feat: surgery pipeline API, video inference, voice confirm, and tests
- Add FastAPI routes for surgery start/end, results, pending confirmation (WAV upload), and health checks.
- Implement RTSP/Hikvision capture, consumable classification, session manager, MinIO/Baidu voice resolution, and DB persistence.
- Add documentation (client API, video backends, staging checklist) and sample camera/RTSP config.
- Add pytest suite (API contract, session manager, voice, repositories, pipeline persistence) and httpx dev dependency.
- Replace deprecated HTTP_422_UNPROCESSABLE_ENTITY with HTTP_422_UNPROCESSABLE_CONTENT.
- Fix SurgeryPipeline DB reads to use an explicit transaction with autobegin disabled.
Made-with: Cursor
2026-04-21 18:33:54 +08:00
|
|
|
|
from __future__ import annotations
|
|
|
|
|
|
|
|
|
|
|
|
from datetime import datetime
|
|
|
|
|
|
|
|
|
|
|
|
from pydantic import BaseModel, ConfigDict, Field
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class HealthResponse(BaseModel):
|
|
|
|
|
|
status: str
|
|
|
|
|
|
database: str
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryStartRequest(BaseModel):
|
|
|
|
|
|
model_config = ConfigDict(
|
|
|
|
|
|
json_schema_extra={
|
|
|
|
|
|
"example": {
|
|
|
|
|
|
"surgery_id": "123456",
|
|
|
|
|
|
"camera_ids": ["or-cam-01", "or-cam-02"],
|
|
|
|
|
|
"candidate_consumables": ["纱布", "缝线", "止血钳"],
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str = Field(
|
|
|
|
|
|
min_length=6,
|
|
|
|
|
|
max_length=6,
|
|
|
|
|
|
pattern=r"^\d{6}$",
|
|
|
|
|
|
description="手术6位号,只允许6位数字。",
|
|
|
|
|
|
)
|
|
|
|
|
|
camera_ids: list[str] = Field(
|
|
|
|
|
|
min_length=1,
|
|
|
|
|
|
description="本次手术需要接入的摄像头 ID 列表。",
|
|
|
|
|
|
)
|
|
|
|
|
|
candidate_consumables: list[str] = Field(
|
|
|
|
|
|
default_factory=list,
|
|
|
|
|
|
description=(
|
|
|
|
|
|
"本次手术可能使用到的耗材清单。"
|
|
|
|
|
|
"服务端仅对该清单内的耗材做自动记账与待确认追问;"
|
|
|
|
|
|
"若为空则不会写入任何消耗(仅拉流推理)。"
|
|
|
|
|
|
),
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryEndRequest(BaseModel):
|
|
|
|
|
|
model_config = ConfigDict(
|
|
|
|
|
|
json_schema_extra={"example": {"surgery_id": "123456"}}
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str = Field(
|
|
|
|
|
|
min_length=6,
|
|
|
|
|
|
max_length=6,
|
|
|
|
|
|
pattern=r"^\d{6}$",
|
|
|
|
|
|
description="手术6位号,只允许6位数字。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryApiResponse(BaseModel):
|
|
|
|
|
|
surgery_id: str = Field(description="手术6位号。")
|
|
|
|
|
|
status: str = Field(description="接口处理状态。")
|
|
|
|
|
|
message: str = Field(description="返回说明。")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryClientErrorDetail(BaseModel):
|
|
|
|
|
|
"""与 `HTTPException(detail={...})` 对应;最终 JSON 为 `{"detail": {...}}`。"""
|
|
|
|
|
|
|
|
|
|
|
|
code: str = Field(description="业务错误码,如 RECORDING_CANNOT_START、RECORDING_NOT_STOPPED、RESULT_NOT_READY。")
|
|
|
|
|
|
message: str = Field(description="人类可读说明。")
|
|
|
|
|
|
surgery_id: str = Field(description="手术 6 位号。")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryClientErrorResponse(BaseModel):
|
|
|
|
|
|
"""FastAPI/Starlette 对 HTTPException 序列化后的常见外形(`detail` 为对象时)。"""
|
|
|
|
|
|
|
|
|
|
|
|
detail: SurgeryClientErrorDetail
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryConsumptionDetail(BaseModel):
|
|
|
|
|
|
"""单条消耗明细(按事件发生,可能多行)。"""
|
|
|
|
|
|
|
|
|
|
|
|
item_id: str = Field(description="物品 ID。")
|
|
|
|
|
|
item_name: str = Field(description="物品名称。")
|
|
|
|
|
|
quantity: int = Field(ge=0, description="本条记录对应的消耗数量。")
|
|
|
|
|
|
doctor_id: str = Field(description="医生 ID。")
|
|
|
|
|
|
timestamp: datetime = Field(description="记录时间(ISO 8601)。")
|
|
|
|
|
|
source: str = Field(
|
|
|
|
|
|
default="vision",
|
|
|
|
|
|
description="记录来源:vision 自动识别;voice 语音确认。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryConsumptionSummary(BaseModel):
|
|
|
|
|
|
"""按物品汇总:该手术下该物品消耗数量合计。"""
|
|
|
|
|
|
|
|
|
|
|
|
item_id: str = Field(description="物品 ID。")
|
|
|
|
|
|
item_name: str = Field(description="物品名称。")
|
|
|
|
|
|
total_quantity: int = Field(ge=0, description="该物品在本台手术中的消耗数量合计。")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def build_consumption_summary(
|
|
|
|
|
|
details: list[SurgeryConsumptionDetail],
|
|
|
|
|
|
) -> list[SurgeryConsumptionSummary]:
|
|
|
|
|
|
"""按 item_id 汇总 total_quantity;名称取该物品首条出现时的 item_name。"""
|
|
|
|
|
|
totals: dict[str, tuple[str, int]] = {}
|
|
|
|
|
|
for row in details:
|
|
|
|
|
|
if row.item_id not in totals:
|
|
|
|
|
|
totals[row.item_id] = (row.item_name, 0)
|
|
|
|
|
|
name, acc = totals[row.item_id]
|
|
|
|
|
|
totals[row.item_id] = (name, acc + row.quantity)
|
|
|
|
|
|
return [
|
|
|
|
|
|
SurgeryConsumptionSummary(
|
|
|
|
|
|
item_id=iid,
|
|
|
|
|
|
item_name=name,
|
|
|
|
|
|
total_quantity=qty,
|
|
|
|
|
|
)
|
|
|
|
|
|
for iid, (name, qty) in sorted(totals.items(), key=lambda x: x[0])
|
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryVoiceStatusResponse(BaseModel):
|
|
|
|
|
|
"""手术进行中人工确认(客户端播报)联调状态。"""
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str = Field(description="手术 6 位号。")
|
|
|
|
|
|
voice_enabled: bool = Field(
|
|
|
|
|
|
description="是否启用了低置信度人工确认(客户端拉取待确认项)。",
|
|
|
|
|
|
)
|
|
|
|
|
|
pending_queue_approx: int = Field(
|
|
|
|
|
|
ge=0,
|
|
|
|
|
|
description="待医生确认的追问任务数量(FIFO 队列长度)。",
|
|
|
|
|
|
)
|
|
|
|
|
|
last_prompt_snippet: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="最近一次生成的待确认话术摘要。",
|
|
|
|
|
|
)
|
|
|
|
|
|
last_asr_text: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="最近一次语音确认接口产生的 ASR 文本。",
|
|
|
|
|
|
)
|
|
|
|
|
|
last_error: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="最近一次语音确认错误说明(如 ASR/解析失败)。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-04-23 14:24:20 +08:00
|
|
|
|
class SurgeryVoiceAuditItem(BaseModel):
|
|
|
|
|
|
"""单条 `voice_confirmation_audits` 行(追溯对账用)。"""
|
|
|
|
|
|
|
|
|
|
|
|
id: int
|
|
|
|
|
|
confirmation_id: str
|
|
|
|
|
|
status: str = Field(
|
|
|
|
|
|
description=(
|
|
|
|
|
|
"recognized / rejected / parse_failed / asr_failed / invalid_audio / "
|
|
|
|
|
|
"upload_failed / client_stt_empty / client_stt_parse_failed 等"
|
|
|
|
|
|
),
|
|
|
|
|
|
)
|
|
|
|
|
|
audio_object_key: str | None = None
|
|
|
|
|
|
audio_content_type: str | None = None
|
|
|
|
|
|
audio_size_bytes: int | None = None
|
|
|
|
|
|
audio_sha256: str | None = None
|
|
|
|
|
|
asr_text: str | None = None
|
|
|
|
|
|
resolved_label: str | None = None
|
|
|
|
|
|
options_snapshot_json: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="当次候选项与置信度 JSON 快照。",
|
|
|
|
|
|
)
|
|
|
|
|
|
error_message: str | None = None
|
|
|
|
|
|
created_at: datetime = Field(
|
|
|
|
|
|
description="记录写入时间(UTC)。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
model_config = ConfigDict(from_attributes=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryVoiceAuditsListResponse(BaseModel):
|
|
|
|
|
|
"""按手术号分页的语音确认审计列表。"""
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str
|
|
|
|
|
|
total: int = Field(ge=0, description="该手术在表中的总条数(不受本页 limit 截断)。")
|
|
|
|
|
|
limit: int = Field(ge=1, le=200)
|
|
|
|
|
|
offset: int = Field(ge=0)
|
|
|
|
|
|
items: list[SurgeryVoiceAuditItem] = Field(
|
|
|
|
|
|
default_factory=list,
|
|
|
|
|
|
description="按 `created_at` 降序。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
feat: surgery pipeline API, video inference, voice confirm, and tests
- Add FastAPI routes for surgery start/end, results, pending confirmation (WAV upload), and health checks.
- Implement RTSP/Hikvision capture, consumable classification, session manager, MinIO/Baidu voice resolution, and DB persistence.
- Add documentation (client API, video backends, staging checklist) and sample camera/RTSP config.
- Add pytest suite (API contract, session manager, voice, repositories, pipeline persistence) and httpx dev dependency.
- Replace deprecated HTTP_422_UNPROCESSABLE_ENTITY with HTTP_422_UNPROCESSABLE_CONTENT.
- Fix SurgeryPipeline DB reads to use an explicit transaction with autobegin disabled.
Made-with: Cursor
2026-04-21 18:33:54 +08:00
|
|
|
|
class PendingConfirmationOption(BaseModel):
|
|
|
|
|
|
label: str
|
|
|
|
|
|
confidence: float
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryPendingConfirmationResponse(BaseModel):
|
|
|
|
|
|
"""当前待医生确认的一条低置信度识别。"""
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str
|
|
|
|
|
|
confirmation_id: str
|
|
|
|
|
|
prompt_text: str = Field(description="可直接用于 TTS 播报的话术。")
|
|
|
|
|
|
options: list[PendingConfirmationOption]
|
|
|
|
|
|
model_top1_label: str = Field(description="模型原始 Top1 标签(可能不在候选清单内)。")
|
|
|
|
|
|
model_top1_confidence: float
|
|
|
|
|
|
created_at: datetime
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-04-23 14:24:20 +08:00
|
|
|
|
class SurgeryPendingResolveTextRequest(BaseModel):
|
|
|
|
|
|
"""由浏览器 Web Speech 等客户端本地识别后提交的文本,语义与经百度 ASR 得到的文本相同。"""
|
|
|
|
|
|
|
|
|
|
|
|
recognized_text: str = Field(
|
|
|
|
|
|
min_length=1,
|
|
|
|
|
|
max_length=2000,
|
|
|
|
|
|
description="识别文本;服务端用与语音接口相同的规则解析候选项。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
feat: surgery pipeline API, video inference, voice confirm, and tests
- Add FastAPI routes for surgery start/end, results, pending confirmation (WAV upload), and health checks.
- Implement RTSP/Hikvision capture, consumable classification, session manager, MinIO/Baidu voice resolution, and DB persistence.
- Add documentation (client API, video backends, staging checklist) and sample camera/RTSP config.
- Add pytest suite (API contract, session manager, voice, repositories, pipeline persistence) and httpx dev dependency.
- Replace deprecated HTTP_422_UNPROCESSABLE_ENTITY with HTTP_422_UNPROCESSABLE_CONTENT.
- Fix SurgeryPipeline DB reads to use an explicit transaction with autobegin disabled.
Made-with: Cursor
2026-04-21 18:33:54 +08:00
|
|
|
|
class SurgeryPendingConfirmationResolveResponse(BaseModel):
|
|
|
|
|
|
surgery_id: str
|
|
|
|
|
|
confirmation_id: str
|
|
|
|
|
|
status: str = Field(description="accepted")
|
|
|
|
|
|
message: str
|
|
|
|
|
|
resolved_label: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="解析并确认后的耗材名称;否认全部候选时为 null。",
|
|
|
|
|
|
)
|
|
|
|
|
|
rejected: bool = Field(
|
|
|
|
|
|
default=False,
|
|
|
|
|
|
description="是否为否认全部候选(不记消耗)。",
|
|
|
|
|
|
)
|
|
|
|
|
|
asr_text: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="服务端语音识别得到的文本。",
|
|
|
|
|
|
)
|
|
|
|
|
|
audio_object_key: str | None = Field(
|
|
|
|
|
|
default=None,
|
|
|
|
|
|
description="MinIO 中原始 WAV 的对象键,用于追溯。",
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SurgeryResultResponse(BaseModel):
|
|
|
|
|
|
model_config = ConfigDict(
|
|
|
|
|
|
json_schema_extra={
|
|
|
|
|
|
"example": {
|
|
|
|
|
|
"surgery_id": "123456",
|
|
|
|
|
|
"status": "completed",
|
|
|
|
|
|
"message": "结果查询成功。",
|
|
|
|
|
|
"details": [
|
|
|
|
|
|
{
|
|
|
|
|
|
"item_id": "HC001",
|
|
|
|
|
|
"item_name": "纱布",
|
|
|
|
|
|
"quantity": 2,
|
|
|
|
|
|
"doctor_id": "D1001",
|
|
|
|
|
|
"timestamp": "2026-04-21T10:30:00+08:00",
|
|
|
|
|
|
},
|
|
|
|
|
|
{
|
|
|
|
|
|
"item_id": "HC001",
|
|
|
|
|
|
"item_name": "纱布",
|
|
|
|
|
|
"quantity": 1,
|
|
|
|
|
|
"doctor_id": "D1002",
|
|
|
|
|
|
"timestamp": "2026-04-21T11:05:00+08:00",
|
|
|
|
|
|
},
|
|
|
|
|
|
{
|
|
|
|
|
|
"item_id": "HC002",
|
|
|
|
|
|
"item_name": "缝线",
|
|
|
|
|
|
"quantity": 1,
|
|
|
|
|
|
"doctor_id": "D1001",
|
|
|
|
|
|
"timestamp": "2026-04-21T10:45:00+08:00",
|
|
|
|
|
|
},
|
|
|
|
|
|
],
|
|
|
|
|
|
"summary": [
|
|
|
|
|
|
{"item_id": "HC001", "item_name": "纱布", "total_quantity": 3},
|
|
|
|
|
|
{"item_id": "HC002", "item_name": "缝线", "total_quantity": 1},
|
|
|
|
|
|
],
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
surgery_id: str = Field(description="手术6位号。")
|
|
|
|
|
|
status: str = Field(description="结果状态,例如 pending / completed / failed。")
|
|
|
|
|
|
message: str = Field(description="返回说明。")
|
|
|
|
|
|
details: list[SurgeryConsumptionDetail] = Field(
|
|
|
|
|
|
default_factory=list,
|
|
|
|
|
|
description="消耗明细行:每条含物品、数量、医生与时间;同一物品可多次出现。",
|
|
|
|
|
|
)
|
|
|
|
|
|
summary: list[SurgeryConsumptionSummary] = Field(
|
|
|
|
|
|
default_factory=list,
|
|
|
|
|
|
description="按物品汇总的消耗合计,应与 details 按 item_id 汇总一致。",
|
|
|
|
|
|
)
|