更新

app/src/doctor_workstation/services/mock_repository.py

@@ -5,7 +5,7 @@ from __future__ import annotations
 import hashlib
 import struct
 import zlib
-from collections.abc import Mapping
+from collections.abc import Mapping, Sequence
 from copy import deepcopy
 from datetime import date, datetime, timedelta
 from os import PathLike
@@ -163,8 +163,27 @@ class DemoDoctorRepository:
                     "room_id": "demo-room-501",
                     "status": 2,
                     "status_text": "已结束",
-                    "recording_status_text": "录制完成",
-                    "recording_urls_list": ["https://media.example.invalid/demo/diagnosis-501.mp4"],
+                    "recording_status_text": "录制完成",
+                    "recording_urls_list": ["https://media.example.invalid/demo/diagnosis-501.mp4"],
+                    "transcription_status": "completed",
+                    "transcription_status_text": "文字已生成",
+                    "transcript_text": "患者：最近睡眠比上周好一些。\n医生：继续记录睡眠和空腹血糖。",
+                    "transcript_segments": [
+                        {
+                            "segment_id": "demo-1",
+                            "speaker_role": "patient",
+                            "speaker_user_id": "patient_301",
+                            "timestamp": 1,
+                            "text": "最近睡眠比上周好一些。",
+                        },
+                        {
+                            "segment_id": "demo-2",
+                            "speaker_role": "doctor",
+                            "speaker_user_id": "doctor_1001",
+                            "timestamp": 2,
+                            "text": "继续记录睡眠和空腹血糖。",
+                        },
+                    ],
                     "start_time_text": f"{self._today.isoformat()} 09:10:00",
                     "end_time_text": f"{self._today.isoformat()} 09:22:00",
                     "duration_text": "12分00秒",
@@ -2233,8 +2252,12 @@ class DemoDoctorRepository:
                 "room_id": "manual-upload",
                 "status": 2,
                 "status_text": "已结束",
-                "recording_status_text": "待上传",
-                "recording_urls_list": [],
+                "recording_status_text": "待上传",
+                "recording_urls_list": [],
+                "transcription_status": "not_started",
+                "transcription_status_text": "未生成文字",
+                "transcript_text": "",
+                "transcript_segments": [],
                 "start_time_text": datetime.now().replace(microsecond=0).isoformat(sep=" "),
                 "end_time_text": datetime.now().replace(microsecond=0).isoformat(sep=" "),
                 "duration_text": "0秒",
@@ -2531,8 +2554,12 @@ class DemoDoctorRepository:
             replay = {
                 **record,
                 "status_text": "呼叫中",
-                "recording_status_text": "未录制",
-                "recording_urls_list": [],
+                "recording_status_text": "未录制",
+                "recording_urls_list": [],
+                "transcription_status": "not_started",
+                "transcription_status_text": "未生成文字",
+                "transcript_text": "",
+                "transcript_segments": [],
                 "start_time_text": datetime.now().replace(microsecond=0).isoformat(sep=" "),
                 "end_time_text": "",
                 "duration_text": "—",
@@ -2567,7 +2594,7 @@ class DemoDoctorRepository:
                 )
             return deepcopy(record)
 
-    def bind_call_room(self, diagnosis_id: int, room_id: str) -> dict[str, Any]:
+    def bind_call_room(self, diagnosis_id: int, room_id: str) -> dict[str, Any]:
         """Persist the room identifier on the active demo call."""
 
         if not room_id.strip():
@@ -2594,11 +2621,133 @@ class DemoDoctorRepository:
                         "status_text": "通话中",
                     }
                 )
-            return {
-                "diagnosis_id": diagnosis_id,
-                "room_id": room_id.strip(),
-                "cloud_recording": {"started": False, "message": "演示模式不录制"},
-            }
+            return {
+                "diagnosis_id": diagnosis_id,
+                "room_id": room_id.strip(),
+                "cloud_recording": {"started": False, "message": "演示模式不录制"},
+            }
+
+    def _demo_call_record(
+        self, diagnosis_id: int, call_record_id: int | str
+    ) -> dict[str, Any]:
+        record = next(
+            (
+                row
+                for row in self._call_records.get(diagnosis_id, [])
+                if str(row.get("id") or row.get("call_record_id") or "")
+                == str(call_record_id)
+            ),
+            None,
+        )
+        if record is None:
+            raise RepositoryNotFoundError(f"call record {call_record_id} not found")
+        return record
+
+    def start_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        *,
+        language: str = "zh-CN",
+    ) -> dict[str, Any]:
+        """Start an in-memory transcript attached to one demo call record."""
+
+        clean_session = transcription_session_id.strip()
+        if not clean_session:
+            raise ValueError("transcription_session_id is required")
+        with self._lock:
+            self._find_consultation(diagnosis_id)
+            record = self._demo_call_record(diagnosis_id, call_record_id)
+            existing = str(record.get("transcription_session_id") or "")
+            if existing and existing != clean_session:
+                raise ValueError("another transcription session already exists")
+            record.update(
+                {
+                    "transcription_session_id": clean_session,
+                    "transcription_language": language.strip() or "zh-CN",
+                    "transcription_status": "running",
+                    "transcription_status_text": "录音转写中",
+                    "transcript_text": "",
+                    "transcript_segments": [],
+                }
+            )
+            return deepcopy(record)
+
+    def upsert_call_transcript_segments(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        segments: Sequence[Mapping[str, Any]],
+    ) -> dict[str, Any]:
+        """Idempotently append/update completed segments in demo mode."""
+
+        with self._lock:
+            record = self._demo_call_record(diagnosis_id, call_record_id)
+            if record.get("transcription_session_id") != transcription_session_id:
+                raise ValueError("transcription session does not match the call record")
+            stored = record.setdefault("transcript_segments", [])
+            for incoming in segments:
+                segment_id = str(incoming.get("segment_id") or "").strip()
+                text = str(incoming.get("text") or "").strip()
+                if not segment_id or not text:
+                    raise ValueError("transcript segment_id and text are required")
+                normalized = {
+                    "segment_id": segment_id,
+                    "speaker_role": str(incoming.get("speaker_role") or "unknown"),
+                    "speaker_user_id": str(incoming.get("speaker_user_id") or ""),
+                    "timestamp": int(incoming.get("timestamp") or 0),
+                    "text": text,
+                }
+                current = next(
+                    (row for row in stored if row.get("segment_id") == segment_id), None
+                )
+                if current is None:
+                    stored.append(normalized)
+                else:
+                    current.update(normalized)
+            stored.sort(key=lambda row: (int(row.get("timestamp") or 0), row["segment_id"]))
+            labels = {"doctor": "医生", "patient": "患者", "unknown": "未知说话人"}
+            record["transcript_text"] = "\n".join(
+                f"{labels.get(str(row.get('speaker_role')), '未知说话人')}：{row['text']}"
+                for row in stored
+            )
+            return deepcopy(record)
+
+    def finish_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        expected_segment_count: int,
+        *,
+        status: str = "completed",
+    ) -> dict[str, Any]:
+        """Finalize demo transcript fields on the same call record."""
+
+        clean_status = status.strip().lower()
+        if clean_status not in {"completed", "partial", "failed"}:
+            raise ValueError("transcription status is invalid")
+        with self._lock:
+            record = self._demo_call_record(diagnosis_id, call_record_id)
+            if record.get("transcription_session_id") != transcription_session_id:
+                raise ValueError("transcription session does not match the call record")
+            actual_count = len(record.get("transcript_segments") or [])
+            final_status = clean_status
+            if clean_status == "completed" and actual_count < expected_segment_count:
+                final_status = "partial"
+            record["transcription_status"] = final_status
+            record["transcription_status_text"] = {
+                "completed": "文字已生成",
+                "partial": "文字部分保存",
+                "failed": "文字生成失败",
+            }[final_status]
+            record["transcription_segment_count"] = actual_count
+            record["transcription_finished_at"] = datetime.now().replace(
+                microsecond=0
+            ).isoformat(sep=" ")
+            return deepcopy(record)
 
     def my_self(self) -> Session:
         """Compatibility alias for :meth:`get_session`."""

app/src/doctor_workstation/services/repository.py

@@ -5,7 +5,7 @@ from __future__ import annotations
 import mimetypes
 import re
 import time
-from collections.abc import Mapping
+from collections.abc import Mapping, Sequence
 from contextlib import suppress
 from datetime import date
 from io import BytesIO
@@ -619,8 +619,38 @@ class DoctorRepository(Protocol):
     def end_call(self, diagnosis_id: int) -> Any:
         """End the active diagnosis call."""
 
-    def bind_call_room(self, diagnosis_id: int, room_id: str) -> Any:
-        """Bind a TRTC room to the active call."""
+    def bind_call_room(self, diagnosis_id: int, room_id: str) -> Any:
+        """Bind a TRTC room to the active call."""
+
+    def start_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        *,
+        language: str = "zh-CN",
+    ) -> Any:
+        """Start a transcript stored on one exact call record."""
+
+    def upsert_call_transcript_segments(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        segments: Sequence[Mapping[str, Any]],
+    ) -> Any:
+        """Idempotently persist completed transcript segments."""
+
+    def finish_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        expected_segment_count: int,
+        *,
+        status: str = "completed",
+    ) -> Any:
+        """Finalize and materialize the transcript text on a call record."""
 
     def my_self(self) -> Session:
         """Compatibility alias for :meth:`get_session`."""
@@ -2171,32 +2201,157 @@ class RemoteDoctorRepository:
             ticket.diagnosis_id = diagnosis_id
         return ticket
 
-    def start_call(self, diagnosis_id: int, patient_id: int, *, call_type: int = 2) -> Any:
-        """Create the server-side call record before ringing participants."""
-
-        return self.client.post(
-            "tcm.diagnosis/startCall",
-            {
-                "diagnosis_id": diagnosis_id,
-                "patient_id": patient_id,
-                "call_type": call_type,
-            },
-        )
+    def start_call(self, diagnosis_id: int, patient_id: int, *, call_type: int = 2) -> Any:
+        """Create the server-side call record before ringing participants."""
+
+        payload = self.client.post(
+            "tcm.diagnosis/startCall",
+            {
+                "diagnosis_id": diagnosis_id,
+                "patient_id": patient_id,
+                "call_type": call_type,
+            },
+        )
+        if not isinstance(payload, Mapping):
+            raise ApiProtocolError(
+                "tcm.diagnosis/startCall returned no valid call_record_id object", data=payload
+            )
+        raw_id = next(
+            (
+                payload[key]
+                for key in ("call_record_id", "callRecordId", "id")
+                if key in payload
+            ),
+            None,
+        )
+        try:
+            if isinstance(raw_id, bool):
+                raise ValueError
+            call_record_id = int(raw_id)
+        except (TypeError, ValueError) as exc:
+            raise ApiProtocolError(
+                "tcm.diagnosis/startCall returned no valid call_record_id",
+                data=dict(payload),
+            ) from exc
+        if call_record_id <= 0:
+            raise ApiProtocolError(
+                "tcm.diagnosis/startCall returned no valid call_record_id",
+                data=dict(payload),
+            )
+        return {"call_record_id": call_record_id}
 
     def end_call(self, diagnosis_id: int) -> Any:
         """End the active call/recording associated with a diagnosis."""
 
         return self.client.post("tcm.diagnosis/endCall", {"diagnosis_id": diagnosis_id})
 
-    def bind_call_room(self, diagnosis_id: int, room_id: str) -> Any:
-        """Bind the actual TRTC room to the active call record."""
+    def bind_call_room(self, diagnosis_id: int, room_id: str) -> Any:
+        """Bind the actual TRTC room to the active call record."""
 
         if not room_id.strip():
             raise ValueError("room_id is required")
-        return self.client.post(
-            "tcm.diagnosis/bindCallRoom",
-            {"diagnosis_id": diagnosis_id, "room_id": room_id.strip()},
-        )
+        return self.client.post(
+            "tcm.diagnosis/bindCallRoom",
+            {"diagnosis_id": diagnosis_id, "room_id": room_id.strip()},
+        )
+
+    @staticmethod
+    def _transcription_identity(
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+    ) -> dict[str, Any]:
+        if diagnosis_id <= 0:
+            raise ValueError("diagnosis_id must be positive")
+        if isinstance(call_record_id, bool) or not str(call_record_id).strip():
+            raise ValueError("call_record_id is required")
+        clean_session = transcription_session_id.strip()
+        if not clean_session or len(clean_session) > 128:
+            raise ValueError("transcription_session_id must contain 1 to 128 characters")
+        return {
+            "diagnosis_id": diagnosis_id,
+            "call_record_id": call_record_id,
+            "transcription_session_id": clean_session,
+        }
+
+    def start_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        *,
+        language: str = "zh-CN",
+    ) -> Any:
+        """Create the server transcript session for one call record."""
+
+        body = self._transcription_identity(
+            diagnosis_id, call_record_id, transcription_session_id
+        )
+        body["language"] = language.strip()[:32] or "zh-CN"
+        return self.client.post("tcm.diagnosis/startCallTranscription", body)
+
+    def upsert_call_transcript_segments(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        segments: Sequence[Mapping[str, Any]],
+    ) -> Any:
+        """Upsert a bounded batch using each segment_id as the idempotency key."""
+
+        body = self._transcription_identity(
+            diagnosis_id, call_record_id, transcription_session_id
+        )
+        normalized: list[dict[str, Any]] = []
+        if len(segments) > 50:
+            raise ValueError("at most 50 transcript segments may be submitted at once")
+        for segment in segments:
+            segment_id = str(segment.get("segment_id") or "").strip()
+            text = str(segment.get("text") or "").strip()
+            if not segment_id or len(segment_id) > 160:
+                raise ValueError("transcript segment_id is invalid")
+            if not text or len(text) > 4_000:
+                raise ValueError("transcript text is invalid")
+            normalized.append(
+                {
+                    "segment_id": segment_id,
+                    "speaker_user_id": str(segment.get("speaker_user_id") or "")[:160],
+                    "speaker_role": str(segment.get("speaker_role") or "unknown")[:20],
+                    "timestamp": max(int(segment.get("timestamp") or 0), 0),
+                    "text": text,
+                }
+            )
+        if not normalized:
+            raise ValueError("at least one transcript segment is required")
+        body["segments"] = normalized
+        return self.client.post("tcm.diagnosis/upsertCallTranscriptSegments", body)
+
+    def finish_call_transcription(
+        self,
+        diagnosis_id: int,
+        call_record_id: int | str,
+        transcription_session_id: str,
+        expected_segment_count: int,
+        *,
+        status: str = "completed",
+    ) -> Any:
+        """Finalize a transcript; repeated requests use the same session identity."""
+
+        body = self._transcription_identity(
+            diagnosis_id, call_record_id, transcription_session_id
+        )
+        clean_status = status.strip().lower()
+        if clean_status not in {"completed", "partial", "failed"}:
+            raise ValueError("transcription status is invalid")
+        if expected_segment_count < 0:
+            raise ValueError("expected_segment_count must not be negative")
+        body.update(
+            {
+                "expected_segment_count": expected_segment_count,
+                "status": clean_status,
+            }
+        )
+        return self.client.post("tcm.diagnosis/finishCallTranscription", body)
 
     # Compatibility aliases keep UI naming independent from endpoint history.
     def my_self(self) -> Session:

app/src/doctor_workstation/ui/dialogs/diagnosis.py

@@ -20,6 +20,7 @@ from PySide6.QtWidgets import (
     QInputDialog,
     QLabel,
     QMessageBox,
+    QPlainTextEdit,
     QPushButton,
     QScrollArea,
     QSizePolicy,
@@ -1411,7 +1412,7 @@ class DiagnosisDialog(QDialog):
         layout.setContentsMargins(16, 14, 16, 18)
         layout.setSpacing(12)
         toolbar = QHBoxLayout()
-        hint = QLabel("每条通话记录可包含多个回放地址，并可追加指定记录的视频。")
+        hint = QLabel("每条通话记录可包含多个回放地址；录音转写完成后可查看本次面诊对话文字。")
         hint.setObjectName("DiagnosisDialogGuidance")
         hint.setWordWrap(True)
         toolbar.addWidget(hint, 1)
@@ -1553,7 +1554,7 @@ class DiagnosisDialog(QDialog):
                 ("房间号", 180),
                 ("时长", 110),
                 ("状态", 90),
-                ("录制", 100),
+                ("录制 / 文字", 120),
                 ("操作", 130),
             ),
             "assign": (
@@ -3365,11 +3366,71 @@ class DiagnosisDialog(QDialog):
                 cell.stop()
         self._inline_recording_cells.clear()
 
+    @staticmethod
+    def _call_transcript_text(row: Any) -> str:
+        raw = first_value(
+            row,
+            "transcript_text",
+            "call_transcript",
+            "conversation_text",
+            "transcript",
+            default="",
+        )
+        if isinstance(raw, Mapping):
+            raw = first_value(raw, "text", "full_text", "content", default="")
+        text = str(raw or "").strip()
+        if text:
+            return text
+        segments = first_value(row, "transcript_segments", "segments", default=[]) or []
+        if isinstance(segments, Sequence) and not isinstance(
+            segments, (str, bytes, bytearray)
+        ):
+            labels = {"doctor": "医生", "patient": "患者", "unknown": "未知说话人"}
+            lines: list[str] = []
+            for segment in segments:
+                content = str(first_value(segment, "text", "sourceText", default="") or "").strip()
+                if not content:
+                    continue
+                role = str(
+                    first_value(segment, "speaker_role", "speakerRole", default="unknown")
+                    or "unknown"
+                )
+                lines.append(f"{labels.get(role, '未知说话人')}：{content}")
+            return "\n".join(lines)
+        return ""
+
+    def _view_call_transcript(self, call_record_id: int, transcript: str) -> None:
+        dialog = QDialog(self)
+        dialog.setObjectName("DiagnosisCallTranscriptDialog")
+        dialog.setWindowTitle(f"面诊对话文字 · 通话记录 #{call_record_id}")
+        dialog.resize(680, 520)
+        layout = QVBoxLayout(dialog)
+        layout.setContentsMargins(20, 18, 20, 18)
+        layout.setSpacing(12)
+        title = QLabel("录音转写对话")
+        title.setObjectName("DiagnosisDailySectionTitle")
+        layout.addWidget(title)
+        hint = QLabel("以下内容由语音自动转写，仅作为面诊记录辅助，请由医生核对。")
+        hint.setObjectName("DiagnosisDialogGuidance")
+        hint.setWordWrap(True)
+        layout.addWidget(hint)
+        content = QPlainTextEdit(dialog)
+        content.setObjectName("DiagnosisCallTranscriptText")
+        content.setReadOnly(True)
+        content.setPlainText(transcript)
+        layout.addWidget(content, 1)
+        close_button = QPushButton("关闭")
+        close_button.setProperty("variant", "primary")
+        close_button.clicked.connect(dialog.accept)
+        layout.addWidget(close_button, 0, Qt.AlignmentFlag.AlignRight)
+        dialog.exec()
+
     def _fill_video(self, rows: Sequence[Any]) -> None:
         self._stop_inline_recordings()
         matrix: list[tuple[Any, ...]] = []
         row_urls: list[list[str]] = []
         record_ids: list[int] = []
+        transcripts: list[str] = []
         for row in rows:
             raw_urls = first_value(row, "recording_urls_list", "recording_urls", default=[]) or []
             if isinstance(raw_urls, str) or not isinstance(raw_urls, Sequence):
@@ -3382,6 +3443,8 @@ class DiagnosisDialog(QDialog):
                     normalized.append(url)
             row_urls.append(normalized)
             record_ids.append(_int(first_value(row, "id", "call_record_id"), 0))
+            transcript = self._call_transcript_text(row)
+            transcripts.append(transcript)
             call_type = _raw_value(row, "call_type_text")
             if call_type is _MISSING:
                 raw_call_type = _raw_value(row, "call_type")
@@ -3398,6 +3461,15 @@ class DiagnosisDialog(QDialog):
                     3: "未接听",
                     4: "已取消",
                 }.get(_int(raw_status, -1), "—")
+            recording_status = first_value(
+                row, "recording_status_text", "record_status_text", "record_status"
+            )
+            transcript_status = first_value(
+                row,
+                "transcription_status_text",
+                "transcript_status_text",
+                default="文字已生成" if transcript else "未生成文字",
+            )
             matrix.append(
                 (
                     "" if normalized else "暂无录制回放",
@@ -3407,9 +3479,7 @@ class DiagnosisDialog(QDialog):
                     first_value(row, "room_id", "room_no"),
                     first_value(row, "duration_text", "duration"),
                     status,
-                    first_value(
-                        row, "recording_status_text", "record_status_text", "record_status"
-                    ),
+                    f"{display_text(recording_status)}\n{display_text(transcript_status)}",
                     "",
                 )
             )
@@ -3434,6 +3504,20 @@ class DiagnosisDialog(QDialog):
                     if item is not None:
                         item.setText("")
                 call_record_id = record_ids[row_index]
+                actions: list[QPushButton] = []
+                transcript = transcripts[row_index]
+                if transcript and call_record_id > 0:
+                    view_transcript = self._action_button(
+                        "查看文字", f"查看通话记录 #{call_record_id} 的录音转写"
+                    )
+                    view_transcript.setObjectName("DiagnosisVideoTranscriptView")
+                    view_transcript.setProperty("callRecordId", call_record_id)
+                    view_transcript.clicked.connect(
+                        lambda _checked=False, selected=call_record_id, text=transcript: self._view_call_transcript(
+                            selected, text
+                        )
+                    )
+                    actions.append(view_transcript)
                 if self._editable and self._can_video_upload and call_record_id > 0:
                     upload = self._action_button(
                         "追加回放", f"上传并绑定通话记录 #{call_record_id}"
@@ -3452,12 +3536,16 @@ class DiagnosisDialog(QDialog):
                             selected
                         )
                     )
-                    upload_host = QWidget()
-                    upload_host.setObjectName("DiagnosisVideoRowUploadCell")
-                    upload_layout = QVBoxLayout(upload_host)
-                    upload_layout.setContentsMargins(0, 0, 0, 0)
-                    upload_layout.addWidget(upload, 0, Qt.AlignmentFlag.AlignCenter)
-                    table.setCellWidget(row_index, 8, upload_host)
+                    actions.append(upload)
+                if actions:
+                    action_host = QWidget()
+                    action_host.setObjectName("DiagnosisVideoRowUploadCell")
+                    action_layout = QVBoxLayout(action_host)
+                    action_layout.setContentsMargins(0, 0, 0, 0)
+                    action_layout.setSpacing(2)
+                    for action in actions:
+                        action_layout.addWidget(action, 0, Qt.AlignmentFlag.AlignCenter)
+                    table.setCellWidget(row_index, 8, action_host)
                     upload_item = table.item(row_index, 8)
                     if upload_item is not None:
                         upload_item.setText("")

app/src/doctor_workstation/video/launcher.py

@@ -130,6 +130,7 @@ def _ticket_mapping(ticket: Any) -> Mapping[str, Any]:
         "patientUserId": "patient_user_id",
         "diagnosisId": "diagnosis_id",
         "patientId": "patient_id",
+        "callRecordId": "call_record_id",
     }
     adapted = {
         json_name: getattr(ticket, attribute_name)
@@ -196,6 +197,7 @@ class VideoCallRequest:
     target_user_id: str
     diagnosis_id: Identifier
     patient_id: Identifier | None = None
+    call_record_id: Identifier | None = None
     backend_mode: BackendMode = BackendMode.EMBEDDED
 
     def __post_init__(self) -> None:
@@ -218,6 +220,12 @@ class VideoCallRequest:
                 "patient_id",
                 _identifier(self.patient_id, "patientId"),
             )
+        if self.call_record_id is not None:
+            object.__setattr__(
+                self,
+                "call_record_id",
+                _identifier(self.call_record_id, "callRecordId"),
+            )
         object.__setattr__(self, "backend_mode", BackendMode.parse(self.backend_mode))
 
     @classmethod
@@ -257,6 +265,7 @@ class VideoCallRequest:
         return {
             "diagnosis_id": self.diagnosis_id,
             "patient_id": self.patient_id,
+            "call_record_id": self.call_record_id,
             "backend_mode": self.backend_mode.value,
         }
 
@@ -285,6 +294,13 @@ def normalize_backend_ticket(
         _identifier,
         required=False,
     )
+    payload_call_record = _read_aliases(
+        payload,
+        ("callRecordId", "call_record_id"),
+        "callRecordId",
+        _identifier,
+        required=False,
+    )
 
     normalized_diagnosis = _merge_identifier(
         payload_diagnosis,
@@ -323,6 +339,7 @@ def normalize_backend_ticket(
         ),
         diagnosis_id=normalized_diagnosis,
         patient_id=normalized_patient,
+        call_record_id=payload_call_record,
         backend_mode=BackendMode.parse(backend_mode),
     )
 

app/src/doctor_workstation/video/lifecycle.py

@@ -74,6 +74,71 @@ def _call_repository_method(method: Callable[..., Any], payload: Mapping[str, An
     return _resolve_result(result)
 
 
+def _mapping_candidate(value: Any) -> Mapping[str, Any] | None:
+    if isinstance(value, Mapping):
+        return value
+    for attribute in ("raw", "data"):
+        candidate = getattr(value, attribute, None)
+        if isinstance(candidate, Mapping):
+            return candidate
+    return None
+
+
+def _extract_call_record_id(result: Any) -> int | str | None:
+    """Read a positive call-record identity from common backend envelopes."""
+
+    pending = [result]
+    visited: set[int] = set()
+    while pending:
+        candidate = pending.pop(0)
+        mapping = _mapping_candidate(candidate)
+        if mapping is None or id(mapping) in visited:
+            continue
+        visited.add(id(mapping))
+        for key in ("call_record_id", "callRecordId", "id"):
+            value = mapping.get(key)
+            if isinstance(value, bool) or value is None:
+                continue
+            if isinstance(value, int) and value > 0:
+                return value
+            if isinstance(value, str) and value.strip():
+                return value.strip()
+        for key in ("data", "result", "record", "call_record", "callRecord"):
+            nested = mapping.get(key)
+            if isinstance(nested, Mapping):
+                pending.append(nested)
+    return None
+
+
+def _clean_transcript_segment(segment: Mapping[str, Any], session_id: str) -> dict[str, Any]:
+    segment_id = str(segment.get("segment_id", segment.get("segmentId", ""))).strip()
+    text = str(segment.get("text", segment.get("sourceText", ""))).strip()
+    speaker_user_id = str(
+        segment.get("speaker_user_id", segment.get("speakerUserId", ""))
+    ).strip()
+    speaker_role = str(segment.get("speaker_role", segment.get("speakerRole", "unknown"))).strip()
+    if not segment_id or len(segment_id) > 160:
+        raise ValueError("transcript segment_id must contain 1 to 160 characters")
+    if not text or len(text) > 4_000:
+        raise ValueError("transcript text must contain 1 to 4000 characters")
+    if len(speaker_user_id) > 160:
+        raise ValueError("transcript speaker_user_id is too long")
+    if speaker_role not in {"doctor", "patient", "unknown"}:
+        speaker_role = "unknown"
+    try:
+        timestamp = int(segment.get("timestamp") or 0)
+    except (TypeError, ValueError):
+        timestamp = 0
+    return {
+        "segment_id": segment_id,
+        "transcription_session_id": session_id,
+        "speaker_user_id": speaker_user_id,
+        "speaker_role": speaker_role,
+        "timestamp": max(timestamp, 0),
+        "text": text,
+    }
+
+
 @dataclass(slots=True)
 class _WorkItem:
     operation: str
@@ -169,11 +234,17 @@ class OrderedCallLifecycle:
         self.logger = logger
         self.started = False
         self.ended = False
+        self.call_record_id: int | str | None = request.call_record_id
         self.bound_room_id: str | None = None
         self._claimed_room_id: str | None = None
         self._start_future: Future[bool] | None = None
         self._bind_future: Future[bool] | None = None
         self._end_future: Future[bool] | None = None
+        self._transcription_start_future: Future[bool] | None = None
+        self._transcription_finish_future: Future[bool] | None = None
+        self._transcription_session_id: str | None = None
+        self._transcription_active = False
+        self._segment_futures: dict[str, Future[bool]] = {}
         self._lock = threading.RLock()
         self._worker = _OrderedDaemonWorker(logger, request.safe_log_context())
 
@@ -181,6 +252,10 @@ class OrderedCallLifecycle:
     def worker_is_daemon(self) -> bool:
         return self._worker.is_daemon
 
+    @property
+    def transcription_session_id(self) -> str | None:
+        return self._transcription_session_id
+
     def start(self) -> Future[bool]:
         with self._lock:
             if self._start_future is not None:
@@ -196,9 +271,13 @@ class OrderedCallLifecycle:
                 payload["patient_id"] = self.request.patient_id
 
             def operation() -> bool:
-                _call_repository_method(method, payload)
+                result = _call_repository_method(method, payload)
+                record_id = _extract_call_record_id(result)
+                if record_id is None:
+                    raise ValueError("startCall response did not include the current call_record_id")
                 with self._lock:
                     self.started = True
+                    self.call_record_id = record_id
                 self.logger.info(
                     "video call record started",
                     extra={"video_call": self.request.safe_log_context()},
@@ -309,10 +388,167 @@ class OrderedCallLifecycle:
 
             return self._worker.submit("screenshot", operation)
 
+    def start_transcription(self, session_id: str, *, language: str = "zh-CN") -> Future[bool]:
+        """Start persisted realtime transcription for this exact call record."""
+
+        clean_session = str(session_id or "").strip()
+        clean_language = str(language or "zh-CN").strip()[:32] or "zh-CN"
+        if not clean_session or len(clean_session) > 128:
+            raise ValueError("transcription session id must contain 1 to 128 characters")
+        with self._lock:
+            if self._end_future is not None:
+                raise RuntimeError("video call has already ended")
+            if self._transcription_start_future is not None:
+                if self._transcription_session_id != clean_session:
+                    raise RuntimeError("another transcription session already exists")
+                return self._transcription_start_future
+            if self._start_future is None:
+                self.start()
+            method = getattr(self.repository, "start_call_transcription", None)
+            if not callable(method):
+                raise ValueError("video repository does not implement call transcription storage")
+            self._transcription_session_id = clean_session
+
+            def operation() -> bool:
+                with self._lock:
+                    started = self.started
+                    record_id = self.call_record_id
+                if not started:
+                    return False
+                if record_id is None:
+                    raise ValueError("server did not return the current call_record_id")
+                _call_repository_method(
+                    method,
+                    {
+                        "diagnosis_id": self.request.diagnosis_id,
+                        "call_record_id": record_id,
+                        "transcription_session_id": clean_session,
+                        "language": clean_language,
+                    },
+                )
+                with self._lock:
+                    self._transcription_active = True
+                self.logger.info(
+                    "video call transcription started",
+                    extra={"video_call": self.request.safe_log_context()},
+                )
+                return True
+
+            self._transcription_start_future = self._worker.submit(
+                "transcription-start", operation
+            )
+            return self._transcription_start_future
+
+    def save_transcript_segment(self, segment: Mapping[str, Any]) -> Future[bool]:
+        """Upsert one completed, bounded transcript segment without logging its text."""
+
+        with self._lock:
+            session_id = self._transcription_session_id
+            if not session_id or self._transcription_start_future is None:
+                raise RuntimeError("call transcription has not started")
+            if self._transcription_finish_future is not None:
+                raise RuntimeError("call transcription has already stopped")
+            cleaned = _clean_transcript_segment(segment, session_id)
+            segment_id = cleaned["segment_id"]
+            existing = self._segment_futures.get(segment_id)
+            if existing is not None:
+                return existing
+            method = getattr(self.repository, "upsert_call_transcript_segments", None)
+            if not callable(method):
+                raise ValueError("video repository does not implement transcript segment storage")
+
+            def operation() -> bool:
+                with self._lock:
+                    active = self._transcription_active
+                    record_id = self.call_record_id
+                if not active:
+                    return False
+                if record_id is None:
+                    raise ValueError("server did not return the current call_record_id")
+                _call_repository_method(
+                    method,
+                    {
+                        "diagnosis_id": self.request.diagnosis_id,
+                        "call_record_id": record_id,
+                        "transcription_session_id": session_id,
+                        "segments": [cleaned],
+                    },
+                )
+                return True
+
+            future = self._worker.submit("transcription-segment", operation)
+            self._segment_futures[segment_id] = future
+
+            def release_failed(completed: Future[bool]) -> None:
+                if completed.cancelled() or completed.exception() is not None:
+                    with self._lock:
+                        if self._segment_futures.get(segment_id) is completed:
+                            self._segment_futures.pop(segment_id, None)
+
+            future.add_done_callback(release_failed)
+            return future
+
+    def finish_transcription(self, *, status: str = "completed") -> Future[bool]:
+        """Flush and finalize the current transcript exactly once."""
+
+        clean_status = str(status or "completed").strip().lower()
+        if clean_status not in {"completed", "partial", "failed"}:
+            raise ValueError("transcription status is invalid")
+        with self._lock:
+            if self._transcription_finish_future is not None:
+                return self._transcription_finish_future
+            if self._transcription_start_future is None or not self._transcription_session_id:
+                return _settled_future(False)
+            method = getattr(self.repository, "finish_call_transcription", None)
+            if not callable(method):
+                raise ValueError("video repository does not implement transcription finalization")
+            session_id = self._transcription_session_id
+
+            def operation() -> bool:
+                with self._lock:
+                    active = self._transcription_active
+                    record_id = self.call_record_id
+                    expected_count = len(self._segment_futures)
+                if not active:
+                    return False
+                if record_id is None:
+                    raise ValueError("server did not return the current call_record_id")
+                _call_repository_method(
+                    method,
+                    {
+                        "diagnosis_id": self.request.diagnosis_id,
+                        "call_record_id": record_id,
+                        "transcription_session_id": session_id,
+                        "expected_segment_count": expected_count,
+                        "status": clean_status,
+                    },
+                )
+                with self._lock:
+                    self._transcription_active = False
+                self.logger.info(
+                    "video call transcription finalized",
+                    extra={
+                        "video_call": self.request.safe_log_context(),
+                        "transcription_status": clean_status,
+                        "segment_count": expected_count,
+                    },
+                )
+                return True
+
+            self._transcription_finish_future = self._worker.submit(
+                "transcription-finish", operation
+            )
+            return self._transcription_finish_future
+
     def end(self, reason: str) -> Future[bool]:
         with self._lock:
             if self._end_future is not None:
                 return self._end_future
+            if (
+                self._transcription_start_future is not None
+                and self._transcription_finish_future is None
+            ):
+                self.finish_transcription(status="partial")
             method = getattr(self.repository, "end_call", None)
 
             def operation() -> bool:

app/src/doctor_workstation/video/window.py

@@ -210,6 +210,7 @@ if WEBENGINE_AVAILABLE:  # pragma: no cover - GUI behavior needs an integration
         call_error = Signal(str)  # type: ignore[misc]
         _start_completed = Signal(bool)  # type: ignore[misc]
         _screenshot_completed = Signal(bool, str)  # type: ignore[misc]
+        _transcription_completed = Signal(str, str, str, bool, str)  # type: ignore[misc]
 
         def __init__(
             self,
@@ -284,6 +285,7 @@ if WEBENGINE_AVAILABLE:  # pragma: no cover - GUI behavior needs an integration
 
             self._start_completed.connect(self._on_lifecycle_started)
             self._screenshot_completed.connect(self._on_screenshot_completed)
+            self._transcription_completed.connect(self._on_transcription_completed)
             self.web_view.loadFinished.connect(self._on_load_finished)
             self.web_view.setUrl(QUrl(self.location.url))
 
@@ -426,6 +428,21 @@ if WEBENGINE_AVAILABLE:  # pragma: no cover - GUI behavior needs an integration
                     str(message.get("message") or "截屏图片无效。")[:200],
                 )
                 return
+            if event == "transcription-start-request":
+                self._start_transcription(
+                    str(message.get("sessionId") or ""),
+                    str(message.get("language") or "zh-CN"),
+                )
+                return
+            if event == "transcription-segment":
+                self._save_transcript_segment(message)
+                return
+            if event == "transcription-stop":
+                self._finish_transcription(
+                    str(message.get("sessionId") or ""),
+                    str(message.get("status") or "completed"),
+                )
+                return
             room_id = message.get("roomId", message.get("room_id"))
             if room_id not in (None, ""):
                 self.lifecycle.bind_room(room_id)
@@ -434,8 +451,6 @@ if WEBENGINE_AVAILABLE:  # pragma: no cover - GUI behavior needs an integration
             if event == "status":
                 status = str(message.get("status", "unknown"))[:80]
                 self.status_changed.emit(status)
-                if status == "idle" and not self.open_im:
-                    self._close_from_companion("remote-idle")
             elif event == "hangup":
                 status = str(message.get("status", "ended"))[:80]
                 self.call_ended.emit(status)
@@ -509,6 +524,109 @@ if WEBENGINE_AVAILABLE:  # pragma: no cover - GUI behavior needs an integration
                 f"window.doctorConsultation?.screenshotResult?.({state}, {payload});"
             )
 
+        def _notify_transcription_completed(
+            self,
+            operation: str,
+            session_id: str,
+            segment_id: str,
+            future: Future[bool],
+        ) -> None:
+            try:
+                succeeded = bool(future.result())
+            except Exception as error:
+                succeeded = False
+                message = str(error)[:200] or "录音文字保存失败。"
+            else:
+                message = {
+                    "start": "录音文字存储已准备。",
+                    "segment": "",
+                    "stop": "本次面诊对话文字已保存。",
+                }.get(operation, "")
+            with suppress(RuntimeError):
+                self._transcription_completed.emit(
+                    operation, session_id, segment_id, succeeded, message
+                )
+
+        def _on_transcription_completed(
+            self,
+            operation: str,
+            session_id: str,
+            segment_id: str,
+            succeeded: bool,
+            message: str,
+        ) -> None:
+            payload = json.dumps(str(message)[:200], ensure_ascii=True)
+            state = "true" if succeeded else "false"
+            self._page.runJavaScript(
+                "window.doctorConsultation?.transcriptionResult?.("
+                f"{json.dumps(operation)}, {json.dumps(session_id)}, "
+                f"{json.dumps(segment_id)}, {state}, {payload});"
+            )
+
+        def _start_transcription(self, session_id: str, language: str) -> None:
+            try:
+                future = self.lifecycle.start_transcription(session_id, language=language)
+            except Exception as error:
+                self._on_transcription_completed(
+                    "start", session_id, "", False, str(error)[:200]
+                )
+                return
+            future.add_done_callback(
+                lambda completed: self._notify_transcription_completed(
+                    "start", session_id, "", completed
+                )
+            )
+
+        def _save_transcript_segment(self, message: Mapping[str, Any]) -> None:
+            session_id = str(message.get("sessionId") or "").strip()
+            segment = message.get("segment")
+            segment_id = (
+                str(segment.get("segment_id") or segment.get("segmentId") or "").strip()
+                if isinstance(segment, Mapping)
+                else ""
+            )
+            if not isinstance(segment, Mapping):
+                self._on_transcription_completed(
+                    "segment", session_id, segment_id, False, "录音文字片段无效。"
+                )
+                return
+            if session_id != str(self.lifecycle.transcription_session_id or ""):
+                self._on_transcription_completed(
+                    "segment", session_id, segment_id, False, "录音会话标识不匹配。"
+                )
+                return
+            try:
+                future = self.lifecycle.save_transcript_segment(segment)
+            except Exception as error:
+                self._on_transcription_completed(
+                    "segment", session_id, segment_id, False, str(error)[:200]
+                )
+                return
+            future.add_done_callback(
+                lambda completed: self._notify_transcription_completed(
+                    "segment", session_id, segment_id, completed
+                )
+            )
+
+        def _finish_transcription(self, session_id: str, status: str) -> None:
+            if session_id.strip() != str(self.lifecycle.transcription_session_id or ""):
+                self._on_transcription_completed(
+                    "stop", session_id, "", False, "录音会话标识不匹配。"
+                )
+                return
+            try:
+                future = self.lifecycle.finish_transcription(status=status)
+            except Exception as error:
+                self._on_transcription_completed(
+                    "stop", session_id, "", False, str(error)[:200]
+                )
+                return
+            future.add_done_callback(
+                lambda completed: self._notify_transcription_completed(
+                    "stop", session_id, "", completed
+                )
+            )
+
         def _close_from_companion(self, reason: str) -> None:
             self._companion_ended = True
             self._close_reason = reason