07c6478742
- 新增 utterance_substance：短时/应答/元话语可跳过记忆检索、阶段 LLM 与资料抽取 LLM；可配置 - 输入归一化：LLM 模式默认仅语音/ASR；配置项写入 .env.example - Memoir Phase1：可选 batch LLM 一次性抽取+分类（失败回退逐段）；Extraction 空槽位时阶段与 current_stage 对齐，prompt 约束收紧 - 叙事与忠实度：narrative_safety、证据重叠/场合锚点、标题 slots 与履历短语 grounded；fidelity 解析失败 fail-open 可配置 - 章节管线：锁 TTL 上调、锁竞争 Celery 重试、Phase2 immediate singleflight 等；story_pipeline_sync / chapter_compose / memoir_tasks 联动 - Memory：compaction / repo / summarizer / evidence 小修；事实 FTS 未命中是否回退最近事实可配置 - 新增 memoir_pipeline_trace；补充 memoir_reliability 文档与多项回归/门控测试
121 lines
4.6 KiB
Python
121 lines
4.6 KiB
Python
from pgvector.sqlalchemy import Vector
|
|
from sqlalchemy import (
|
|
JSON,
|
|
Boolean,
|
|
Column,
|
|
DateTime,
|
|
Float,
|
|
ForeignKey,
|
|
Integer,
|
|
String,
|
|
Text,
|
|
)
|
|
from sqlalchemy.dialects.postgresql import TSVECTOR as TSVector
|
|
from sqlalchemy.orm import relationship
|
|
|
|
from app.core.db import Base, utc_now
|
|
from app.core.embedding import MEMORY_EMBEDDING_DIMENSION
|
|
|
|
pgvector_type = Vector(MEMORY_EMBEDDING_DIMENSION)
|
|
|
|
|
|
class MemorySource(Base):
|
|
__tablename__ = "memory_sources"
|
|
id = Column(String, primary_key=True)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
source_type = Column(String, nullable=False) # transcript / note / draft
|
|
raw_text = Column(Text, nullable=True)
|
|
storage_key = Column(String, nullable=True)
|
|
speaker = Column(String, nullable=True)
|
|
captured_at = Column(DateTime(timezone=True), nullable=True)
|
|
status = Column(String, default="active")
|
|
conversation_id = Column(String, ForeignKey("conversations.id"), nullable=True)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|
|
chunks = relationship(
|
|
"MemoryChunk", back_populates="source", cascade="all, delete-orphan"
|
|
)
|
|
|
|
|
|
class MemoryChunk(Base):
|
|
__tablename__ = "memory_chunks"
|
|
id = Column(String, primary_key=True)
|
|
source_id = Column(
|
|
String, ForeignKey("memory_sources.id"), nullable=False, index=True
|
|
)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
content = Column(Text, nullable=False)
|
|
# pgvector embedding — Alembic migration 负责 CREATE EXTENSION vector 及列类型
|
|
embedding = Column(pgvector_type, nullable=True)
|
|
# PostgreSQL FTS — Alembic migration 负责 generated tsvector 列 + GIN index
|
|
content_tsv = Column(TSVector, nullable=True)
|
|
chunk_index = Column(Integer, nullable=False)
|
|
speaker = Column(String, nullable=True)
|
|
event_year = Column(Integer, nullable=True)
|
|
metadata_json = Column(JSON, nullable=True)
|
|
is_excluded = Column(Boolean, default=False)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|
|
source = relationship("MemorySource", back_populates="chunks")
|
|
|
|
|
|
class MemorySummary(Base):
|
|
__tablename__ = "memory_summaries"
|
|
id = Column(String, primary_key=True)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
summary_type = Column(String, nullable=False) # session / rolling / topic
|
|
content = Column(Text, nullable=False)
|
|
source_chunk_ids = Column(JSON, nullable=True)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|
|
updated_at = Column(DateTime(timezone=True), default=utc_now, onupdate=utc_now)
|
|
|
|
|
|
class MemoryFact(Base):
|
|
__tablename__ = "memory_facts"
|
|
id = Column(String, primary_key=True)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
fact_type = Column(
|
|
String, nullable=False
|
|
) # person / event / relation / place / milestone
|
|
subject = Column(String, nullable=True)
|
|
predicate = Column(String, nullable=True)
|
|
object_json = Column(JSON, nullable=True)
|
|
confidence = Column(Float, default=0.0)
|
|
source_chunk_id = Column(String, ForeignKey("memory_chunks.id"), nullable=True)
|
|
status = Column(
|
|
String, default="candidate"
|
|
) # candidate / confirmed / rejected / stale (chunk excluded / superseded)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|
|
|
|
|
|
class TimelineEvent(Base):
|
|
__tablename__ = "timeline_events"
|
|
id = Column(String, primary_key=True)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
memory_source_id = Column(
|
|
String,
|
|
ForeignKey("memory_sources.id", ondelete="SET NULL"),
|
|
nullable=True,
|
|
index=True,
|
|
)
|
|
event_year = Column(Integer, nullable=True)
|
|
event_date = Column(String, nullable=True)
|
|
title = Column(String, nullable=False)
|
|
description = Column(Text, nullable=True)
|
|
person_refs = Column(JSON, nullable=True)
|
|
source_fact_ids = Column(JSON, nullable=True)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|
|
|
|
|
|
class MemoryCurationAction(Base):
|
|
__tablename__ = "memory_curation_actions"
|
|
id = Column(String, primary_key=True)
|
|
user_id = Column(String, ForeignKey("users.id"), nullable=False, index=True)
|
|
action_type = Column(
|
|
String, nullable=False
|
|
) # exclude / restore / correct / merge / confirm / reject
|
|
target_type = Column(
|
|
String, nullable=False
|
|
) # chunk / fact / summary / timeline_event
|
|
target_id = Column(String, nullable=False)
|
|
details = Column(JSON, nullable=True)
|
|
created_at = Column(DateTime(timezone=True), default=utc_now)
|