kevin/life-echo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(api+app): 对话阶段化、回忆录流水线与客户端会话体验

Browse Source 
- DB: segments 用户输入文本（Alembic 0002）
- Chat: 阶段检测/阶段提示/回复限制，编排与访谈/画像 prompts 调整
- Memoir: 忠实度检查 agent，叙事与分类等链路更新
- Core: agent 日志、Alembic 启动、LangChain/日志/配置等
- Story: time_hints；Memory 检索与相关测试
- Expo: 助手头像、会话页与消息拆分、实时会话与文案/i18n
- Docs/scripts/tests: 迁移脚本、LLM JSON/记忆检索文档、新增单测
This commit is contained in:
Kevin
2026-03-26 12:13:36 +08:00
parent 49b089354c
commit a3f61fcc0f
94 changed files with 3332 additions and 672 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
api/app/features/conversation/ws/pipeline.py
View File

@@ -2,6 +2,7 @@
import asyncio
import base64
import time
import uuid
from dataclasses import dataclass, field
from datetime import datetime, timezone
@@ -16,6 +17,7 @@ from sqlalchemy import select, update
from sqlalchemy.ext.asyncio import AsyncSession
from app.agents.chat import ChatOrchestrator
from app.core.agent_logging import agent_summary_enabled
from app.core.config import settings
from app.core.db import AsyncSessionLocal
from app.core.dependencies import get_asr_provider, get_object_storage, get_tts_provider
@@ -93,11 +95,11 @@ async def _send_tts_audio(
err_str = str(e)
if "PkgExhausted" in err_str:
logger.warning(
"TTS skipped: 腾讯云语音合成资源包已用尽，请在控制台购买或开通后付费: %s",
"TTS skipped: 腾讯云语音合成资源包已用尽，请在控制台购买或开通后付费: {}",
err_str[:100],
)
else:
logger.error("TTS synthesize failed: %s", e)
logger.error("TTS synthesize failed: {}", e)
return None
@@ -438,20 +440,20 @@ async def process_audio_segment(
async with state.lock:
state.processed_indices.add(segment_index)
logger.debug(
"分段已存在，按幂等跳过: conversation_id=%s voice_session_id=%s "
"segment_index=%s segment_id=%s transcript=%s",
"分段已存在，按幂等跳过: conversation_id={} voice_session_id={} "
"segment_index={} segment_id={} transcript={}",
conversation_id,
voice_session_id,
segment_index,
existing_segment.id,
existing_segment.transcript_text or "",
existing_segment.user_input_text or "",
)
return
else:
segment = Segment(
id=str(uuid.uuid4()),
conversation_id=conversation_id,
transcript_text=transcript_text or "",
user_input_text=transcript_text or "",
audio_url=_build_segment_audio_url(voice_session_id, segment_index),
audio_duration_seconds=audio_duration
if audio_duration > 0
@@ -531,6 +533,7 @@ async def process_user_message(
is_from_voice = bool(segment.audio_url)
voice_session_id = _voice_session_id_from_audio_url(segment.audio_url)
audio_dur = getattr(segment, "audio_duration_seconds", None)
t_pipeline = time.perf_counter()
turn = await chat_orchestrator.process_user_message(
conversation_id=conversation_id,
user_message=user_message,
@@ -545,6 +548,18 @@ async def process_user_message(
user_message_timestamp=user_message_timestamp,
audio_duration_seconds=audio_dur,
)
if agent_summary_enabled():
logger.info(
"pipeline.process_user_message duration_ms={:.2f} "
"conversation_id={} segment_id={} user_msg_len={} "
"response_segments={} skip_tts={}",
(time.perf_counter() - t_pipeline) * 1000,
conversation_id,
segment.id,
len(user_message or ""),
len(turn.messages),
turn.skip_tts,
)
responses = turn.messages
skip_tts = turn.skip_tts
@@ -618,7 +633,7 @@ async def process_user_message(
)
await db.commit()
except Exception as persist_error:
logger.warning("补写 TTS 元数据失败: %s", persist_error)
logger.warning("补写 TTS 元数据失败: {}", persist_error)
logger.error(f"处理用户消息失败: {e}", exc_info=True)
if conversation_id in manager.active_connections:
try:

46
api/app/features/conversation/ws/router.py
View File

@@ -218,7 +218,7 @@ async def websocket_endpoint(
try:
if websocket.application_state != WebSocketState.CONNECTED:
logger.debug(
"WebSocket 已非连接状态，退出循环: conversation_id=%s",
"WebSocket 已非连接状态，退出循环: conversation_id={}",
conversation_id,
)
break
@@ -251,7 +251,7 @@ async def websocket_endpoint(
segment = Segment(
id=str(uuid.uuid4()),
conversation_id=conversation_id,
transcript_text=text_message,
user_input_text=text_message,
processed=False,
)
db.add(segment)
@@ -414,8 +414,8 @@ async def websocket_endpoint(
if not should_process:
logger.debug(
"收到重复分段，跳过: conversation_id=%s voice_session_id=%s "
"segment_index=%s audio_b64_len=%s duration=%s",
"收到重复分段，跳过: conversation_id={} voice_session_id={} "
"segment_index={} audio_b64_len={} duration={}",
conversation_id,
voice_session_id,
segment_index,
@@ -470,7 +470,7 @@ async def websocket_endpoint(
continue
logger.debug(
"收到音频消息: conversation_id=%s duration_s=%s",
"收到音频消息: conversation_id={} duration_s={}",
conversation_id,
audio_duration,
)
@@ -478,18 +478,16 @@ async def websocket_endpoint(
try:
asr = get_asr_provider()
audio_bytes = base64.b64decode(audio_base64)
transcript_text = await asr.transcribe(
audio_bytes, "m4a"
asr_text = await asr.transcribe(audio_bytes, "m4a")
logger.debug(
"ASR 转写完成: conversation_id={} chars={}",
conversation_id,
len(asr_text or ""),
)
logger.debug(
"ASR 转写完成: conversation_id=%s chars=%s",
"ASR 转写全文: conversation_id={} text={}",
conversation_id,
len(transcript_text or ""),
)
logger.debug(
"ASR 转写全文: conversation_id=%s text=%s",
conversation_id,
transcript_text,
asr_text,
)
await manager.send_message(
@@ -498,7 +496,7 @@ async def websocket_endpoint(
"type": MessageType.TRANSCRIPT,
"conversation_id": conversation_id,
"data": {
"text": transcript_text,
"text": asr_text,
"audio_duration": audio_duration,
},
"timestamp": datetime.now(
@@ -514,7 +512,7 @@ async def websocket_endpoint(
segment = Segment(
id=str(uuid.uuid4()),
conversation_id=conversation_id,
transcript_text=transcript_text,
user_input_text=asr_text,
audio_url=f"audio:{audio_duration}s",
audio_duration_seconds=ads if ads > 0 else None,
processed=False,
@@ -529,12 +527,10 @@ async def websocket_endpoint(
conversation.user_id, segment.id
)
if transcript_text and not transcript_text.startswith(
"转写失败"
):
if asr_text and not asr_text.startswith("转写失败"):
await process_user_message(
conversation_id=conversation_id,
user_message=transcript_text,
user_message=asr_text,
conversation=conversation,
segment=segment,
db=db,
@@ -587,13 +583,13 @@ async def websocket_endpoint(
try:
asr = get_asr_provider()
audio_bytes = base64.b64decode(audio_base64)
transcript_text = await asr.transcribe(audio_bytes, "m4a")
asr_text = await asr.transcribe(audio_bytes, "m4a")
await manager.send_message(
conversation_id,
{
"type": MessageType.TRANSCRIPT,
"conversation_id": conversation_id,
"data": {"text": transcript_text or ""},
"data": {"text": asr_text or ""},
"timestamp": datetime.now(timezone.utc).isoformat(),
},
)
@@ -637,7 +633,7 @@ async def websocket_endpoint(
and "not connected" in error_msg.lower()
):
logger.debug(
"WebSocket 连接已断开或未就绪: conversation_id=%s error=%s",
"WebSocket 连接已断开或未就绪: conversation_id={} error={}",
conversation_id,
error_msg,
)
@@ -661,7 +657,7 @@ async def websocket_endpoint(
break
except WebSocketDisconnect:
logger.debug(
"WebSocket 断开连接: conversation_id=%s", conversation_id
"WebSocket 断开连接: conversation_id={}", conversation_id
)
break
except Exception as e:
@@ -681,7 +677,7 @@ async def websocket_endpoint(
break
except WebSocketDisconnect:
logger.debug("WebSocket 断开连接: conversation_id=%s", conversation_id)
logger.debug("WebSocket 断开连接: conversation_id={}", conversation_id)
await manager.disconnect(conversation_id)
cleanup_segment_states(conversation_id)
except Exception as e: