69a673e6c6
Chat 访谈 - 新增 persona 系统（default / warm_listener / curious_guide）与 background_voice 语气层 - 回复长度由 compute_reply_plan 统一决策（brief / standard / expanded），融合信息密度启发式 - 输入净稿（input_normalize）：编排层可选 rules/llm 归一用户口语后再喂模型与记忆检索 - 记忆证据注入：按用户话检索 memory evidence 并注入 prompt Memoir 回忆录 - 口述归一（oral_normalize）：segment 原文保留，story 管线取派生净稿作叙事输入 - segment 入队批次门闸：累计字数 + 最长等待秒数，减少零碎提交 - fidelity_check / prompts / narrative_agent 微调 - Alembic 0005：清理跨章节 story 外键 Infra - Dockerfile 加入 ffmpeg - pyproject.toml 新增依赖并同步 uv.lock - .env.example / .env.production 补全新配置项 Tests - 新增 test_background_voice、test_chat_input_normalize、test_experience_regressions - 扩展 test_interview_prompts、test_interview_reply_length、test_story_route_oral_invariant Made-with: Cursor
291 lines
11 KiB
Python
291 lines
11 KiB
Python
"""Conversation service — 对话编排（列表、创建、结束、删除、消息、整理）。"""
|
||
|
||
import uuid
|
||
from datetime import datetime, timezone
|
||
|
||
from fastapi import HTTPException
|
||
from sqlalchemy.ext.asyncio import AsyncSession
|
||
|
||
from app.core.cos_url_keys import (
|
||
collect_cos_keys_from_conversation_history,
|
||
collect_cos_keys_from_tts_url_list,
|
||
extract_cos_object_key_if_owned,
|
||
)
|
||
from app.core.logging import get_logger
|
||
from app.core.redis import redis_service
|
||
from app.core.storage_purge import delete_object_storage_keys_best_effort
|
||
from app.features.conversation import repo
|
||
from app.features.conversation.models import Conversation
|
||
from app.features.conversation.session_history import (
|
||
conversation_messages_to_redis_history,
|
||
)
|
||
from app.features.conversation.tts_delivery import apply_presigned_tts_urls_to_messages
|
||
from app.features.memory import repo as memory_repo
|
||
from app.features.quota.service import QuotaService
|
||
from app.ports.storage import ObjectStorage
|
||
from app.tasks.memoir_tasks import process_memoir_segments
|
||
|
||
logger = get_logger(__name__)
|
||
|
||
|
||
def _datetime_to_timestamp_ms(value: datetime | None) -> int:
|
||
if value is None:
|
||
return int(datetime.now(timezone.utc).timestamp() * 1000)
|
||
if value.tzinfo is None:
|
||
value = value.replace(tzinfo=timezone.utc)
|
||
return int(value.timestamp() * 1000)
|
||
|
||
|
||
def _message_timestamp_ms(msg: dict, fallback: datetime | None) -> int:
|
||
raw_timestamp = msg.get("timestamp")
|
||
if isinstance(raw_timestamp, (int, float)):
|
||
return int(raw_timestamp)
|
||
if isinstance(raw_timestamp, str):
|
||
try:
|
||
return int(
|
||
datetime.fromisoformat(raw_timestamp.replace("Z", "+00:00")).timestamp()
|
||
* 1000
|
||
)
|
||
except ValueError:
|
||
pass
|
||
return _datetime_to_timestamp_ms(fallback)
|
||
|
||
|
||
def _latest_message_time_ms(conversation: Conversation, history: list[dict]) -> int:
|
||
last_at = getattr(conversation, "last_message_at", None)
|
||
if last_at:
|
||
return _datetime_to_timestamp_ms(last_at)
|
||
if history:
|
||
return _message_timestamp_ms(history[-1], conversation.started_at)
|
||
return _datetime_to_timestamp_ms(conversation.started_at)
|
||
|
||
|
||
def _build_messages_from_history(
|
||
conversation_id: str,
|
||
history: list[dict],
|
||
fallback_timestamp: datetime | None,
|
||
) -> list[dict]:
|
||
messages: list[dict] = []
|
||
seen_audio_sessions: set[str] = set()
|
||
for idx, msg in enumerate(history):
|
||
role = msg.get("role")
|
||
message_type = msg.get("messageType", "text")
|
||
voice_session_id = msg.get("voiceSessionId")
|
||
if role == "human" and message_type == "audio" and voice_session_id:
|
||
if voice_session_id in seen_audio_sessions:
|
||
continue
|
||
seen_audio_sessions.add(voice_session_id)
|
||
item: dict = {
|
||
"id": f"{conversation_id}_msg_{idx}",
|
||
"conversationId": conversation_id,
|
||
"content": msg.get("content", ""),
|
||
"senderType": "user" if role == "human" else "assistant",
|
||
"timestamp": _message_timestamp_ms(msg, fallback_timestamp),
|
||
"messageType": message_type,
|
||
}
|
||
if voice_session_id and role == "human":
|
||
item["voiceSessionId"] = voice_session_id
|
||
ds = msg.get("durationSeconds")
|
||
if isinstance(ds, (int, float)) and ds > 0:
|
||
item["durationSeconds"] = int(ds)
|
||
if role == "ai":
|
||
tts = msg.get("ttsAudioUrls")
|
||
if isinstance(tts, list) and tts:
|
||
item["ttsAudioUrls"] = [x for x in tts if isinstance(x, str)]
|
||
messages.append(item)
|
||
return messages
|
||
|
||
|
||
class ConversationService:
|
||
def __init__(
|
||
self,
|
||
db: AsyncSession,
|
||
quota_service: QuotaService,
|
||
*,
|
||
object_storage: ObjectStorage | None = None,
|
||
):
|
||
self._db = db
|
||
self._quota = quota_service
|
||
self._object_storage = object_storage
|
||
|
||
async def _clear_history(self, conversation_id: str) -> None:
|
||
try:
|
||
await redis_service.clear_conversation_history(conversation_id)
|
||
except Exception:
|
||
pass
|
||
|
||
async def ensure_redis_history_from_db(self, conversation_id: str) -> list[dict]:
|
||
"""
|
||
供 WS 与 get_messages 使用：优先 Redis；若为空则用 DB conversation_messages 重建并写回。
|
||
"""
|
||
try:
|
||
history = await redis_service.get_conversation_history(conversation_id)
|
||
except Exception as exc:
|
||
logger.warning("conversation history cache read skipped: {}", exc)
|
||
history = []
|
||
if history:
|
||
return history
|
||
|
||
rows = await repo.get_conversation_messages(conversation_id, self._db)
|
||
if rows:
|
||
rebuilt = conversation_messages_to_redis_history(rows)
|
||
try:
|
||
await redis_service.set_conversation_history(conversation_id, rebuilt)
|
||
except Exception as exc:
|
||
logger.warning("conversation history cache write skipped: {}", exc)
|
||
return rebuilt
|
||
|
||
return []
|
||
|
||
async def list_for_user(self, user_id: str) -> list[dict]:
|
||
conversations = await repo.get_user_conversations(user_id, self._db)
|
||
result = []
|
||
for conv in conversations:
|
||
history: list[dict] = []
|
||
try:
|
||
history = await self.ensure_redis_history_from_db(conv.id)
|
||
except Exception:
|
||
pass
|
||
latest_message = history[-1].get("content", "")[:50] if history else None
|
||
has_user_message = any((msg.get("role") == "human") for msg in history)
|
||
result.append(
|
||
{
|
||
"id": conv.id,
|
||
"title": (conv.summary or "")[:30] or "岁月知己",
|
||
"avatarUrl": None,
|
||
"latestMessagePreview": latest_message or conv.summary,
|
||
"latestMessageTime": _latest_message_time_ms(conv, history),
|
||
# 对话「初次创建」时间（ms），供客户端按日历日区分「打个招呼 / 继续对话」
|
||
"startedAt": _datetime_to_timestamp_ms(conv.started_at),
|
||
"unreadCount": 0,
|
||
"isDefaultAssistant": conv.summary is None,
|
||
"hasUserMessage": has_user_message,
|
||
}
|
||
)
|
||
return result
|
||
|
||
async def create(self, user_id: str) -> dict:
|
||
conv = Conversation(
|
||
id=str(uuid.uuid4()),
|
||
user_id=user_id,
|
||
started_at=datetime.now(timezone.utc),
|
||
status="active",
|
||
)
|
||
repo.add_conversation(conv, self._db)
|
||
await self._db.commit()
|
||
await self._db.refresh(conv)
|
||
return {
|
||
"id": conv.id,
|
||
"user_id": conv.user_id,
|
||
"started_at": conv.started_at.isoformat(),
|
||
"status": conv.status,
|
||
}
|
||
|
||
async def get_or_404(self, conversation_id: str, user_id: str) -> Conversation:
|
||
conv = await repo.get_conversation(conversation_id, self._db)
|
||
if not conv or conv.user_id != user_id or conv.deleted_at is not None:
|
||
raise HTTPException(status_code=404, detail="Conversation not found")
|
||
return conv
|
||
|
||
async def get_one(self, conversation_id: str, user_id: str) -> dict:
|
||
conv = await self.get_or_404(conversation_id, user_id)
|
||
return {
|
||
"id": conv.id,
|
||
"user_id": conv.user_id,
|
||
"started_at": conv.started_at.isoformat(),
|
||
"ended_at": conv.ended_at.isoformat() if conv.ended_at else None,
|
||
"duration_seconds": conv.duration_seconds,
|
||
"summary": conv.summary,
|
||
"status": conv.status,
|
||
"current_topic": conv.current_topic,
|
||
"conversation_stage": conv.conversation_stage,
|
||
}
|
||
|
||
async def end(self, conversation_id: str, user_id: str) -> dict:
|
||
conv = await self.get_or_404(conversation_id, user_id)
|
||
conv.status = "ended"
|
||
conv.ended_at = datetime.now(timezone.utc)
|
||
if conv.started_at:
|
||
conv.duration_seconds = int(
|
||
(conv.ended_at - conv.started_at).total_seconds()
|
||
)
|
||
await self._db.commit()
|
||
return {
|
||
"id": conv.id,
|
||
"status": conv.status,
|
||
"ended_at": conv.ended_at.isoformat(),
|
||
"duration_seconds": conv.duration_seconds,
|
||
}
|
||
|
||
async def delete(self, conversation_id: str, user_id: str) -> None:
|
||
conv = await self.get_or_404(conversation_id, user_id)
|
||
cos_keys: set[str] = set(
|
||
await memory_repo.list_storage_keys_for_conversation(
|
||
self._db, conversation_id
|
||
)
|
||
)
|
||
try:
|
||
hist = await redis_service.get_conversation_history(conversation_id)
|
||
cos_keys |= collect_cos_keys_from_conversation_history(hist)
|
||
except Exception:
|
||
pass
|
||
segments = await repo.get_segments_for_conversation(conversation_id, self._db)
|
||
for seg in segments:
|
||
k = extract_cos_object_key_if_owned(seg.audio_url)
|
||
if k:
|
||
cos_keys.add(k)
|
||
raw_tts = getattr(seg, "tts_audio_urls", None)
|
||
if isinstance(raw_tts, list):
|
||
cos_keys |= collect_cos_keys_from_tts_url_list(
|
||
[str(x) for x in raw_tts if isinstance(x, str)]
|
||
)
|
||
|
||
await self._clear_history(conversation_id)
|
||
conv.deleted_at = datetime.now(timezone.utc)
|
||
await self._db.commit()
|
||
|
||
delete_object_storage_keys_best_effort(
|
||
self._object_storage,
|
||
sorted(cos_keys),
|
||
log_prefix=f"conversation_soft_delete id={conversation_id}",
|
||
)
|
||
|
||
async def get_messages(self, conversation_id: str, user_id: str) -> list[dict]:
|
||
conv = await self.get_or_404(conversation_id, user_id)
|
||
try:
|
||
history = await self.ensure_redis_history_from_db(conversation_id)
|
||
messages = _build_messages_from_history(
|
||
conversation_id=conversation_id,
|
||
history=history,
|
||
fallback_timestamp=conv.started_at,
|
||
)
|
||
apply_presigned_tts_urls_to_messages(messages, self._object_storage)
|
||
return messages
|
||
except Exception:
|
||
return []
|
||
|
||
async def organize(
|
||
self, conversation_id: str, user_id: str, subscription_type: str
|
||
) -> dict:
|
||
conv = await self.get_or_404(conversation_id, user_id)
|
||
segments = await repo.get_segments_for_organize(conversation_id, self._db)
|
||
if not segments:
|
||
raise HTTPException(status_code=400, detail="该对话没有可整理的内容")
|
||
can_submit, quota_message = await self._quota.check_can_submit_organize(
|
||
user_id, subscription_type
|
||
)
|
||
if not can_submit:
|
||
raise HTTPException(status_code=403, detail=quota_message)
|
||
segment_ids = [s.id for s in segments]
|
||
process_memoir_segments.delay(conv.user_id, segment_ids)
|
||
logger.info(
|
||
"手动触发对话整理: conversation_id={}, segments={}",
|
||
conversation_id,
|
||
len(segment_ids),
|
||
)
|
||
return {
|
||
"message": "对话整理任务已提交",
|
||
"conversation_id": conversation_id,
|
||
"segments_count": len(segment_ids),
|
||
}
|