feat(plugins): add source-aware evidence and recording timelines

This commit is contained in:
chenmingxuan committed 2026-10-16 10:20:11 +08:00
1 parent 72ff342605
commit ef7af9988e
61 files changed
+3847 -108

No files matched your search

@@ -809,6 +809,7 @@ limitations under the License.
<div class="action-card-header" (click)="toggleActionExpanded('action-' + block.id + '-' + itemIdx, $event)">
<span class="material-symbols-outlined action-icon" [class.is-failed]="isActionFailed(item.data, block.data)">{{ getActionIcon(item.data) }}</span>
<span class="action-title">{{ getActionTitle(item.data) }}</span>
<button class="evidence-source" [title]="evidenceSourceLabel(item.data) + ' · 查看对应录像'" (click)="openEvidenceVideo(item.data, block.data, $event)">{{ evidenceSourceLabel(item.data) }}</button>
@if (isActionFailed(item.data, block.data)) {
<span class="action-status-badge is-failed">
<span class="material-symbols-outlined status-icon">error</span>
@@ -1050,7 +1051,7 @@ limitations under the License.
</div>
}
<!-- Generic Tool Calls (Text format for note tools, video analyzer, and other sub-agents) -->
@if (shouldShowTool(item.data, block.data) && !isDeviceActionTool(item.data)) {
@if (shouldShowTool(item.data, block.data) && !isEvidenceTool(item.data)) {
<div class="step-generic-tool"
[class.is-running]="isToolLineRunning(item, block)"
[class.is-note-tool]="isNoteTool(item.data)"
@@ -1091,9 +1092,11 @@ limitations under the License.
</div>
</div>
}
<!-- Action Card format (For Device Actions & ADB Commands) -->
@if (shouldShowTool(item.data, block.data) && isDeviceActionTool(item.data)) {
<!-- Device actions and plugin calls share evidence cards with explicit sources. -->
@if (shouldShowTool(item.data, block.data) && isEvidenceTool(item.data)) {
<div class="step-android-action-card tool-action-card"
[class.plugin-tool-card]="getPluginToolView(item.data)"
[class.is-auxiliary]="getPluginToolView(item.data)?.auxiliary"
[class.is-running]="isItemRunning(item, block)"
[class.has-no-details]="isAdbCommandTool(item.data) ? !getAdbCommandLine(item.data) : (!getToolTargetText(item.data) && !getToolInputText(item.data) && !getToolCoords(item.data) && !getToolAnalysisText(item.data) && !getToolGenericDetails(item.data) && getToolExtraParams(item.data).length === 0)"
[class.is-failed]="isToolFailed(item.data)"
@@ -1102,6 +1105,10 @@ limitations under the License.
<div class="action-card-header" (click)="toggleActionExpanded('tool-' + block.id + '-' + itemIdx, $event)">
<span class="material-symbols-outlined action-icon" [class.is-failed]="isToolFailed(item.data)">{{ getToolIcon(item.data) }}</span>
<span class="action-title">{{ getToolTitle(item.data) }}</span>
<button class="evidence-source" [title]="evidenceSourceLabel(item.data) + ' · 查看对应证据'" (click)="openEvidenceVideo(item.data, block.data, $event)">{{ evidenceSourceLabel(item.data) }}</button>
@if (getPluginToolView(item.data); as plugin) {
<span class="plugin-call-kind">{{ plugin.kind }} · {{ plugin.status }}</span>
}
@if (isToolFailed(item.data)) {
<span class="action-status-badge is-failed">
<span class="material-symbols-outlined status-icon">error</span>
@@ -1114,7 +1121,7 @@ limitations under the License.
</span>
</div>
<!-- Compact preview body (when collapsed) -->
@if (!isActionExpanded('tool-' + block.id + '-' + itemIdx) && (isAdbCommandTool(item.data) ? getAdbCommandLine(item.data) : (getToolTargetText(item.data) || getToolInputText(item.data) || getToolCoords(item.data) || getToolAnalysisText(item.data) || getToolGenericDetails(item.data)))) {
@if (!getPluginToolView(item.data) && !isActionExpanded('tool-' + block.id + '-' + itemIdx) && (isAdbCommandTool(item.data) ? getAdbCommandLine(item.data) : (getToolTargetText(item.data) || getToolInputText(item.data) || getToolCoords(item.data) || getToolAnalysisText(item.data) || getToolGenericDetails(item.data)))) {
<div class="action-card-body">
<!-- ADB command details -->
@if (isAdbCommandTool(item.data)) {
@@ -1196,7 +1203,21 @@ limitations under the License.
</div>
}
}
@if (!isAdbCommandTool(item.data)) {
@if (getPluginToolView(item.data); as plugin) {
@for (param of plugin.args; track param.key) {
<div class="grid-item">
<span class="grid-label">{{ param.key }}</span>
<span class="grid-value code-pill">{{ param.value }}</span>
</div>
}
@if (plugin.result) {
<div class="grid-item full-width">
<span class="grid-label">实际结果</span>
<span class="grid-value plugin-result">{{ plugin.result }}</span>
</div>
}
}
@if (!isAdbCommandTool(item.data) && !getPluginToolView(item.data)) {
@if (getToolTargetText(item.data)) {
<div class="grid-item">
<span class="grid-label">目标</span>
@@ -1251,7 +1272,7 @@ limitations under the License.
}
@if (getStepPostImageUrl(block.data, item.data)) {
<div class="screenshot-column">
<span class="screenshot-label">操作后</span>
<span class="screenshot-label">{{ getPluginToolView(item.data)?.auxiliary ? '观察截图' : '操作后' }}</span>
<div class="image-wrapper" (click)="openImageModal(getStepPostImageUrl(block.data, item.data)!, null, $event)">
<div class="img-relative-box">
<img [src]="getStepPostImageUrl(block.data, item.data)" alt="操作后截图" />
@@ -4053,3 +4053,16 @@ $stream-row-height: 25px;
th { background: #dbeafe; }
.md-para-gap { height: 8px; }
}
.evidence-source { max-width: 180px; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; padding: 2px 7px; border: 1px solid #cbd5e1; border-radius: 6px; background: #f1f5f9; color: #475569; font-size: 11px; cursor: pointer; }
.plugin-tool-card {
.action-card-header { flex-wrap: wrap; }
.plugin-call-kind { font-size: 10px; color: #64748b; }
.plugin-result { white-space: pre-wrap; overflow-wrap: anywhere; }
&.is-auxiliary:not(.is-expanded):not(.is-failed):not(.is-running) {
background: #f8fafc;
border-color: #e2e8f0;
box-shadow: none;
}
}
@@ -162,6 +162,7 @@ import {
extractCheckerResult,
renderMarkdownToHtml
} from '../../utils/markdown-parser.util';
import { getPluginToolView } from '../../utils/plugin-tool-view.util';
import {
getActionObject,
@@ -182,7 +183,8 @@ import {
getStepPostImageUrl,
isReportStatusAction,
getReportStatusValue,
getReportStatusExplanation
getReportStatusExplanation,
actionEvidenceSource
} from '../../utils/action-formatter.util';
import {
@@ -1612,6 +1614,17 @@ export class AgentStreamComponent implements AfterViewInit {
}
}
public evidenceSourceLabel(item: any): string { return actionEvidenceSource(item).label; }
public openEvidenceVideo(item: any, step: any, event: Event): void {
event.stopPropagation();
const started = this.agentService.currentSession()?.start_time;
const value = item?.timestamp || item?.start_time || step?.timestamp;
const timestamp = typeof value === 'number' ? value : Date.parse(value) / 1000;
if (!started || !Number.isFinite(timestamp)) return;
this.agentService.openVideoPlayer(undefined, undefined, undefined, Math.max(0, timestamp - started), undefined, actionEvidenceSource(item).id);
}
public getScreenRecordingButtonTitle(): string {
if (this.agentService.isCurrentSessionRunning()) {
return this.agentService.currentSession()?.recording_status === 'recording' ? "任务正在运行(录屏中)" : "任务正在运行";
@@ -1637,6 +1650,12 @@ export class AgentStreamComponent implements AfterViewInit {
return isDeviceActionTool(tool);
}
public getPluginToolView = getPluginToolView;
public isEvidenceTool(tool: any): boolean {
return isDeviceActionTool(tool) || getPluginToolView(tool) !== null;
}
public isFailureAnalyzerActionTool(tool: any): boolean {
return isDeviceActionTool(tool);
}
@@ -2033,11 +2052,12 @@ export class AgentStreamComponent implements AfterViewInit {
}
public getToolIcon(tool: any): string {
return getToolIcon(tool);
const plugin = getPluginToolView(tool);
return plugin ? (plugin.auxiliary ? 'visibility' : 'extension') : getToolIcon(tool);
}
public getToolTitle(tool: any): string {
return getToolTitle(tool);
return getPluginToolView(tool)?.title || getToolTitle(tool);
}
public getToolTargetText(tool: any): string {
@@ -95,6 +95,22 @@ limitations under the License.
</button>
</div>
</header>
@if (sources().length && !agentService.isVideoMinimized()) {
<div class="recording-sources">
<label for="recording-source">录像来源</label>
<select id="recording-source" [ngModel]="agentService.selectedRecordingSource()" (ngModelChange)="selectSource($event)">
@for (source of sources(); track source.id) {
<option [value]="source.id">{{ source.label }}{{ source.status === 'failed' ? '(录制异常)' : '' }}</option>
}
</select>
</div>
@if (agentService.recordingPlaybackMessage()) {
<div class="recording-notice" role="status">{{ agentService.recordingPlaybackMessage() }}</div>
}
@if (!agentService.activeVideoUrl()) {
<div class="recording-notice">此来源暂无可播放录像,请查看步骤截图或日志。</div>
}
}
<!-- Video Main Content Area -->
@if (!agentService.isVideoMinimized()) {
<div class="video-body">
@@ -332,8 +348,10 @@ limitations under the License.
@if (agentService.playerMode() === 'video' && agentService.recordingPlaybackStatus() === 'ready' && !agentService.isVideoLoading() && agentService.activeVideoUrl(); as videoUrl) {
@if (!videoLoadError()) {
<div class="video-container">
@if (gapMessage()) { <div class="recording-gap" role="status">{{ gapMessage() }}</div> }
<video
#videoRef
[style.visibility]="gapMessage() ? 'hidden' : 'visible'"
[src]="currentVideoUrl()"
playsinline
(loadedmetadata)="onLoadedMetadata()"
@@ -1017,3 +1017,42 @@
}
}
}
.recording-sources {
display: flex;
gap: 8px;
align-items: center;
padding: 8px 12px;
background: #f8fafc;
border-bottom: 1px solid #e2e8f0;
font-size: 12px;
color: #475569;
select {
min-width: 0;
flex: 1;
border: 1px solid #cbd5e1;
border-radius: 6px;
padding: 4px;
background: white;
}
}
.recording-notice {
padding: 8px 12px;
background: #fffbeb;
color: #92400e;
font-size: 12px;
}
.recording-gap {
position: absolute;
inset: 0;
display: flex;
align-items: center;
justify-content: center;
padding: 24px;
color: #475569;
background: #f1f5f9;
text-align: center;
z-index: 2;
}
@@ -1,6 +1,6 @@
import { ComponentFixture, TestBed } from '@angular/core/testing';
import { signal } from '@angular/core';
import { AgentService } from '../../services/agent.service';
import { AgentService, RecordingSource, VideoSegment } from '../../services/agent.service';
import { StepReplayFrame } from '../../core/models/stream.model';
import { FloatingVideoPlayerComponent } from './floating-video-player.component';
@@ -20,7 +20,7 @@ describe('FloatingVideoPlayerComponent video markers', () => {
index: 0, stepNumber: 1, title: '步骤 1: 点击设置', imageUrl: 'step.png',
actionText: '点击设置', timestamp: start + 5
};
return {
const service = {
isVideoWindowOpen: signal(true),
isVideoMinimized: signal(false),
recordingPlaybackStatus: signal('ready'),
@@ -29,14 +29,30 @@ describe('FloatingVideoPlayerComponent video markers', () => {
activeVideoTitle: signal('测试录像'),
isVideoLoading: signal(false),
activeVideoUrl: signal('/test.mp4'),
activeVideoSegments: signal([{ url: '/test.mp4', start: 0, duration: 20, width: 400, height: 800 }]),
activeVideoSegments: signal<VideoSegment[]>([{ url: '/test.mp4', start: 0, duration: 20, width: 400, height: 800 }]),
recordingSources: signal<RecordingSource[]>([]),
selectedRecordingSource: signal(''),
usesSessionTimeline: signal(false),
recordingSessionDuration: signal(0),
recordingPlaybackMessage: signal(''),
selectRecordingSource: jasmine.createSpy('selectRecordingSource'),
currentSessionStepFrames: signal([frame]),
currentSession: signal({ session_id: 'test', start_time: start }),
currentSessionId: signal('test'),
videoSeekRequest: signal(null),
videoSeekRequest: signal<{ seconds: number; requestId: number; sourceId?: string } | null>(null),
stepSeekRequest: signal(null),
consumeVideoAutoplay: jasmine.createSpy('consumeVideoAutoplay').and.returnValue(false)
};
service.selectRecordingSource.and.callFake((id: string) => {
const source = service.recordingSources().find(item => item.id === id);
if (!source) return false;
if (service.selectedRecordingSource() === id) return true;
service.selectedRecordingSource.set(id);
service.activeVideoSegments.set(source.segments);
service.activeVideoUrl.set(source.url || source.segments[0]?.url || '');
return true;
});
return service;
}
beforeEach(() => {
@@ -58,6 +74,7 @@ describe('FloatingVideoPlayerComponent video markers', () => {
seeking: { configurable: true, get: () => seeking },
ended: { configurable: true, get: () => false },
duration: { configurable: true, get: () => 20 },
readyState: { configurable: true, get: () => 1 },
currentTime: {
configurable: true, get: () => time,
set: (value: number) => { time = value; seeking = true; }
@@ -72,6 +89,7 @@ describe('FloatingVideoPlayerComponent video markers', () => {
video.dispatchEvent(new Event('play'));
return Promise.resolve();
});
video.dispatchEvent(new Event('loadedmetadata'));
});
function finishSeek(): void {
@@ -189,6 +207,7 @@ describe('FloatingVideoPlayerComponent video markers', () => {
it('resumes after seeking a recording without a segment manifest', () => {
agentService.activeVideoSegments.set([]);
agentService.usesSessionTimeline.set(true);
fixture.detectChanges();
video.dispatchEvent(new Event('loadedmetadata'));
component.togglePlay();
@@ -201,6 +220,115 @@ describe('FloatingVideoPlayerComponent video markers', () => {
expect(play).toHaveBeenCalledTimes(1);
});
function setSources(computerStart = 0): void {
agentService.recordingSources.set([
{ id: 'android', label: '安卓', kind: 'android', target_ids: ['android'], status: 'ready',
segments: [{ url: '/test.mp4', start: 0, duration: 20, width: 400, height: 800 }] },
{ id: 'screen', label: '电脑', kind: 'computer', target_ids: ['window-1'], status: 'ready',
segments: [{ url: '/computer.mp4', start: computerStart, duration: 20, width: 1280, height: 720 }] }
]);
agentService.usesSessionTimeline.set(true);
agentService.recordingSessionDuration.set(40);
agentService.selectRecordingSource('android');
fixture.detectChanges();
}
it('handles a seek request once without seeking back on play or pause', () => {
const seek = spyOn(component, 'seekToSessionTime').and.callThrough();
agentService.videoSeekRequest.set({ seconds: 5, requestId: 1 });
fixture.detectChanges();
finishSeek();
time = 9;
component.togglePlay();
fixture.detectChanges();
expect(time).toBe(9);
component.togglePlay();
fixture.detectChanges();
expect(time).toBe(9);
expect(seek).toHaveBeenCalledOnceWith(5, false);
});
it('defers a seek command until the recording is ready', () => {
const seek = spyOn(component, 'seekToSessionTime');
agentService.recordingPlaybackStatus.set('processing');
agentService.videoSeekRequest.set({ seconds: 5, requestId: 2 });
fixture.detectChanges();
expect(seek).not.toHaveBeenCalled();
agentService.recordingPlaybackStatus.set('ready');
fixture.detectChanges();
component.onLoadedMetadata();
expect(seek).toHaveBeenCalledOnceWith(5, false);
});
it('switches to the requested source, waits for seeked and allows later manual source changes', () => {
setSources(2);
component.togglePlay();
agentService.videoSeekRequest.set({ seconds: 7, requestId: 3, sourceId: 'window-1' });
fixture.detectChanges();
play.calls.reset();
video.dispatchEvent(new Event('loadedmetadata'));
expect(time).toBe(5);
expect(play).not.toHaveBeenCalled();
finishSeek();
expect(component.currentTime()).toBe(7);
expect(paused).toBeFalse();
expect(play).toHaveBeenCalledTimes(1);
component.selectSource('android');
fixture.detectChanges();
expect(agentService.selectedRecordingSource()).toBe('android');
video.dispatchEvent(new Event('loadedmetadata'));
finishSeek();
expect(time).toBe(7);
expect(agentService.selectedRecordingSource()).toBe('android');
});
it('preserves task time when switching into a gap and advances only on explicit play', () => {
setSources(20);
time = 12;
component.onTimeUpdate();
component.togglePlay();
component.selectSource('screen');
fixture.detectChanges();
play.calls.reset();
video.dispatchEvent(new Event('loadedmetadata'));
expect(component.currentTime()).toBe(12);
expect(component.gapMessage()).toContain('未录制');
expect(paused).toBeTrue();
expect(play).not.toHaveBeenCalled();
component.togglePlay();
finishSeek();
expect(component.currentTime()).toBe(20);
expect(component.gapMessage()).toBe('');
expect(paused).toBeFalse();
});
it('retains a paused source switch and resumes with one play click', () => {
setSources();
time = 6;
component.onTimeUpdate();
component.selectSource('screen');
fixture.detectChanges();
video.dispatchEvent(new Event('loadedmetadata'));
finishSeek();
expect(component.currentTime()).toBe(6);
expect(paused).toBeTrue();
expect(play).not.toHaveBeenCalled();
component.togglePlay();
expect(play).toHaveBeenCalledTimes(1);
expect(paused).toBeFalse();
});
it('does not play another source when the requested source has no recording', () => {
setSources();
agentService.videoSeekRequest.set({ seconds: 5, requestId: 4, sourceId: 'missing-window' });
fixture.detectChanges();
component.togglePlay();
expect(component.gapMessage()).toContain('来源没有录像');
expect(component.currentTime()).toBe(5);
expect(paused).toBeTrue();
expect(play).not.toHaveBeenCalled();
});
it('resets a rejected play request so a single subsequent click can resume', async () => {
play.and.callFake(() => {
paused = false;
@@ -25,6 +25,7 @@ import {
ViewChild,
HostListener,
effect,
untracked,
OnDestroy
} from '@angular/core';
@@ -33,7 +34,7 @@ import { AgentService } from '../../services/agent.service';
import { StepReplayFrame } from '../../core/models/stream.model';
import { drawActionCoordinatesOnOverlay } from '../../utils/image-overlay.util';
import { getActionIcon } from '../../utils/action-formatter.util';
import { locateTimelineTime, sessionTimeToTimelineTime } from '../../utils/recording-timeline.util';
import { locateCoveredSessionTime, locateTimelineTime, sessionTimeToTimelineTime } from '../../utils/recording-timeline.util';
import { buildVideoStepMarkers, VideoStepMarker } from '../../utils/video-step-markers.util';
@Component({
@@ -61,6 +62,11 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
public isMuted = signal<boolean>(false);
public isTheaterMode = signal<boolean>(false);
public videoLoadError = signal<boolean>(false);
public gapMessage = signal('');
private sourceSwitch: { time: number; play: boolean } | null = null;
private lastHandledVideoSeekRequestId: number | null = null;
private gapCanAdvance = true;
public sources = computed(() => this.agentService.recordingSources?.() || []);
public liveStreamUrl = signal<string>('/api/stream/device-live');
public liveStreamError = signal<boolean>(false);
public activeSegmentIndex = signal<number>(0);
@@ -71,13 +77,14 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
private pendingLocalTime: number | null = null;
private pendingAutoplay = false;
private pendingAbsoluteSeek: number | null = null;
private waitingForMetadata = true;
private seekPlaybackIntent: boolean | null = null;
private playRequestId = 0;
// Step Replay state signals
public activeStepIndex = signal<number>(0);
public isStepPlaying = signal<boolean>(false);
public viewMode = signal<'pre' | 'post'>('pre'); // 'pre' = action preview with coords, 'post' = action result
public viewMode = signal<'pre' | 'post'>('post'); // default to the observed result
public isCardHovered = signal<boolean>(false);
public isCardPinned = signal<boolean>(false);
public isCardVisible = computed(() => this.isCardHovered() || this.isCardPinned());
@@ -87,10 +94,14 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
public stepFrames = computed(() => this.agentService.currentSessionStepFrames());
public videoStepMarkers = computed(() => buildVideoStepMarkers(
this.stepFrames(),
this.stepFrames().filter(frame => {
const source = this.sources().find(s => s.id === this.agentService.selectedRecordingSource?.());
return !source || source.target_ids.includes(frame.sourceId || 'android');
}),
this.agentService.currentSession()?.start_time,
this.agentService.activeVideoSegments(),
this.duration()
this.duration(),
this.agentService.usesSessionTimeline?.() || false
));
public totalStepFrames = computed(() => this.stepFrames().length);
public currentStepFrame = computed<StepReplayFrame | null>(() => {
@@ -136,12 +147,23 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
const url = this.agentService.activeVideoUrl();
const segments = this.agentService.activeVideoSegments();
this.videoLoadError.set(false);
this.waitingForMetadata = true;
this.activeSegmentIndex.set(0);
this.currentTime.set(0);
this.duration.set(segments.reduce((sum, segment) => sum + segment.duration, 0));
const switching = this.sourceSwitch;
this.sourceSwitch = null;
this.currentTime.set(switching?.time || 0);
this.gapMessage.set('');
this.gapCanAdvance = true;
this.duration.set(this.agentService.usesSessionTimeline?.()
? Math.max(this.agentService.recordingSessionDuration?.() || 0, ...segments.map(s => s.start + s.duration))
: segments.reduce((sum, segment) => sum + segment.duration, 0));
this.isPlaying.set(false);
this.pendingLocalTime = null;
this.pendingAutoplay = false;
if (switching) {
this.pendingAbsoluteSeek = switching.time;
this.pendingAutoplay = switching.play;
}
this.seekPlaybackIntent = null;
this.playRequestId++;
},
@@ -157,7 +179,7 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
if (curSessionId !== lastTrackedSessionId) {
lastTrackedSessionId = curSessionId;
this.activeStepIndex.set(0);
this.viewMode.set('pre');
this.viewMode.set('post');
this.pauseStepPlay();
} else if (total > 0 && this.activeStepIndex() >= total) {
this.activeStepIndex.set(total - 1);
@@ -186,15 +208,42 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
effect(() => {
const request = this.agentService.videoSeekRequest();
if (!request) return;
this.pendingAbsoluteSeek = request.seconds;
if (this.agentService.recordingPlaybackStatus() !== 'ready') return;
const video = this.videoRef?.nativeElement;
if (video && video.readyState >= 1) {
const target = this.pendingAbsoluteSeek;
this.pendingAbsoluteSeek = null;
this.seekToSessionTime(target, true);
}
const status = this.agentService.recordingPlaybackStatus();
const sources = this.sources();
if (!request || status !== 'ready' || request.requestId === this.lastHandledVideoSeekRequestId) return;
// A seek is a command, not a binding to playback/source state. Reading
// those signals in this effect would replay the command after play/pause
// or a manual source selection.
untracked(() => {
this.lastHandledVideoSeekRequestId = request.requestId;
this.pendingAbsoluteSeek = request.seconds;
if (request.sourceId && sources.length) {
const source = sources.find(s => s.id === request.sourceId || s.target_ids.includes(request.sourceId!));
if (!source) {
this.pendingAbsoluteSeek = null;
this.pendingLocalTime = null;
this.pendingAutoplay = false;
this.seekPlaybackIntent = null;
this.playRequestId++;
this.gapCanAdvance = false;
this.currentTime.set(request.seconds);
this.gapMessage.set('此步骤的来源没有录像,请查看步骤截图或日志。');
this.videoRef?.nativeElement.pause();
this.isPlaying.set(false);
return;
}
if (source.id !== this.agentService.selectedRecordingSource?.()) {
this.sourceSwitch = { time: request.seconds, play: this.isPlaying() };
this.agentService.selectRecordingSource(source.id);
return;
}
}
const video = this.videoRef?.nativeElement;
if (video && video.readyState >= 1 && !this.waitingForMetadata) {
this.pendingAbsoluteSeek = null;
this.seekToSessionTime(request.seconds, this.isPlaying());
}
});
});
effect(() => {
@@ -305,6 +354,7 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
public onLoadedMetadata(): void {
const v = this.videoRef?.nativeElement;
if (v) {
this.waitingForMetadata = false;
const segments = this.agentService.activeVideoSegments();
if (!segments.length) {
this.duration.set(v.duration || 0);
@@ -313,14 +363,6 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
v.playbackRate = this.playbackRate();
v.loop = this.isLooping() && !segments.length;
v.muted = this.isMuted();
if (this.pendingAbsoluteSeek !== null) {
const target = this.pendingAbsoluteSeek;
this.pendingAbsoluteSeek = null;
this.seekToSessionTime(target, true);
// A seek into another finalized segment changes the video source. Let
// that segment's metadata event apply pendingLocalTime.
if (this.pendingLocalTime !== null) return;
}
const autoplay = this.agentService.consumeVideoAutoplay();
if (autoplay) {
this.isMuted.set(true);
@@ -328,6 +370,14 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
}
const shouldPlay = this.pendingAutoplay || autoplay;
this.pendingAutoplay = false;
if (this.pendingAbsoluteSeek !== null) {
const target = this.pendingAbsoluteSeek;
this.pendingAbsoluteSeek = null;
this.seekToSessionTime(target, shouldPlay);
// Same-source seeks resume on seeked; cross-segment seeks resume after
// the destination metadata and seeked. Never play the old frame here.
return;
}
if (this.pendingLocalTime !== null) {
const localTime = Math.max(0, Math.min(v.duration || this.pendingLocalTime, this.pendingLocalTime));
this.pendingLocalTime = null;
@@ -339,6 +389,7 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
}
public onTimeUpdate(): void {
if (this.gapMessage()) return;
const v = this.videoRef?.nativeElement;
if (v) {
const segments = this.agentService.activeVideoSegments();
@@ -375,6 +426,14 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
this.isPlaying.set(false);
const segments = this.agentService.activeVideoSegments();
if (segments.length && this.activeSegmentIndex() < segments.length - 1) {
const current = segments[this.activeSegmentIndex()];
const next = segments[this.activeSegmentIndex() + 1];
if (this.agentService.usesSessionTimeline?.() && next.start - current.start - current.duration > 0.05) {
this.currentTime.set(current.start + current.duration);
this.gapCanAdvance = true;
this.gapMessage.set('此处录像存在中断。点击播放可前往下一段,或选择其他步骤。');
return;
}
this.pendingLocalTime = 0;
this.pendingAutoplay = true;
this.activeSegmentIndex.update(index => index + 1);
@@ -400,6 +459,12 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
* Playback actions
*/
public togglePlay(): void {
if (this.gapMessage()) {
if (!this.gapCanAdvance) return;
const next = this.agentService.activeVideoSegments().find(s => s.start > this.currentTime());
if (next) this.seek(next.start, true);
return;
}
const v = this.videoRef?.nativeElement;
if (!v) return;
if (this.pendingLocalTime !== null) {
@@ -462,7 +527,13 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
*/
public seekToSessionTime(sessionSeconds: number, autoplay = false): void {
const segments = this.agentService.activeVideoSegments();
this.seek(sessionTimeToTimelineTime(segments, sessionSeconds), autoplay);
this.seek(this.agentService.usesSessionTimeline?.() ? sessionSeconds : sessionTimeToTimelineTime(segments, sessionSeconds), autoplay);
}
public selectSource(id: string): void {
if (id === this.agentService.selectedRecordingSource?.()) return;
this.sourceSwitch = { time: this.currentTime(), play: this.isPlaying() };
if (!this.agentService.selectRecordingSource(id)) this.sourceSwitch = null;
}
public seekToVideoMarker(marker: VideoStepMarker): void {
@@ -484,7 +555,23 @@ export class FloatingVideoPlayerComponent implements OnDestroy {
if (!v) return;
const target = Math.max(0, Math.min(this.duration(), seconds));
const segments = this.agentService.activeVideoSegments();
const location = locateTimelineTime(segments, target);
// URL-only legacy recordings have no coverage manifest. Preserve their
// native video seeking instead of treating every timestamp as a gap.
const sessionTimeline = this.agentService.usesSessionTimeline?.() && segments.length > 0;
const location = sessionTimeline ? locateCoveredSessionTime(segments, target) : locateTimelineTime(segments, target);
if (sessionTimeline && !location) {
this.playRequestId++;
this.pendingLocalTime = null;
this.pendingAutoplay = false;
this.seekPlaybackIntent = null;
v.pause();
this.isPlaying.set(false);
this.currentTime.set(target);
this.gapCanAdvance = true;
this.gapMessage.set('此时间段未录制(尚未开始、暂停或录制中断),可查看步骤截图与日志。');
return;
}
this.gapMessage.set('');
const shouldPlay = autoplay || this.pendingAutoplay || (this.seekPlaybackIntent ?? !v.paused);
if (!location) {
this.seekWithinVideo(target, shouldPlay);
@@ -4,6 +4,7 @@ import { HttpClient } from '@angular/common/http';
import { finalize } from 'rxjs';
interface Capabilities {
recording?: { codec: string; max_long_edge: number; fps: number; h265_kbps: number; h264_kbps: number; computer_scope: string; minimum_disk_mb: number; hevc_playback: boolean };
adb_auto_connect: boolean;
adb_discovery_interval_seconds: number;
adb_discovery_timeout_seconds: number;
@@ -33,6 +34,20 @@ interface Capabilities {
<span>视频分析<small>允许智能体分析任务画面。运行期间会临时录制,完成后不保留录像(除非开启录屏)。</small></span></label>
<label><input type="checkbox" role="switch" [(ngModel)]="config.recording_enabled" [disabled]="busy() || loading() || !loaded()" />
<span>保存录屏<small>保留任务录像以便复盘,可单独开启,不要求启用视频分析。</small></span></label>
@if (config.recording; as recording) {
<details class="execution-settings">
<summary>录像留档设置</summary>
<p>默认低码率留档,不影响模型观察截图。仅录实际使用的来源;屏幕模式包含该屏幕其他可见内容。</p>
<div class="number-fields">
<label>编码<select [(ngModel)]="recording.codec" [disabled]="busy() || !loaded()"><option value="auto">自动选择兼容编码</option><option value="h265">优先 H.265</option><option value="h264">H.264</option></select></label>
<label>电脑录制范围<select [(ngModel)]="recording.computer_scope" [disabled]="busy() || !loaded()"><option value="screen">当前控制的屏幕</option><option value="window">仅目标窗口</option></select></label>
<label>最长边(像素)<input type="number" min="320" max="4096" [(ngModel)]="recording.max_long_edge" [disabled]="busy() || !loaded()" /></label>
<label>最高帧率<input type="number" min="5" max="60" [(ngModel)]="recording.fps" [disabled]="busy() || !loaded()" /></label>
<label>H.265 目标码率(kbps)<input type="number" min="128" max="10000" [(ngModel)]="recording.h265_kbps" [disabled]="busy() || !loaded()" /></label>
<label>H.264 目标码率(kbps)<input type="number" min="128" max="20000" [(ngModel)]="recording.h264_kbps" [disabled]="busy() || !loaded()" /></label>
</div>
</details>
}
<label><input type="checkbox" role="switch" [(ngModel)]="config.outputter_enabled" [disabled]="busy() || loading() || !loaded()" />
<span>最终报告生成<small>根据执行记录生成任务结果报告。</small></span></label>
<label><input type="checkbox" role="switch" [(ngModel)]="config.step_summarizer_enabled" [disabled]="busy() || loading() || !loaded()" />
@@ -68,7 +83,7 @@ interface Capabilities {
summary { cursor: pointer; font-weight: 600; color: #334155; }
.number-fields { display: grid; grid-template-columns: 1fr 1fr; gap: 12px 16px; }
.number-fields label { display: block; padding: 0; color: #334155; }
.number-fields input { box-sizing: border-box; display: block; width: 100%; height: auto; margin-top: 8px; padding: 9px; border: 1px solid #cbd5e1; border-radius: 8px; font: inherit; }
.number-fields input, .number-fields select { box-sizing: border-box; display: block; width: 100%; height: auto; margin-top: 8px; padding: 9px; border: 1px solid #cbd5e1; border-radius: 8px; font: inherit; }
@media(max-width: 480px) { .number-fields { grid-template-columns: 1fr; } }`]
})
export class CapabilitySettingsComponent implements OnInit {
@@ -83,6 +98,9 @@ export class CapabilitySettingsComponent implements OnInit {
}
save(): void {
if (this.busy() || this.loading() || !this.loaded() || !this.valid()) return;
if (this.config.recording) {
this.config.recording.hevc_playback = !!document.createElement('video').canPlayType('video/mp4; codecs="hvc1"');
}
this.busy.set(true); this.message.set(''); this.error.set('');
this.http.post<Capabilities>('/api/system/capability-settings', this.config).pipe(finalize(() => this.busy.set(false))).subscribe({
next: config => { this.config = config; this.changed.emit(config); this.message.set('功能设置已保存,新任务将使用此配置。'); },
@@ -91,6 +109,12 @@ export class CapabilitySettingsComponent implements OnInit {
}
valid(): boolean {
const c = this.config;
const r = c.recording;
if (r && (!['auto', 'h264', 'h265'].includes(r.codec) || !['screen', 'window'].includes(r.computer_scope)
|| !Number.isInteger(r.max_long_edge) || r.max_long_edge < 320 || r.max_long_edge > 4096
|| !Number.isInteger(r.fps) || r.fps < 5 || r.fps > 60
|| !Number.isInteger(r.h265_kbps) || r.h265_kbps < 128 || r.h265_kbps > 10000
|| !Number.isInteger(r.h264_kbps) || r.h264_kbps < 128 || r.h264_kbps > 20000)) return false;
return Number.isInteger(c.task_timeout_seconds) && c.task_timeout_seconds >= 30 && c.task_timeout_seconds <= 86400
&& c.observation_timeout_seconds > 0 && c.observation_timeout_seconds <= 300
&& Number.isInteger(c.observation_max_attempts) && c.observation_max_attempts >= 1 && c.observation_max_attempts <= 5
@@ -225,6 +225,8 @@ export interface CheckerResult {
}
export interface StepReplayFrame {
sourceId?: string;
sourceLabel?: string;
index: number;
stepNumber: number;
rawStepNumber?: number;
@@ -230,6 +230,17 @@ describe('AgentService recording finalization lifecycle', () => {
(service as any).videoWaitStartedAt = Date.now();
service.activeVideoUrl = signal<string | null>(null);
service.activeVideoSegments = signal<any[]>([]);
service.recordingSources = signal([]);
service.selectedRecordingSource = signal('');
service.usesSessionTimeline = signal(false);
service.recordingSessionDuration = signal(0);
service.activeVideoTitle = signal('');
service.isVideoWindowOpen = signal(false);
service.isVideoMinimized = signal(false);
service.videoSeekRequest = signal(null);
service.currentSessionId = signal<string | null>('session-1');
service.sessions = computed(() => (service as any).rawSessions());
(service as any).videoSeekRequestId = 0;
service.isVideoLoading = signal(false);
service.recordingPlaybackStatus = signal('idle');
service.recordingPlaybackMessage = signal('');
@@ -293,6 +304,89 @@ describe('AgentService recording finalization lifecycle', () => {
expect(service.isVideoLoading()).toBeFalse();
expect(service.recordingPlaybackMessage()).toBe('ffmpeg failed');
});
it('publishes task-time sources and selects a computer recording by its target alias', () => {
const android = {
id: 'android', label: '安卓', kind: 'android', target_ids: ['android'], status: 'ready',
segments: [{ url: '/videos/android.mp4', start: 0, duration: 10, width: 720, height: 1280 }]
};
const computer = {
id: 'screen-1', label: '电脑', kind: 'computer', target_ids: ['window-1', 'window-2'], status: 'ready',
message: '已使用兼容编码',
segments: [{ url: '/videos/computer.mp4', start: 12, duration: 10, width: 1280, height: 720 }]
};
const service = createVideoService({
session_id: 'session-1', status: 'ready', has_video: true,
video_url: android.segments[0].url, video_segments: android.segments,
sources: [android, computer], session_timeline: true, session_duration: 25
});
(service as any).requestSessionVideo('session-1', 1);
expect(service.recordingSources()).toEqual([android, computer]);
expect(service.selectedRecordingSource()).toBe('android');
expect(service.usesSessionTimeline()).toBeTrue();
expect(service.recordingSessionDuration()).toBe(25);
expect(service.selectRecordingSource('window-2')).toBeTrue();
expect(service.selectedRecordingSource()).toBe('screen-1');
expect(service.activeVideoUrl()).toBe('/videos/computer.mp4');
expect(service.activeVideoSegments()).toBe(computer.segments);
expect(service.recordingPlaybackMessage()).toBe('已使用兼容编码');
expect(service.selectRecordingSource('screen-1')).toBeTrue();
expect(service.activeVideoSegments()).toBe(computer.segments);
expect(service.selectRecordingSource('missing-source')).toBeFalse();
expect(service.selectedRecordingSource()).toBe('screen-1');
expect(service.activeVideoUrl()).toBe('/videos/computer.mp4');
});
it('keeps a failed source explicit without retaining another source video', () => {
const service = createVideoService({});
service.activeVideoUrl.set('/videos/android.mp4');
service.recordingSources.set([{
id: 'screen-1', label: '电脑', kind: 'computer', target_ids: ['window-1'], status: 'failed',
segments: [], message: '录制进程未产生画面'
}]);
expect(service.selectRecordingSource('window-1')).toBeTrue();
expect(service.activeVideoUrl()).toBeNull();
expect(service.activeVideoSegments()).toEqual([]);
expect(service.recordingPlaybackMessage()).toBe('录制进程未产生画面');
});
it('clears the previous task source and seek command when opening another task normally', () => {
const service = createVideoService({});
const load = spyOn<any>(service, 'requestSessionVideo');
service.recordingSources.set([{
id: 'screen-old', label: '电脑', kind: 'computer', target_ids: ['window-old'], status: 'ready',
segments: [{ url: '/videos/old.mp4', start: 0, duration: 10, width: 1280, height: 720 }]
}]);
service.selectRecordingSource('screen-old');
service.usesSessionTimeline.set(true);
service.recordingSessionDuration.set(10);
service.requestVideoSeek(6, 'window-old');
service.openVideoPlayer('session-2');
expect(service.videoSeekRequest()).toBeNull();
expect(service.recordingSources()).toEqual([]);
expect(service.selectedRecordingSource()).toBe('');
expect(service.usesSessionTimeline()).toBeFalse();
expect(service.recordingSessionDuration()).toBe(0);
expect(service.activeVideoSegments()).toEqual([]);
expect(service.activeVideoUrl()).toBeNull();
expect(load).toHaveBeenCalledOnceWith('session-2', 2);
});
it('replaces the previous task seek with the new source-specific command', () => {
const service = createVideoService({});
spyOn<any>(service, 'requestSessionVideo');
service.requestVideoSeek(6, 'window-old');
const oldRequestId = service.videoSeekRequest()!.requestId;
service.openVideoPlayer('session-2', undefined, undefined, 14, undefined, 'window-new');
expect(service.videoSeekRequest()).toEqual({ seconds: 14, sourceId: 'window-new', requestId: oldRequestId + 1 });
});
});
describe('AgentService video analysis seeking', () => {
@@ -39,6 +39,18 @@ export interface VideoSegment {
duration_ms?: number;
}
export interface RecordingSource {
id: string;
label: string;
kind: string;
target_ids: string[];
status: string;
message?: string;
codec?: string;
segments: VideoSegment[];
url?: string;
}
export type RecordingPlaybackStatus =
| 'idle'
| 'live'
@@ -64,6 +76,9 @@ export interface GuidanceEvent {
}
interface SessionVideoResponse {
sources?: RecordingSource[];
session_timeline?: boolean;
session_duration?: number;
session_id: string;
status?: 'processing' | 'ready' | 'failed' | 'unavailable';
has_video: boolean;
@@ -272,12 +287,27 @@ export class AgentService {
public isVideoMinimized = signal<boolean>(false);
public activeVideoUrl = signal<string | null>(null);
public activeVideoSegments = signal<VideoSegment[]>([]);
public recordingSources = signal<RecordingSource[]>([]);
public selectedRecordingSource = signal<string>('');
public usesSessionTimeline = signal(false);
public recordingSessionDuration = signal(0);
public selectRecordingSource(id: string): boolean {
const source = this.recordingSources().find(s => s.id === id || s.target_ids?.includes(id));
if (!source) return false;
if (this.selectedRecordingSource() === source.id) return true;
this.selectedRecordingSource.set(source.id);
this.activeVideoSegments.set(source.segments);
this.activeVideoUrl.set(source.url || source.segments[0]?.url || null);
this.recordingPlaybackMessage.set(source.message || '');
return true;
}
public activeVideoTitle = signal<string>('');
public isVideoLoading = signal<boolean>(false);
public recordingPlaybackStatus = signal<RecordingPlaybackStatus>('idle');
public recordingPlaybackMessage = signal<string>('');
public shouldAutoplayVideo = signal<boolean>(false);
public videoSeekRequest = signal<{ seconds: number; requestId: number } | null>(null);
public videoSeekRequest = signal<{ seconds: number; requestId: number; sourceId?: string } | null>(null);
public stepSeekRequest = signal<{ index: number; requestId: number } | null>(null);
public playerMode = signal<'video' | 'steps'>('video');
public streamResetEvent = signal<LLMStreamResetEventData | null>(null);
@@ -1838,7 +1868,8 @@ export class AgentService {
videoUrl?: string,
title?: string,
seekSeconds?: number,
stepIndex?: number
stepIndex?: number,
sourceId?: string
): void {
const targetSessionId = sessionId || this.currentSessionId();
const session = this.sessions().find(s => s.session_id === targetSessionId);
@@ -1854,7 +1885,12 @@ export class AgentService {
this.shouldAutoplayVideo.set(false);
this.recordingPlaybackMessage.set('');
this.activeVideoSegments.set([]);
if (Number.isFinite(seekSeconds)) this.requestVideoSeek(Number(seekSeconds));
this.recordingSources.set([]);
this.selectedRecordingSource.set('');
this.usesSessionTimeline.set(false);
this.recordingSessionDuration.set(0);
if (Number.isFinite(seekSeconds)) this.requestVideoSeek(Number(seekSeconds), sourceId);
else this.videoSeekRequest.set(null);
if (Number.isFinite(stepIndex)) {
this.playerMode?.set('steps');
this.requestStepSeek(Number(stepIndex));
@@ -1941,6 +1977,9 @@ export class AgentService {
return;
}
const status = res.status || (res.has_video && res.video_url ? 'ready' : 'unavailable');
this.recordingSources.set(res.sources || []);
this.usesSessionTimeline.set(!!res.session_timeline);
this.recordingSessionDuration.set(res.session_duration || 0);
if (status === 'ready' && res.video_url) {
this.cancelVideoRetry();
this.isVideoLoading.set(false);
@@ -1948,6 +1987,7 @@ export class AgentService {
this.recordingPlaybackMessage.set('');
this.activeVideoUrl.set(res.video_url);
this.activeVideoSegments.set(res.video_segments || []);
this.selectedRecordingSource.set(res.sources?.find(s => s.segments.length || s.url)?.id || '');
this.playerMode?.set('video');
this.rawSessions.update((list) =>
list.map((s) => s.session_id === sessionId
@@ -2041,10 +2081,11 @@ export class AgentService {
return shouldAutoplay;
}
public requestVideoSeek(seconds: number): void {
public requestVideoSeek(seconds: number, sourceId?: string): void {
if (!Number.isFinite(seconds)) return;
this.videoSeekRequest.set({
seconds: Math.max(0, seconds),
sourceId,
requestId: ++this.videoSeekRequestId
});
}
@@ -2,13 +2,109 @@ import {
getStepPreImageUrl,
getStepPostImageUrl,
extractStepReplayFrames,
getActionTitle,
getActionCoords,
getActionTargetText,
isActionFailed,
getActionErrorMessage,
extractActionExtraParams
} from './action-formatter.util';
describe('plugin replay marker descriptions', () => {
it('does not parse digits in real hashed click and observation names as coordinates', () => {
const click = { action: 'plug_96c522c2e2_click', plugin_id: 'open-computer-use', args: { app: 'Demo' } };
const observe = { action: 'plug_5f6ec6ac87_get_app_state', plugin_id: 'open-computer-use', args: { app: 'Demo' } };
expect(getActionTitle(click)).toBe('点击');
expect(getActionTitle(observe)).toBe('观察电脑窗口');
expect(getActionCoords(click)).toBe('');
expect(getActionCoords(observe)).toBe('');
expect(getActionTargetText(observe)).toBe('Demo');
expect(getActionCoords({ ...click, args: { app: 'Demo', x: 100, y: 200 } })).toBe('[100, 200]');
});
it('uses the original tool name from the trace rather than its dynamic binding name', () => {
const trace = { name: 'plug_96c522c2e2_click', type: 'tool',
payload: { plugin_id: 'desktop', tool_name: 'double_click', args: { x: 70, y: 80 } } };
expect(getActionTitle(trace)).toBe('双击');
expect(getActionCoords(trace)).toBe('[70, 80]');
});
it('counts a Flash action and its tool trace as one frame using their shared call id', () => {
const callId = 'tool-call-1';
const action = { action: 'plug_96c522c2e2_click', call_id: callId, plugin_id: 'open-computer-use',
source_id: 'computer-window', args: { app: 'Demo', x: 100, y: 200 } };
const trace = { type: 'tool', name: action.action, trace_id: 'different-trace-uuid', timestamp: 109,
payload: { plugin_id: action.plugin_id, call_id: callId, source_id: action.source_id,
args: action.args, post_image_name: 'computer-shot' } };
const frames = extractStepReplayFrames([{ step_id: 'step-4', step_number: 4, timestamp: 109,
action_taken: action, generic_tools: [trace], post_image_name: 'computer-shot' }]);
expect(frames.length).toBe(1);
expect(frames[0].sourceId).toBe('computer-window');
expect(frames[0].postImageUrl).toBe('/images/computer-shot');
expect(frames[0].actionText).toBe('点击 ([100, 200])');
expect(frames[0].rawStepNumber).toBe(4);
});
it('preserves repeated same-source actions when they have different call ids', () => {
const trace = (callId: string, timestamp: number) => ({ type: 'tool', name: 'plug_96c522c2e2_click',
trace_id: `trace-${callId}`, timestamp, payload: { plugin_id: 'desktop', call_id: callId,
source_id: 'computer-window', args: { x: 100, y: 200 }, post_image_name: `shot-${callId}` } });
const first = trace('one', 101);
const second = trace('two', 102);
const frames = extractStepReplayFrames([{ step_id: 'multi', step_number: 1, timestamp: 101,
action_taken: first, generic_tools: [first, second] }]);
expect(frames.length).toBe(2);
expect(frames.map(frame => frame.timestamp)).toEqual([101, 102]);
});
});
describe('action-formatter.util screenshot chaining', () => {
for (const update of [
{ pre_image_name: null },
{ pre_screenshot: null },
{ last_execution_result: { pre_image_valid: false } }
]) {
it(`clears invalidated before-image aliases in merged step updates: ${JSON.stringify(update)}`, () => {
const frames = extractStepReplayFrames([
{ type: 'step_recorded', data: {
step_id: 'cross-window', step_number: 1, timestamp: 100,
pre_image_name: 'window-a', pre_screenshot: 'legacy-window-a',
action_taken: { action: 'click' }
} },
{ type: 'step_updated', data: {
step_id: 'cross-window', post_image_name: 'window-b', ...update
} },
{ type: 'step_updated', data: { step_id: 'cross-window', summary: 'Completed' } }
]);
expect(frames.length).toBe(1);
expect(frames[0].preImageUrl).toBeNull();
expect(frames[0].postImageUrl).toBe('/images/window-b');
expect(frames[0].imageUrl).toBe('/images/window-b');
});
}
it('retains a valid before-image when an update omits image fields', () => {
const frames = extractStepReplayFrames([
{ step_id: 'same-window', step_number: 1, timestamp: 100,
pre_image_name: 'before', action_taken: { action: 'click' } },
{ step_id: 'same-window', summary: 'Completed', post_image_name: 'after' }
]);
expect(frames[0].preImageUrl).toBe('/images/before');
expect(frames[0].postImageUrl).toBe('/images/after');
});
it('keeps distinct windows from the same plugin as distinct evidence and replay frames', () => {
const first = { type: 'tool', trace_id: 'w1', name: 'plug_click', timestamp: 101,
payload: { plugin_id: 'desktop', source_id: 'window-1', post_image_name: 'one' } };
const second = { type: 'tool', trace_id: 'w2', name: 'plug_click', timestamp: 102,
payload: { plugin_id: 'desktop', source_id: 'window-2', post_image_name: 'two' } };
const step = { step_id: 'multi', step_number: 1, timestamp: 101, action_taken: first, generic_tools: [first, second] };
expect(getStepPostImageUrl(step, first)).toBe('/images/one');
expect(getStepPostImageUrl(step, second)).toBe('/images/two');
const frames = extractStepReplayFrames([step]);
expect(frames.map(frame => frame.sourceId)).toEqual(['window-1', 'window-2']);
expect(frames.map(frame => frame.postImageUrl)).toEqual(['/images/one', '/images/two']);
});
it('keeps plugin observations separate from the phone image chain', () => {
const desktop = { type: 'tool', trace_id: 'desktop', name: 'plug_observe', timestamp: 1001,
payload: { plugin_id: 'desktop-control', post_image_name: 'desktop-screen' } };
@@ -17,6 +17,7 @@
import { ActionParam, StepReplayFrame } from '../core/models/stream.model';
import { extractNumbersFromCoordinateValue, isPureDirectionString, parseSequenceCoordinates, unwrapTraceAction } from './image-overlay.util';
import { cleanErrorMessage, joinTargetDescriptions } from './tool-formatter.util';
import { getPluginToolView } from './plugin-tool-view.util';
/**
* Safely parse JSON string if it looks like an object/array, otherwise return original value
@@ -166,6 +167,8 @@ export function getActionIcon(action: any): string {
export function getActionTitle(action: any): string {
const act = getActionObject(action);
if (!act) return "操作";
const plugin = getPluginToolView(act);
if (plugin) return plugin.title;
const name = (act.name || act.action || '').toLowerCase();
switch (name) {
case 'tap':
@@ -233,6 +236,11 @@ export function getActionTargetText(action: any): string {
const name = (act.name || act.action || '').toLowerCase();
const args = act.args && typeof act.args === 'object' ? act.args : {};
if (getPluginToolView(act)) {
return act.target_text || act.target_description || args.target_text || args.target_description
|| act.app_name || act.app || args.app_name || args.app || '';
}
if (name.includes('app')) {
const rawApp = act.app_name || act.package_name || act.app || args.app_name || args.package_name || args.app || '';
const app = rawApp ? (rawApp.charAt(0).toUpperCase() + rawApp.slice(1)) : '';
@@ -321,6 +329,7 @@ export function getActionCoords(action: any): string {
if (!act) return '';
const args = act.args && typeof act.args === 'object' ? act.args : {};
const name = (act.name || act.action || '').toLowerCase();
const plugin = getPluginToolView(act) !== null;
// Check sequence first if action is sequence or sequence arg is present
const isSequenceAction = name === 'click_sequence' || name === 'tap_sequence' || Boolean(act.sequence || args.sequence || act.normalized_sequence || args.normalized_sequence);
@@ -344,15 +353,18 @@ export function getActionCoords(action: any): string {
}
const normCoords = extractNumbersFromCoordinateValue(act.normalized_coordinates || args.normalized_coordinates);
const point = [act.x ?? args.x, act.y ?? args.y];
const explicitPoint = point.every(value => typeof value === 'number' && Number.isFinite(value)) ? point : null;
const coords = normCoords ||
explicitPoint ||
extractNumbersFromCoordinateValue(act.coordinates) ||
extractNumbersFromCoordinateValue(args.coordinates) ||
extractNumbersFromCoordinateValue(act.coords) ||
extractNumbersFromCoordinateValue(args.coords) ||
extractNumbersFromCoordinateValue(act.target) ||
extractNumbersFromCoordinateValue(args.target) ||
extractNumbersFromCoordinateValue(args.action) ||
extractNumbersFromCoordinateValue(act.action) ||
extractNumbersFromCoordinateValue(plugin ? null : args.action) ||
extractNumbersFromCoordinateValue(plugin ? null : act.action) ||
extractNumbersFromCoordinateValue(args.gesture) ||
extractNumbersFromCoordinateValue(act.gesture);
@@ -613,6 +625,29 @@ function pluginSource(item: any): string | null {
return act?.plugin_id || payload?.plugin_id || payload?.action?.plugin_id || result?.plugin_id || null;
}
function toolCallIdentity(item: any): string {
const act = getActionObject(item);
const payload = safeParseJson(act?.payload);
const result = safeParseJson(act?.result || payload?.result);
return String(act?.call_id || payload?.call_id || result?.call_id || act?.trace_id || '');
}
export function actionEvidenceSource(item: any): { id: string; label: string } {
const act = getActionObject(item);
const payload = safeParseJson(act?.payload) || {};
const result = safeParseJson(act?.result || payload?.result) || {};
const source = act?.source || payload?.source || payload?.action?.source || result?.source;
const plugin = pluginSource(item);
const android = !item || isAndroidAction(act) || isReportStatusAction(act);
const recording = act?.recording || result?.recording;
const recordingLabel = recording?.status === 'recording' ? (recording.scope === 'screen' ? ' · 屏幕录制中' : ' · 窗口录制中') : recording?.status === 'failed' ? ' · 录像异常' : '';
const labels: Record<string, string> = { computer: '电脑', android: '安卓', terminal: '终端', api: '接口', file: '文件', other: '插件' };
return {
id: act?.source_id || payload?.source_id || payload?.action?.source_id || result?.source_id || (plugin ? `plugin:${plugin}` : android ? 'android' : 'auxiliary'),
label: (source ? (labels[source.kind] || '插件') + (source.label ? ` · ${source.label}` : '') : (plugin ? `插件 · ${plugin}` : android ? '安卓' : '辅助')) + recordingLabel
};
}
export function extractItemPreImage(item: any): string | null {
const act = getActionObject(item);
if (!act) return null;
@@ -792,7 +827,7 @@ export function resolveStepImageMap(stepData: any): Map<any, ResolvedImageEntry>
}
// Ensure stepData.action_taken cross-resolves to the action images
if (stepData.action_taken && lastActionItem && pluginSource(stepData.action_taken) === pluginSource(lastActionItem)) {
if (stepData.action_taken && lastActionItem && actionEvidenceSource(stepData.action_taken).id === actionEvidenceSource(lastActionItem).id) {
const lastEntry = map.get(lastActionItem);
if (lastEntry) {
map.set(stepData.action_taken, lastEntry);
@@ -905,19 +940,23 @@ export function extractStepReplayFrames(logsOrSteps: any[]): StepReplayFrame[] {
const mergedStepNum = (stepNum !== null) ? stepNum : existing.step_number;
const mergedStepId = existing.step_id || step.step_id;
// Explicit invalidation must clear both aliases, otherwise an old window
// screenshot can reappear through the legacy pre_screenshot fallback.
const clearPreImage = safeParseJson(step.last_execution_result)?.pre_image_valid === false ||
(!preferExistingAction && (step.pre_image_name === null || step.pre_screenshot === null));
const merged = {
...existing,
...step,
step_id: mergedStepId,
step_number: mergedStepNum,
pre_image_name: preferExistingAction
pre_image_name: clearPreImage ? null : preferExistingAction
? existing.pre_image_name
: (step.pre_image_name || existing.pre_image_name),
post_image_name: preferExistingAction
? (existing.post_image_name || step.post_image_name)
: (step.post_image_name || existing.post_image_name),
pre_screenshot: preferExistingAction
pre_screenshot: clearPreImage ? null : preferExistingAction
? existing.pre_screenshot
: (step.pre_screenshot || existing.pre_screenshot),
post_screenshot: preferExistingAction
@@ -978,6 +1017,20 @@ export function extractStepReplayFrames(logsOrSteps: any[]): StepReplayFrame[] {
primaryImg
});
}
// A Pro step may contain actions from several targets. Each tool retains its
// own frame and timestamp instead of borrowing the step's last phone image.
const seen = new Set<string>([toolCallIdentity(act)]);
for (const tool of stepData.generic_tools || []) {
if (!pluginSource(tool)) continue;
const key = toolCallIdentity(tool);
if (tool === act || (key && seen.has(key))) continue;
if (key) seen.add(key);
const toolPre = getStepPreImageUrl(stepData, tool);
const toolPost = getStepPostImageUrl(stepData, tool);
if (!toolPre && !toolPost) continue;
visualCandidates.push({ stepData: { ...stepData, timestamp: tool.timestamp || stepData.timestamp }, act: tool,
preUrl: toolPre, postUrl: toolPost, primaryImg: toolPost || toolPre! });
}
}
// Sort strictly by original step_number ascending, then by timestamp
@@ -1004,6 +1057,8 @@ export function extractStepReplayFrames(logsOrSteps: any[]): StepReplayFrame[] {
: stepNum;
frames.push({
sourceId: actionEvidenceSource(act).id,
sourceLabel: actionEvidenceSource(act).label,
index: i,
stepNumber: stepNum,
rawStepNumber: rawStepNum,
@@ -0,0 +1,71 @@
import { actionEvidenceSource, getStepPostImageUrl, getStepPreImageUrl } from './action-formatter.util';
import { isDeviceActionTool } from './tool-formatter.util';
import { getPluginToolView } from './plugin-tool-view.util';
describe('plugin evidence timeline cards', () => {
const observation = () => ({
name: 'plug_abcdef1234_get_app_state', type: 'tool', status: 'success',
payload: {
plugin_id: 'desktop', tool_name: 'get_app_state', effect: 'read', observation: true,
args: { app: 'Demo' }, source_id: 'plugin:desktop:window-1',
source: { kind: 'computer', label: 'Demo' }, post_image_name: 'desktop-screen',
pre_image_valid: false,
result: { plugin_id: 'desktop', execution: 'completed', side_effect: 'none',
content: [{ type: 'text', text: '当前窗口已截图' }] }
}
});
it('renders real dynamic plugin observations with computer evidence, not Android actions', () => {
const tool = observation();
expect(isDeviceActionTool(tool)).toBeFalse();
expect(getPluginToolView(tool)).toEqual(jasmine.objectContaining({
title: '观察电脑窗口', kind: '辅助观察', auxiliary: true,
status: '已执行', result: '当前窗口已截图', args: [{ key: 'app', value: 'Demo' }]
}));
expect(actionEvidenceSource(tool)).toEqual({ id: 'plugin:desktop:window-1', label: '电脑 · Demo' });
const step = { pre_image_name: 'phone-before', post_image_name: 'phone-after', generic_tools: [tool] };
expect(getStepPreImageUrl(step, tool)).toBeNull();
expect(getStepPostImageUrl(step, tool)).toContain('desktop-screen');
});
it('uses the declared write contract even before a tool result arrives', () => {
const view = getPluginToolView({ name: 'plug_abcdef1234_click', type: 'tool', status: 'running',
payload: { plugin_id: 'desktop', tool_name: 'click', effect: 'write', observation: false,
args: { x: 30, y: 40 } } });
expect(view).toEqual(jasmine.objectContaining({ title: '点击', kind: '插件操作', auxiliary: false, status: '执行中' }));
});
it('keeps read-only file calls folded without labelling them as screen observations', () => {
const view = getPluginToolView({ name: 'plug_abcdef1234_read_file',
payload: { plugin_id: 'files', effect: 'read', observation: false, args: { path: 'result.txt' } } });
expect(view).toEqual(jasmine.objectContaining({ kind: '辅助读取', auxiliary: true, status: '待返回' }));
});
it('uses stored legacy results and keeps unavailable outcomes uncertain', () => {
const view = getPluginToolView({ name: 'plug_abcdef1234_click', type: 'tool', status: 'failed',
payload: JSON.stringify({ plugin_id: 'desktop', args: { x: 30 },
result: { tool: 'click', execution: 'unknown', side_effect: 'possible', message: '连接中断,需重新观察' } }) });
expect(view).toEqual(jasmine.objectContaining({ kind: '插件操作', auxiliary: false,
status: '结果不确定', result: '连接中断,需重新观察', args: [{ key: 'x', value: '30' }] }));
});
it('does not guess write intent from a dynamic tool name with missing metadata', () => {
expect(getPluginToolView({ name: 'plug_abcdef1234_click', status: 'running' }))
.toEqual(jasmine.objectContaining({ kind: '插件调用', auxiliary: true, args: [] }));
});
it('does not treat ordinary Android or model tools as plugin calls', () => {
expect(getPluginToolView({ name: 'click', type: 'tool', payload: { args: { x: 30 } } })).toBeNull();
expect(getPluginToolView({ name: 'plug_abcdef1234_click', type: 'llm_call' })).toBeNull();
});
it('reflects a completed receipt even when stream updates reuse the tool object', () => {
const tool: any = { name: 'plug_abcdef1234_click', status: 'running',
payload: { plugin_id: 'desktop', effect: 'write', args: {} } };
expect(getPluginToolView(tool)?.status).toBe('执行中');
tool.status = 'success';
tool.payload.result = { execution: 'completed', message: '已点击' };
expect(getPluginToolView(tool)?.status).toBe('已执行');
expect(getPluginToolView(tool)?.result).toBe('已点击');
});
});
@@ -0,0 +1,58 @@
export interface PluginToolView {
title: string;
kind: string;
auxiliary: boolean;
result: string;
status: string;
args: Array<{ key: string; value: string }>;
}
function objectValue(value: any): Record<string, any> {
if (typeof value === 'string') {
try { value = JSON.parse(value); } catch { return {}; }
}
return value && typeof value === 'object' && !Array.isArray(value) ? value : {};
}
/** Display plugin calls from their declared contract, never from Android tool names. */
export function getPluginToolView(tool: any): PluginToolView | null {
if (!tool || typeof tool !== 'object' || tool.type === 'llm_call') return null;
const payload = objectValue(tool.payload);
const result = objectValue(payload['result'] ?? tool.result);
const pluginId = payload['plugin_id'] || result['plugin_id'] || tool.plugin_id;
if (!pluginId && !/^plug_[a-f0-9]{10}_/.test(tool.name || tool.action || '')) {
return null;
}
const name = String(payload['tool_name'] || result['tool'] || tool.name || tool.action || '插件调用')
.replace(/^plug_[a-f0-9]{10}_/, '');
const effect = payload['effect'] || (result['side_effect'] === 'none' ? 'read' :
['possible', 'confirmed'].includes(result['side_effect']) ? 'write' : '');
const auxiliary = effect !== 'write';
const titles: Record<string, string> = {
get_app_state: '观察电脑窗口', list_apps: '列出电脑应用',
click: '点击', double_click: '双击', drag: '拖动', scroll: '滚动',
type_text: '输入文本', press_key: '按键', screenshot: '截取屏幕',
save_screenshot: '保存截图', launch_app: '打开应用'
};
const content = Array.isArray(result['content']) ? result['content'] : [];
const text = content.filter((item: any) => item?.type === 'text' && typeof item.text === 'string')
.map((item: any) => item.text).join('\n');
const state = result['execution'];
const status = tool.status === 'running' ? '执行中' :
result['status'] === 'uncertain' || state === 'unknown' ? '结果不确定' :
tool.status === 'failed' || tool.status === 'error' || result['status'] === 'error' ? '失败' :
state === 'partial' ? '部分完成' : state === 'not_started' ? '未执行' :
state === 'completed' || tool.status === 'success' ? '已执行' : '待返回';
const view: PluginToolView = {
title: titles[name] || name.replace(/_/g, ' '),
kind: payload['observation'] === true ? '辅助观察' : effect === 'read' ? '辅助读取' :
effect === 'write' ? '插件操作' : '插件调用',
auxiliary,
result: [result['message'], text].filter(Boolean).join('\n') || String(tool.error || ''),
status,
args: Object.entries(objectValue(payload['args'] ?? tool.args)).map(([key, value]) => ({
key, value: typeof value === 'string' ? value : JSON.stringify(value)
}))
};
return view;
}
@@ -47,6 +47,19 @@ export interface SegmentLocation {
const clamp = (value: number, min: number, max: number): number => Math.max(min, Math.min(max, value));
/** A real task-time seek must never silently jump across an unrecorded interval. */
export function locateCoveredSessionTime(segments: readonly TimelineSegment[], seconds: number): SegmentLocation | null {
if (!Number.isFinite(seconds)) return null;
const index = segments.findIndex(s => {
const start = s.offset_ms === undefined ? s.start : s.offset_ms / 1000;
return seconds >= start && seconds < start + s.duration;
});
if (index < 0) return null;
const s = segments[index];
const start = s.offset_ms === undefined ? s.start : s.offset_ms / 1000;
return { index, localTime: seconds - start, timelineTime: seconds };
}
/** True when every segment carries the session-relative v2 fields. */
export function hasSessionOffsets(segments: readonly TimelineSegment[]): boolean {
return segments.length > 0 && segments.every(
@@ -1,5 +1,5 @@
import { StepReplayFrame } from '../core/models/stream.model';
import { hasSessionOffsets, sessionTimeToTimelineTime, TimelineSegment } from './recording-timeline.util';
import { hasSessionOffsets, locateCoveredSessionTime, sessionTimeToTimelineTime, TimelineSegment } from './recording-timeline.util';
export interface VideoStepMarker {
time: number;
@@ -13,7 +13,8 @@ export function buildVideoStepMarkers(
frames: readonly StepReplayFrame[],
sessionStart: number | undefined,
segments: readonly TimelineSegment[],
duration: number
duration: number,
sessionTimeline = false
): VideoStepMarker[] {
const start = Number(sessionStart);
if (!Number.isFinite(start) || !Number.isFinite(duration) || duration <= 0) return [];
@@ -31,7 +32,8 @@ export function buildVideoStepMarkers(
const sessionTime = timestamp - start;
// Do not create misleading ticks at the start/end for steps outside the recording.
if (sessionTime < firstOffset || sessionTime > lastOffset) return [];
const time = sessionTimeToTimelineTime(segments, sessionTime);
if (sessionTimeline && !locateCoveredSessionTime(segments, sessionTime)) return [];
const time = sessionTimeline ? sessionTime : sessionTimeToTimelineTime(segments, sessionTime);
if (time < 0 || time > duration) return [];
return [{ time, frame }];
}).sort((a, b) => a.time - b.time);