-123")
client._start_awake_heartbeat = MagicMock()
client._ensure_connected()
mock_remove_maestro.assert_called_once_with("device-123")
mock_keep_awake.assert_called_once_with("device-123")
mock_connect.assert_called_once_with("device-123")
client._start_awake_heartbeat.assert_called_once_with()
assert client._awake_strategy == AWAKE_STRATEGY_HEARTBEAT
@patch("artemis.clients.ui_automator_client._release_screen_wake_lock")
def test_disconnect_releases_screen_wake_lock(mock_release):
client = UIAutomatorClient("device-123")
client._awake_strategy = AWAKE_STRATEGY_WAKE_LOCK
client._stop_awake_heartbeat = MagicMock()
client.disconnect()
client._stop_awake_heartbeat.assert_called_once_with()
mock_release.assert_called_once_with("device-123")
assert client._awake_strategy is None
@patch("artemis.clients.ui_automator_client.threading.Thread")
@patch("artemis.clients.ui_automator_client.threading.Event")
def test_heartbeat_thread_is_daemonized_and_stoppable(mock_event, mock_thread):
stop_event = MagicMock()
thread = MagicMock()
mock_event.return_value = stop_event
mock_thread.return_value = thread
client = UIAutomatorClient("device-123")
client._start_awake_heartbeat()
client._stop_awake_heartbeat()
assert mock_thread.call_args.kwargs["daemon"] is True
thread.start.assert_called_once_with()
stop_event.set.assert_called_once_with()
thread.join.assert_called_once_with(timeout=2.0)
@pytest.mark.asyncio
async def test_android_driver_disconnect_cleans_up_ui_client():
ui_client = MagicMock()
driver = AndroidAdbDriver(
device_id="device-123",
adb_client=MagicMock(),
ui_adb_client=ui_client,
)
await driver.disconnect()
ui_client.disconnect.assert_called_once_with()
@patch("artemis.clients.ui_automator_client.subprocess.run")
def test_keep_device_awake_continues_after_an_adb_failure(mock_run):
mock_run.side_effect = [
completed(returncode=1, stderr="not allowed"),
OSError("adb temporarily unavailable"),
completed(),
completed("mStayOn=true\n"),
]
assert _keep_device_awake("device-123") == AWAKE_STRATEGY_STAY_ON
assert mock_run.call_count == 4
+168 -20
View File
@@ -14,23 +14,161 @@
"""Unit tests for UnifiedMobileController video recording and playback features.""" """Unit tests for UnifiedMobileController video recording and playback features."""
import subprocess
import time import time
from unittest.mock import AsyncMock, MagicMock, patch from unittest.mock import AsyncMock, MagicMock, patch
from uuid import uuid4 from uuid import uuid4
import pytest import pytest
import cv2
from artemis.context import ArtemisContext from artemis.context import ArtemisContext
from artemis.controllers.unified_controller import UnifiedMobileController from artemis.controllers.unified_controller import UnifiedMobileController
from artemis.drivers.mock.mock_driver import MockDeviceDriver from artemis.drivers.mock.mock_driver import MockDeviceDriver
from artemis.utils.video import ( from artemis.utils.video import (
RecordingSession, RecordingSession,
build_scrcpy_record_command,
get_ffmpeg_path,
get_active_session, get_active_session,
normalize_recording_to_mp4,
render_timeline_clip,
remove_active_session, remove_active_session,
set_active_session, set_active_session,
) )
def test_scrcpy_recording_locks_each_segment_orientation(tmp_path):
output_path = tmp_path / "recording.mkv"
command = build_scrcpy_record_command("scrcpy", "device-1", output_path)
assert "--capture-orientation=@" in command
assert command[command.index("--record") + 1] == str(output_path)
@pytest.mark.asyncio
async def test_normalize_recording_handles_resolution_change(tmp_path):
"""A portrait/landscape H.264 track must become one fixed-size MP4."""
ffmpeg = get_ffmpeg_path()
portrait = tmp_path / "portrait.mkv"
landscape = tmp_path / "landscape.mkv"
dynamic = tmp_path / "dynamic.mkv"
output = tmp_path / "recording.mp4"
for size, path in (("108x242", portrait), ("242x108", landscape)):
subprocess.run(
[
ffmpeg,
"-y",
"-f",
"lavfi",
"-i",
f"testsrc=size={size}:rate=10:duration=0.5",
"-c:v",
"libx264",
"-pix_fmt",
"yuv420p",
str(path),
],
check=True,
capture_output=True,
)
concat_file = tmp_path / "segments.txt"
concat_file.write_text(
f"file '{portrait.as_posix()}'\nfile '{landscape.as_posix()}'\n",
encoding="utf-8",
)
subprocess.run(
[
ffmpeg,
"-y",
"-f",
"concat",
"-safe",
"0",
"-i",
str(concat_file),
"-c",
"copy",
str(dynamic),
],
check=True,
capture_output=True,
)
assert await normalize_recording_to_mp4(dynamic, output, 108, 242)
capture = cv2.VideoCapture(str(output))
dimensions = set()
frame_count = 0
while True:
ok, frame = capture.read()
if not ok:
break
frame_count += 1
dimensions.add((frame.shape[1], frame.shape[0]))
capture.release()
assert frame_count >= 8
assert dimensions == {(108, 242)}
@pytest.mark.asyncio
async def test_analyzer_clip_stays_continuous_across_orientation_segments(tmp_path):
"""The analyzer must still receive one decodable MP4 across a rotation."""
ffmpeg = get_ffmpeg_path()
portrait = tmp_path / "recording.mkv"
landscape = tmp_path / "recording_001.mkv"
output = tmp_path / "agent_clip.mp4"
for color, size, path in (
("red", "108x242", portrait),
("blue", "242x108", landscape),
):
subprocess.run(
[
ffmpeg,
"-y",
"-f",
"lavfi",
"-i",
f"color={color}:size={size}:rate=15:duration=1",
"-c:v",
"libx264",
"-pix_fmt",
"yuv420p",
str(path),
],
check=True,
capture_output=True,
)
assert await render_timeline_clip(
[
{"path": portrait, "start": 0.0, "end": 1.0},
{"path": landscape, "start": 1.0, "end": 2.0},
],
0.25,
1.75,
output,
canvas_width=180,
canvas_height=320,
)
capture = cv2.VideoCapture(str(output))
dimensions = set()
frames = 0
while True:
ok, frame = capture.read()
if not ok:
break
dimensions.add((frame.shape[1], frame.shape[0]))
frames += 1
capture.release()
assert dimensions == {(180, 320)}
assert frames >= 20
@pytest.fixture @pytest.fixture
def mock_ctx(tmp_path): def mock_ctx(tmp_path):
ctx = MagicMock(spec=ArtemisContext) ctx = MagicMock(spec=ArtemisContext)
@@ -60,17 +198,21 @@ async def test_unified_controller_start_recording(mock_ctx, tmp_path):
mock_proc.returncode = None mock_proc.returncode = None
with patch("asyncio.create_subprocess_exec", AsyncMock(return_value=mock_proc)): with patch("asyncio.create_subprocess_exec", AsyncMock(return_value=mock_proc)):
with patch("asyncio.sleep", AsyncMock()): with patch(
res = await controller.start_video_recording(output_dir=tmp_path) "artemis.controllers.unified_controller.get_android_display_state",
AsyncMock(return_value=(0, 1080, 2424)),
):
with patch("asyncio.sleep", AsyncMock()):
res = await controller.start_video_recording(output_dir=tmp_path)
assert res.success is True assert res.success is True
assert get_active_session("emulator-5554") is not None assert get_active_session("emulator-5554") is not None
assert mock_ctx.data_engine.record_video_start.called assert mock_ctx.data_engine.record_video_start.called
# Calling start again should report already in progress # Calling start again should report already in progress
res2 = await controller.start_video_recording(output_dir=tmp_path) res2 = await controller.start_video_recording(output_dir=tmp_path)
assert res2.success is False assert res2.success is False
assert "already in progress" in res2.message assert "already in progress" in res2.message
remove_active_session("emulator-5554") remove_active_session("emulator-5554")
@@ -99,10 +241,11 @@ async def test_unified_controller_stop_recording(mock_ctx, tmp_path):
) )
set_active_session("emulator-5554", session) set_active_session("emulator-5554", session)
with patch.object( with patch(
controller, "artemis.controllers.unified_controller.remux_recording_to_mp4",
"_convert_mkv_to_mp4",
AsyncMock(side_effect=lambda src, dst: dst.write_bytes(b"mp4 content") or True), AsyncMock(side_effect=lambda src, dst: dst.write_bytes(b"mp4 content") or True),
), patch(
"artemis.controllers.unified_controller.write_recording_manifest", AsyncMock()
): ):
res = await controller.stop_video_recording() res = await controller.stop_video_recording()
@@ -135,7 +278,7 @@ async def test_unified_controller_extract_segment_metadata(mock_ctx, tmp_path):
set_active_session("emulator-5554", session) set_active_session("emulator-5554", session)
with patch( with patch(
"artemis.controllers.unified_controller.trim_video", "artemis.controllers.unified_controller.render_timeline_clip",
AsyncMock(side_effect=lambda src, s, e, dst: dst.write_bytes(b"segment mp4") or True), AsyncMock(side_effect=lambda src, s, e, dst: dst.write_bytes(b"segment mp4") or True),
): ):
res = await controller.extract_segment_metadata(start_time=2.0, end_time=8.0) res = await controller.extract_segment_metadata(start_time=2.0, end_time=8.0)
@@ -204,6 +347,14 @@ async def test_unified_controller_crash_recovery_and_multi_segment(mock_ctx, tmp
data_engine_start_time=time.time() - 30.0, data_engine_start_time=time.time() - 30.0,
local_video_path=seg2, local_video_path=seg2,
android_video_segments=[seg1], android_video_segments=[seg1],
android_segment_records=[
{
"path": seg1,
"output_path": tmp_path / "recording.mp4",
"start": 0.0,
"end": 15.0,
}
],
android_segment_index=1, android_segment_index=1,
process=mock_proc, process=mock_proc,
) )
@@ -211,14 +362,10 @@ async def test_unified_controller_crash_recovery_and_multi_segment(mock_ctx, tmp
with ( with (
patch( patch(
"artemis.controllers.unified_controller.concatenate_videos", "artemis.controllers.unified_controller.remux_recording_to_mp4",
AsyncMock(side_effect=lambda segs, out: out.write_bytes(b"combined mkv") or True),
),
patch.object(
controller,
"_convert_mkv_to_mp4",
AsyncMock(side_effect=lambda src, dst: dst.write_bytes(b"final mp4") or True), AsyncMock(side_effect=lambda src, dst: dst.write_bytes(b"final mp4") or True),
), ),
patch("artemis.controllers.unified_controller.write_recording_manifest", AsyncMock()),
): ):
res = await controller.stop_video_recording() res = await controller.stop_video_recording()
@@ -261,7 +408,8 @@ async def test_unified_controller_timeline_alignment(mock_ctx, tmp_path):
return True return True
with patch( with patch(
"artemis.controllers.unified_controller.trim_video", AsyncMock(side_effect=mock_trim) "artemis.controllers.unified_controller.render_timeline_clip",
AsyncMock(side_effect=mock_trim),
): ):
# Agent asks for system time range [5.0s, 10.0s] # Agent asks for system time range [5.0s, 10.0s]
res = await controller.extract_segment_metadata(start_time=5.0, end_time=10.0) res = await controller.extract_segment_metadata(start_time=5.0, end_time=10.0)