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

apps/admin_console/routers/media.py

@@ -115,6 +115,74 @@ async def get_session_video(session_id: str):
 
 
 def _get_session_video_sync(session_id: str):
+    from apps.admin_console.services.evidence_service import plugin_sources
+    from apps.admin_console.services.report_export import session_videos
+
+    legacy = _get_legacy_session_video_sync(session_id)
+    sources = []
+    if legacy.get("video_segments") or legacy.get("video_url"):
+        segments = legacy.get("video_segments") or []
+        # Pauses create sibling resume directories. Only collect files already
+        # attached to this task; unrelated recordings never enter its timeline.
+        if segments and all("offset_ms" in segment for segment in segments):
+            combined = {}
+            for path, _ in session_videos({"id": session_id}):
+                for segment in media_service.resolve_video_segments(
+                    media_service.path_to_video_url(path)
+                ):
+                    if "offset_ms" in segment:
+                        combined[segment["url"]] = segment
+            if combined:
+                segments = sorted(combined.values(), key=lambda segment: segment["offset_ms"])
+        sources.append(
+            {
+                "id": "android",
+                "kind": "android",
+                "label": "安卓",
+                "target_ids": ["android"],
+                "status": "ready",
+                "message": "",
+                "segments": [
+                    {**s, "start": float(s.get("offset_ms", s["start"] * 1000)) / 1000}
+                    for s in segments
+                ],
+                "url": legacy.get("video_url"),
+            }
+        )
+    sources += plugin_sources(session_id)
+    legacy["sources"] = sources
+    legacy["session_timeline"] = any(source["segments"] for source in sources)
+    row = dict(session_repo.get_session_by_id(session_id) or {})
+    legacy["session_duration"] = max(
+        max(
+            0.0,
+            float(row.get("end_time") or row.get("start_time") or 0)
+            - float(row.get("start_time") or 0),
+        ),
+        max(
+            (s["start"] + s["duration"] for source in sources for s in source["segments"]),
+            default=0,
+        ),
+    )
+    ready = next((s for s in sources if s.get("segments") or s.get("url")), None)
+    if ready:
+        legacy.update(
+            status="ready",
+            has_video=True,
+            video_url=ready.get("url") or ready["segments"][0]["url"],
+            video_segments=ready["segments"],
+        )
+    elif any(s.get("status") == "recording" for s in sources):
+        legacy.update(status="processing", retry_after_ms=1000)
+    elif sources:
+        legacy.update(
+            status="failed",
+            message="；".join(str(s.get("message") or "录像不可用") for s in sources),
+        )
+    return legacy
+
+
+def _get_legacy_session_video_sync(session_id: str):
     video_rec_map = session_repo.get_video_recordings_map()
     video_idx = media_service.build_video_index()
     row = session_repo.get_session_by_id(session_id)

apps/admin_console/services/evidence_service.py

@@ -0,0 +1,151 @@
+"""Read task-owned plugin evidence without executing tools or converting media."""
+
+import json
+import math
+from pathlib import Path
+import re
+
+from artemis.config import TRACES_PATH
+
+
+def task_root(session_id: str) -> Path:
+    if not re.fullmatch(r"[A-Za-z0-9_-]{1,100}", session_id):
+        raise ValueError("无效任务编号")
+    root = TRACES_PATH.resolve() / session_id
+    if root.resolve() != root:
+        raise ValueError("任务证据目录不能是符号链接")
+    return root
+
+
+def evidence_file(root: Path, relative: str) -> Path | None:
+    path = root / relative
+    try:
+        if (
+            path.is_file()
+            and path.resolve() == path.absolute()
+            and path.resolve().is_relative_to(root.resolve())
+        ):
+            return path
+    except OSError:
+        return None
+    return None
+
+
+def plugin_sources(session_id: str) -> list[dict]:
+    root = task_root(session_id)
+    path = evidence_file(root, "evidence/recordings/recordings.json")
+    if not path or path.stat().st_size > 4 * 1024 * 1024:
+        return []
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+        if data.get("version") != 1 or not isinstance(data.get("sources"), list):
+            return []
+        from apps.admin_console.services.media_service import MediaService
+
+        sources = []
+        for item in data["sources"]:
+            if not isinstance(item, dict) or not isinstance(item.get("id"), str):
+                continue
+            source = {**item, "segments": []}
+            for segment in item.get("segments", []):
+                if not isinstance(segment, dict):
+                    continue
+                video = evidence_file(root, str(segment.get("file") or ""))
+                if video is None or video.suffix.lower() != ".mp4":
+                    continue
+                duration = float(segment.get("duration", 0))
+                start = float(segment.get("offset_ms", 0)) / 1000
+                if (
+                    not math.isfinite(duration)
+                    or duration <= 0
+                    or not math.isfinite(start)
+                    or start < 0
+                ):
+                    continue
+                source["segments"].append(
+                    {
+                        **segment,
+                        "url": MediaService.path_to_video_url(video),
+                        "start": start,
+                        "duration": duration,
+                    }
+                )
+            sources.append(source)
+        return sources
+    except (OSError, ValueError, TypeError, KeyError):
+        return []
+
+
+def plugin_files(session_id: str) -> list[tuple[Path, str]]:
+    root = task_root(session_id)
+    files = {}
+    for source in plugin_sources(session_id):
+        for segment in source["segments"]:
+            path = evidence_file(root, segment["file"])
+            if path:
+                files[path] = "任务录屏/电脑/" + path.name
+    for directory in (root / "artifacts" / "plugins", root / "evidence"):
+        if not directory.is_dir():
+            continue
+        for path in directory.rglob("*"):
+            if path.suffix.lower() == ".tmp" or (
+                directory.name == "evidence" and path.suffix.lower() in {".mp4", ".mkv"}
+            ):
+                continue
+            relative = path.relative_to(root).as_posix()
+            checked = evidence_file(root, relative)
+            if checked:
+                files[checked] = (
+                    "任务文件/插件/" + path.relative_to(root / "artifacts" / "plugins").as_posix()
+                    if directory.name == "plugins"
+                    else "证据索引/" + path.relative_to(root / "evidence").as_posix()
+                )
+    return list(files.items())
+
+
+def export_index(session_id: str, report: dict, android_videos: list) -> dict:
+    from apps.admin_console.services.report_export import session_artifacts, session_screenshots
+
+    extras = plugin_files(session_id)
+    recordings = plugin_sources(session_id)
+    root = task_root(session_id)
+    paths = {path.resolve(): name for path, name in extras}
+    for source in recordings:
+        for segment in source["segments"]:
+            segment["file"] = paths.get((root / segment["file"]).resolve())
+            segment.pop("url", None)
+    images = session_screenshots(report)
+    artifacts = session_artifacts(report)
+    files = [(path, "步骤截图/" + name) for path, name in images]
+    files += [(path, "任务录屏/" + name) for path, name in android_videos]
+    files += [(path, "任务文件/安卓/" + name) for path, name in artifacts]
+    files += extras
+    calls = []
+    ledger = evidence_file(root, "evidence/calls.jsonl")
+    if ledger:
+        with ledger.open(encoding="utf-8") as handle:
+            for line in handle:
+                try:
+                    call = json.loads(line)
+                    if not isinstance(call, dict):
+                        continue
+                    for attachment in call.get("attachments", []):
+                        relative = str(attachment.get("path") or "")
+                        attachment["file"] = paths.get((root / relative).resolve())
+                    calls.append(call)
+                except (ValueError, TypeError, AttributeError):
+                    continue  # A partial final line after a crash is not a valid receipt.
+    return {
+        "version": 1,
+        "session_id": session_id,
+        "name": report["name"],
+        "status": report["status"],
+        "verdict": report["verdict"],
+        "steps": report.get("steps", []),
+        "calls": calls,
+        "recordings": recordings,
+        "android_videos": ["任务录屏/" + name for _, name in android_videos],
+        "image_references": {path.stem: "步骤截图/" + name for path, name in images},
+        "files": [{"path": name, "bytes": path.stat().st_size} for path, name in files],
+        "note": "工具执行完成不等于用例通过。录像保留真实时间与缺口；各来源截图不能相互替代。",
+    }

apps/admin_console/services/model_settings_service.py

@@ -175,6 +175,7 @@ class CapabilitySettings(BaseModel):
     )
     video_enabled: bool = False
     recording_enabled: bool = False
+    recording: dict = Field(default_factory=dict)
     outputter_enabled: bool = True
     step_summarizer_enabled: bool = True
     screenshot_limit_enabled: bool = True
@@ -204,6 +205,7 @@ def get_capability_settings() -> CapabilitySettings:
     return CapabilitySettings(
         video_enabled=settings.ARTEMIS_VIDEO_ENABLED,
         recording_enabled=settings.ARTEMIS_RECORDING_ENABLED,
+        recording=load_agent_config().recording.model_dump(),
         outputter_enabled=settings.ARTEMIS_OUTPUTTER_ENABLED,
         step_summarizer_enabled=settings.ARTEMIS_STEP_SUMMARIZER_ENABLED,
         screenshot_limit_enabled=screenshot.enabled,
@@ -212,6 +214,12 @@ def get_capability_settings() -> CapabilitySettings:
 
 
 def save_capability_settings(update: CapabilitySettings) -> CapabilitySettings:
+    from artemis.config.agent import RecordingConfig
+    import json
+
+    recording = RecordingConfig.model_validate(
+        {**load_agent_config().recording.model_dump(), **update.recording}
+    ).model_dump()
     values = {
         "ARTEMIS_ADB_AUTO_CONNECT": update.adb_auto_connect,
         "ARTEMIS_ADB_DISCOVERY_INTERVAL_SECONDS": update.adb_discovery_interval_seconds,
@@ -229,6 +237,8 @@ def save_capability_settings(update: CapabilitySettings) -> CapabilitySettings:
         "ARTEMIS_WORKER_CLEANUP_TIMEOUT_SECONDS": update.worker_cleanup_timeout_seconds,
     }
     persist_environment({key: str(value).lower() for key, value in values.items()})
+    persist_environment({"ARTEMIS_RECORDING_CONFIG": json.dumps(recording)})
+    settings.ARTEMIS_RECORDING_CONFIG = recording
     for key, value in values.items():
         setattr(settings, key, value)
     return get_capability_settings()

apps/admin_console/services/report_export.py

@@ -475,6 +475,24 @@ def plugin_operations(steps: list[dict]) -> list[dict]:
                 for block in result.get("content", [])
                 if isinstance(block, dict) and block.get("type") == "text"
             )
+            source = result.get("source") or payload.get("source") or {}
+            evidence = []
+            if result.get("call_id"):
+                evidence.append("调用：" + str(result["call_id"]))
+            if source:
+                evidence.append(
+                    f"来源：{source.get('kind', '')} · {source.get('label') or source.get('target_id', '')}"
+                )
+            if result.get("execution"):
+                evidence.append(
+                    f"执行：{result['execution']}；副作用：{result.get('side_effect', 'unknown')}；错误码：{result.get('error_code') or '无'}"
+                )
+            for attachment in result.get("attachments") or []:
+                evidence.append(
+                    f"文件：{attachment.get('name')}；SHA-256：{attachment.get('sha256')}"
+                )
+            if evidence:
+                content = "\n".join([content or result.get("message") or "", *evidence])
             operations.append(
                 {
                     "step": step.get("step_number"),
@@ -809,6 +827,8 @@ def session_artifacts(report: dict) -> list[tuple[Path, str]]:
 
 
 def export_session_media(session_id: str, kind: str) -> Path:
+    from apps.admin_console.services.evidence_service import export_index, plugin_files
+
     if kind not in {"screenshots", "video", "all"}:
         raise ValueError("不支持的下载类型")
     with EXPORT_LOCK:
@@ -818,7 +838,10 @@ def export_session_media(session_id: str, kind: str) -> Path:
         images = session_screenshots(report) if kind in {"screenshots", "all"} else []
         videos = session_videos(report) if kind in {"video", "all"} else []
         artifacts = session_artifacts(report) if kind == "all" else []
-        if kind == "video" and not videos:
+        extras = plugin_files(session_id) if kind in {"video", "all"} else []
+        if kind == "video":
+            extras = [(path, name) for path, name in extras if path.suffix.lower() == ".mp4"]
+        if kind == "video" and not videos and not extras:
             raise ValueError("此任务没有已保存的录屏")
         if kind == "screenshots" and not images:
             raise ValueError("此任务没有可下载的步骤截图")
@@ -830,7 +853,8 @@ def export_session_media(session_id: str, kind: str) -> Path:
             "status": report["status"],
             "verdict": report["verdict"],
             "screenshots": [name for _, name in images],
-            "videos": [name for _, name in videos],
+            "videos": ["任务录屏/" + name for _, name in videos]
+            + [name for path, name in extras if path.suffix.lower() == ".mp4"],
             "artifacts": [name for _, name in artifacts],
             "note": "录屏按时间顺序分段；暂停期间未占用设备，也未录制其他任务。",
         }
@@ -842,9 +866,27 @@ def export_session_media(session_id: str, kind: str) -> Path:
             for source, name in images:
                 archive.write(source, f"{prefix}步骤截图/{name}")
             for source, name in videos:
-                archive.write(source, f"{prefix}任务录屏/{name}")
+                archive.write(source, f"{prefix}任务录屏/{name}", compress_type=zipfile.ZIP_STORED)
             for source, name in artifacts:
                 archive.write(source, f"{prefix}任务文件/安卓/{name}")
+            for source, name in extras:
+                archive.write(
+                    source,
+                    prefix + name,
+                    compress_type=zipfile.ZIP_STORED
+                    if source.suffix.lower() == ".mp4"
+                    else zipfile.ZIP_DEFLATED,
+                )
+            if kind == "all":
+                archive.writestr(
+                    prefix + "证据索引.json",
+                    json.dumps(
+                        export_index(session_id, report, videos),
+                        ensure_ascii=False,
+                        indent=2,
+                        default=str,
+                    ),
+                )
             archive.writestr(
                 prefix + "文件清单.json", json.dumps(manifest, ensure_ascii=False, indent=2)
             )
@@ -853,6 +895,8 @@ def export_session_media(session_id: str, kind: str) -> Path:
 
 
 def export_batch(batch_id: str, progress: Callable[[int, int, str], None] | None = None) -> Path:
+    from apps.admin_console.services.evidence_service import export_index, plugin_files
+
     with EXPORT_LOCK:
         batch = store.get("batch", identifier(batch_id))
         if not batch:
@@ -901,9 +945,28 @@ def export_batch(batch_id: str, progress: Callable[[int, int, str], None] | None
                 for source, name in session_screenshots(report):
                     archive.write(source, f"{prefix}/步骤截图/{name}")
                 for source, name in session_videos(report):
-                    archive.write(source, f"{prefix}/任务录屏/{name}")
+                    archive.write(
+                        source, f"{prefix}/任务录屏/{name}", compress_type=zipfile.ZIP_STORED
+                    )
                 for source, name in session_artifacts(report):
                     archive.write(source, f"{prefix}/任务文件/安卓/{name}")
+                for source, name in plugin_files(report["id"]):
+                    archive.write(
+                        source,
+                        f"{prefix}/{name}",
+                        compress_type=zipfile.ZIP_STORED
+                        if source.suffix.lower() == ".mp4"
+                        else zipfile.ZIP_DEFLATED,
+                    )
+                archive.writestr(
+                    f"{prefix}/证据索引.json",
+                    json.dumps(
+                        export_index(report["id"], report, session_videos(report)),
+                        ensure_ascii=False,
+                        indent=2,
+                        default=str,
+                    ),
+                )
                 if progress:
                     progress(number, len(records), f"已打包任务 {number}/{len(records)}")
         if progress:

apps/admin_console/services/task_import.py

@@ -8,8 +8,9 @@ import zipfile
 from openpyxl import Workbook, load_workbook
 from openpyxl.styles import Alignment, Font, PatternFill
 from openpyxl.utils import get_column_letter
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, ConfigDict, Field, field_serializer
 from artemis.plugins.models import ExecutionTarget, PluginSelection
+from artemis.config.agent import RecordingConfig
 
 HEADERS = (
     "用例编号",
@@ -38,6 +39,7 @@ class TaskOptions(BaseModel):
     explorer_mode: str | None = Field(default=None, pattern="^(flash|pro|ultra)$")
     locked_app_package: str | None = Field(default=None, min_length=1, max_length=200)
     record_video: bool | None = None
+    recording: RecordingConfig | None = None
     analyze_video: bool | None = None
     request_timeout_seconds: float | None = Field(default=None, gt=0, le=3600)
     max_retries: int | None = Field(default=None, ge=0, le=10)
@@ -51,6 +53,11 @@ class TaskOptions(BaseModel):
     plugins: list[PluginSelection] | None = None
     execution_target: ExecutionTarget | None = None
 
+    @field_serializer("recording", when_used="unless-none")
+    def serialize_recording(self, value: RecordingConfig) -> dict:
+        # A per-task partial override must not replace other saved recording defaults.
+        return value.model_dump(exclude_unset=True)
+
 
 class TaskCase(BaseModel):
     model_config = ConfigDict(extra="forbid")

apps/admin_console/services/task_queue_service.py

@@ -586,6 +586,10 @@ class TaskQueueService:
         env["ARTEMIS_TASK_INGRESS"] = str(task_item.get("ingress", "frontend"))
         env["ARTEMIS_TASK_WORKER"] = "1"
         overrides = task_item.get("runtime_overrides") or {}
+        if overrides.get("recording") is not None:
+            env["ARTEMIS_TASK_RECORDING_CONFIG"] = json.dumps(overrides["recording"])
+        else:
+            env.pop("ARTEMIS_TASK_RECORDING_CONFIG", None)
         for option, variable in {
             "model": "OPENAI_MODEL",
             "reasoning_effort": "OPENAI_REASONING_EFFORT",
@@ -869,6 +873,10 @@ class TaskQueueService:
         registered at recording start; remuxing that file is all that is
         needed to publish the full video.
         """
+        from apps.admin_console.services.evidence_service import task_root
+        from artemis.plugins.recording import recover_interrupted_recordings
+
+        await recover_interrupted_recordings(task_root(str(sess_id)))
         rec_info = session_repo.get_video_recording_for_session(sess_id)
         rec_status = (rec_info or {}).get("status")
         if rec_status == "ready":
@@ -1890,6 +1898,14 @@ class TaskQueueService:
         recordings published.
         """
         recovered = 0
+        from artemis.config import TRACES_PATH
+        from artemis.plugins.recording import recover_interrupted_recordings
+
+        # This hook runs in the startup worker thread, outside the server event loop.
+        for manifest in TRACES_PATH.glob("*/evidence/recordings/recordings.json"):
+            root = manifest.parent.parent.parent
+            if asyncio.run(recover_interrupted_recordings(root)):
+                recovered += 1
         try:
             pending = session_repo.get_unfinalized_video_recordings()
         except Exception as exc: