e4bf0710c7
数据库 - 新增迁移 0003：timeline_events.memory_source_id 外键 → memory_sources，便于按 ingest 源做时间线幂等 后端 - 记忆 - 新增 ingest 后 LLM 富化（摘要/事实/时间线），可配置开关与最大字符数 - 新增证据包组装：合并 chunk、摘要、事实、时间线、故事等检索结果；支持空 query 时是否仍带 rolling 等开关 - repo/retriever/service/router/schemas/summarizer/timeline/extractor 等扩展；文档 memory-retrieval.md 更新 后端 - 对话 WS - 增加 PING/PONG；分段 ASR 日志与空音频处理；转写失败与「无助手回复」错误提示更明确 - 助手多段回复持久化使用统一分隔符，与分段逻辑一致 后端 - Agent - reply_limits：按 [SPLIT] 与段落拆段，并保证非空 fallback，供 WS 与 TTS 多段下发 后端 - 回忆录任务 - transcript ingest 记录 source_id；任务成功结?
132 lines
4.7 KiB
Python
132 lines
4.7 KiB
Python
"""会话摘要与滚动摘要（LLM + JSON）。"""
|
||
|
||
from __future__ import annotations
|
||
|
||
from typing import Any
|
||
|
||
from app.core.langchain_llm import ainvoke_json_object, invoke_json_object
|
||
from app.core.logging import get_logger
|
||
from app.features.memory.llm_schemas import (
|
||
RollingSummaryPayload,
|
||
SessionSummaryPayload,
|
||
parse_json_payload,
|
||
)
|
||
|
||
logger = get_logger(__name__)
|
||
|
||
|
||
def _max_input_chars() -> int:
|
||
from app.core.config import settings
|
||
|
||
return settings.memory_enrichment_max_chars
|
||
|
||
|
||
def generate_session_summary_sync(llm: Any, chunk_texts: list[str]) -> str:
|
||
"""为本批块生成 session 级短摘要。"""
|
||
if not llm:
|
||
return ""
|
||
lim = _max_input_chars()
|
||
combined = "\n\n".join(t for t in chunk_texts if t).strip()[:lim]
|
||
if not combined:
|
||
return ""
|
||
prompt = (
|
||
"用 2～8 句中文概括下列口述/对话要点，不编造、不评价。只输出 JSON："
|
||
'{"summary":"..."}\n\n文本：\n'
|
||
f"{combined}"
|
||
)
|
||
try:
|
||
raw = invoke_json_object(
|
||
llm, prompt, max_tokens=2048, agent="memory.session_summary_sync"
|
||
)
|
||
parsed = parse_json_payload(raw, SessionSummaryPayload)
|
||
if parsed is None:
|
||
return ""
|
||
return str(parsed.summary or "").strip()
|
||
except (TypeError, ValueError) as e:
|
||
logger.warning("generate_session_summary_sync 失败: {}", e)
|
||
return ""
|
||
|
||
|
||
async def generate_session_summary_async(llm: Any, chunk_texts: list[str]) -> str:
|
||
if not llm:
|
||
return ""
|
||
lim = _max_input_chars()
|
||
combined = "\n\n".join(t for t in chunk_texts if t).strip()[:lim]
|
||
if not combined:
|
||
return ""
|
||
prompt = (
|
||
"用 2～8 句中文概括下列口述/对话要点，不编造、不评价。只输出 JSON："
|
||
'{"summary":"..."}\n\n文本：\n'
|
||
f"{combined}"
|
||
)
|
||
try:
|
||
raw = await ainvoke_json_object(
|
||
llm, prompt, max_tokens=2048, agent="memory.session_summary_async"
|
||
)
|
||
parsed = parse_json_payload(raw, SessionSummaryPayload)
|
||
if parsed is None:
|
||
return ""
|
||
return str(parsed.summary or "").strip()
|
||
except (TypeError, ValueError) as e:
|
||
logger.warning("generate_session_summary_async 失败: {}", e)
|
||
return ""
|
||
|
||
|
||
def generate_rolling_summary_sync(
|
||
llm: Any, existing_summary: str | None, new_chunk_texts: list[str]
|
||
) -> str:
|
||
"""合并已有滚动摘要与新材料。"""
|
||
if not llm:
|
||
return (existing_summary or "").strip()
|
||
lim = _max_input_chars()
|
||
new_t = "\n\n".join(t for t in new_chunk_texts if t).strip()[:lim]
|
||
if not new_t and not (existing_summary or "").strip():
|
||
return ""
|
||
ex = (existing_summary or "").strip()[:lim]
|
||
prompt = (
|
||
"将「已有滚动摘要」与「新材料」合并为更新后的滚动摘要（中文，段落）。"
|
||
"保留人物与时间线索；不编造；可省略无关细节。\n"
|
||
'只输出 JSON：{"rolling_summary":"..."}\n\n'
|
||
f"【已有摘要】\n{ex}\n\n【新材料】\n{new_t}"
|
||
)
|
||
try:
|
||
raw = invoke_json_object(
|
||
llm, prompt, max_tokens=3072, agent="memory.rolling_summary_sync"
|
||
)
|
||
parsed = parse_json_payload(raw, RollingSummaryPayload)
|
||
if parsed is None:
|
||
return (existing_summary or "").strip()
|
||
return str(parsed.rolling_summary or "").strip()
|
||
except (TypeError, ValueError) as e:
|
||
logger.warning("generate_rolling_summary_sync 失败: {}", e)
|
||
return (existing_summary or "").strip()
|
||
|
||
|
||
async def generate_rolling_summary_async(
|
||
llm: Any, existing_summary: str | None, new_chunk_texts: list[str]
|
||
) -> str:
|
||
if not llm:
|
||
return (existing_summary or "").strip()
|
||
lim = _max_input_chars()
|
||
new_t = "\n\n".join(t for t in new_chunk_texts if t).strip()[:lim]
|
||
if not new_t and not (existing_summary or "").strip():
|
||
return ""
|
||
ex = (existing_summary or "").strip()[:lim]
|
||
prompt = (
|
||
"将「已有滚动摘要」与「新材料」合并为更新后的滚动摘要（中文，段落）。"
|
||
"保留人物与时间线索；不编造。\n"
|
||
'只输出 JSON：{"rolling_summary":"..."}\n\n'
|
||
f"【已有摘要】\n{ex}\n\n【新材料】\n{new_t}"
|
||
)
|
||
try:
|
||
raw = await ainvoke_json_object(
|
||
llm, prompt, max_tokens=3072, agent="memory.rolling_summary_async"
|
||
)
|
||
parsed = parse_json_payload(raw, RollingSummaryPayload)
|
||
if parsed is None:
|
||
return (existing_summary or "").strip()
|
||
return str(parsed.rolling_summary or "").strip()
|
||
except (TypeError, ValueError) as e:
|
||
logger.warning("generate_rolling_summary_async 失败: {}", e)
|
||
return (existing_summary or "").strip()
|