CI / backend (push) Canceled after 0s
之前知识只能"存",不能"找"。这一版把链路补上:切片 → 向量 → 打分 → 带来源返回。
- chunk 表 + 迁移 0006:正文按标题归段切、代码块整块保留、超长按句切、
相邻片重叠 80 字;RLS 与 GRANT 与其它租户表一致
- 向量提供方做成协议:内置 hash(字符二元组,零依赖可复现,作为基线)
与 OpenAI 兼容端点两种,换实现不动业务代码
- 检索是混合打分:向量 0.6 + pg_trgm 0.4。专有编号、单号这类必须原样命中的词
靠词法兜,换个说法表达的同一件事靠向量兜
- 三条纪律:只搜 published(过滤写在 SQL 里)、低于阈值宁可不返回、
每条命中都带来源标题与 item_id
- 发布时自动建索引,改完正文用 POST /knowledge/{id}/reindex 重建
- Web 知识页加检索框与命中卡片;新增 e2e/retrieval.mjs 6 项
为什么不是 pgvector:本机 Docker 起不来、fnsvr-01 没有该扩展且明确不动。
先用 real[] + SQL 余弦把链路跑通,将来换列类型与索引即可,接口不变。
后端 79 passed,e2e smoke/knowledge/richtext/retrieval 全绿。
113 lines
5.3 KiB
Python
113 lines
5.3 KiB
Python
"""知识域模型(M3):knowledge_doc。
|
||
|
||
三条设计约束(对应 PRODUCT-PLAN 的 D1/D3):
|
||
|
||
1. **知识也是 `item`**。标题、正文、标签、评论、事件流全部复用;`knowledge_doc`
|
||
只补"从哪来的、是什么类型、什么时候被人认可过"。同一套列表、搜索、详情、权限对知识依然成立。
|
||
2. **知识必须有来源**(D3)。`source_kind` 为 `task` / `ticket` 时必须给出 `source_item_id`,
|
||
只有 `manual`(自己记的笔记)允许留空。这条写在数据库 CHECK 里,不靠 API 自律——
|
||
将来接 LLM 自动归档时,没有来源的知识等于"没人能核对的断言",比没有更糟。
|
||
3. **状态机是 `draft → in_review → published → archived`**。草稿可以随时改,发布的要有
|
||
人来点头(`published_at` 记时间),归档的不删只藏(历史结论不该消失,只该退出检索)。
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import uuid
|
||
from datetime import datetime
|
||
|
||
from sqlalchemy import (
|
||
CheckConstraint,
|
||
DateTime,
|
||
Float,
|
||
ForeignKey,
|
||
Index,
|
||
String,
|
||
Text,
|
||
UniqueConstraint,
|
||
)
|
||
from sqlalchemy.dialects.postgresql import ARRAY
|
||
from sqlalchemy.dialects.postgresql import UUID as PgUUID
|
||
from sqlalchemy.orm import Mapped, mapped_column
|
||
|
||
from .models import Base, Timestamps, new_uuid
|
||
|
||
# 知识状态机(与任务/工单共用 item.status 一列,取值集合不同)
|
||
KNOWLEDGE_STATUSES: tuple[str, ...] = ("draft", "in_review", "published", "archived")
|
||
|
||
DOC_TYPES: tuple[str, ...] = ("note", "howto", "faq", "postmortem", "decision")
|
||
SOURCE_KINDS: tuple[str, ...] = ("manual", "task", "ticket")
|
||
GENERATED_BY: tuple[str, ...] = ("human", "rule")
|
||
|
||
|
||
class KnowledgeDoc(Base, Timestamps):
|
||
"""知识的附加信息,与 item 一比一。"""
|
||
|
||
__tablename__ = "knowledge_doc"
|
||
__table_args__ = (
|
||
Index("ix_knowledge_ws_status", "workspace_id", "status_hint"),
|
||
Index("ix_knowledge_ws_source", "workspace_id", "source_item_id"),
|
||
CheckConstraint(
|
||
"source_kind = 'manual' OR source_item_id IS NOT NULL",
|
||
name="ck_knowledge_source_required",
|
||
),
|
||
CheckConstraint("chunk_count >= 0", name="ck_knowledge_chunk_count"),
|
||
)
|
||
|
||
# item_id 既是主键又是外键:一条 item 最多一条 knowledge_doc,天然 1:1
|
||
item_id: Mapped[uuid.UUID] = mapped_column(
|
||
PgUUID(as_uuid=True), ForeignKey("item.id", ondelete="CASCADE"), primary_key=True
|
||
)
|
||
workspace_id: Mapped[uuid.UUID] = mapped_column(
|
||
PgUUID(as_uuid=True), ForeignKey("workspace.id", ondelete="CASCADE"), nullable=False
|
||
)
|
||
author_id: Mapped[uuid.UUID] = mapped_column(
|
||
PgUUID(as_uuid=True), ForeignKey("app_user.id", ondelete="RESTRICT"), nullable=False
|
||
)
|
||
# 知识的来源。NULL 只允许出现在 source_kind='manual' 上(见 CHECK)
|
||
source_item_id: Mapped[uuid.UUID | None] = mapped_column(
|
||
PgUUID(as_uuid=True), ForeignKey("item.id", ondelete="SET NULL")
|
||
)
|
||
source_kind: Mapped[str] = mapped_column(String(16), nullable=False, default="manual")
|
||
doc_type: Mapped[str] = mapped_column(String(16), nullable=False, default="note")
|
||
# 冗余一份状态:列表页要按知识自己的口径筛,不想每次都回 item 表
|
||
status_hint: Mapped[str] = mapped_column(String(16), nullable=False, default="draft")
|
||
summary: Mapped[str] = mapped_column(Text, nullable=False, default="")
|
||
generated_by: Mapped[str] = mapped_column(String(16), nullable=False, default="human")
|
||
published_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
|
||
archived_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
|
||
chunk_count: Mapped[int] = mapped_column(nullable=False, default=0)
|
||
|
||
|
||
class Chunk(Base):
|
||
"""知识正文的切片。
|
||
|
||
为什么切片与文档同库:检索要按工作区过滤,跨库做权限校验就得把两边数据捞到应用里,
|
||
既不安全也不快。向量本身是普通数组,换 pgvector 时只换列类型与索引。
|
||
"""
|
||
|
||
__tablename__ = "chunk"
|
||
__table_args__ = (
|
||
UniqueConstraint("item_id", "idx", name="uq_chunk_item_idx"),
|
||
Index("ix_chunk_ws_item", "workspace_id", "item_id"),
|
||
{"comment": "知识正文切片,检索的最小单位"},
|
||
)
|
||
|
||
id: Mapped[uuid.UUID] = mapped_column(PgUUID(as_uuid=True), primary_key=True, default=new_uuid)
|
||
workspace_id: Mapped[uuid.UUID] = mapped_column(PgUUID(as_uuid=True), nullable=False)
|
||
item_id: Mapped[uuid.UUID] = mapped_column(
|
||
PgUUID(as_uuid=True), ForeignKey("item.id", ondelete="CASCADE"), nullable=False
|
||
)
|
||
# 片内顺序,用于把命中片段按原顺序还原上下文
|
||
idx: Mapped[int] = mapped_column(nullable=False)
|
||
heading: Mapped[str | None] = mapped_column(String(200))
|
||
content: Mapped[str] = mapped_column(Text, nullable=False)
|
||
char_count: Mapped[int] = mapped_column(nullable=False, default=0)
|
||
# 向量。hash 提供方是 256 维,OpenAI 兼容端点通常是 1024/1536 维,
|
||
# 所以不固定维度——换模型时老切片要么重算,要么维度不匹配(检索会拒绝并提示重建)。
|
||
embedding: Mapped[list[float] | None] = mapped_column(ARRAY(Float))
|
||
embed_model: Mapped[str | None] = mapped_column(String(64))
|
||
|
||
|
||
KNOWLEDGE_TENANT_TABLES: tuple[str, ...] = ("knowledge_doc", "chunk")
|