xsl
aa1287d83d
DeterministicEmbedder（可换 BGE-M3 的接口位）+ kb_chunk 双索引写入 + 硬过滤→向量→关键词→RRF 重排管线 + kb_gap 缺口埋点（T-P2-03/04/06，TD-08）。 实现： - embedding.py：EmbeddingProvider Protocol + DeterministicEmbedder （sha256 种子→归一化 1024 维假向量，同文本同向量/异文本近似正交） + build_embedding_provider 工厂（未知模型名抛 CONFIG_ERROR）。 - repo.py：upsert_chunks（整实体先删后插，逐行带 embedding_model/version， 可重跑判据 T-P2-03）；hybrid_search（§8.3 CTE：硬过滤 published/ entity_types/occasion/avoid_body_shapes → <=> 余弦 topN → ts_rank topN → RRF）；record_gap（咨询锁串行化同 query+filter_ctx 的 hit_count 递增）。 - rerank.py：RRF/权重/新鲜度纯数学（独立于 SQL 以便单测逐项断言）。 - service.py：RetrievalService.search 编排 embed→检索→（空结果）gap 埋点， SearchOutcome.degraded 标记降级。 测试：单测 +30（test_embedding 确定性/维度/归一化/工厂拒绝、test_rerank 公式逐项断言、test_retrieval_service 编排打桩）；集成 +5 （test_hybrid_retrieval：禁忌实体同文最近向量仍零出现、draft 零出现、 关键词路径、换模型整组替换、kb_gap hit_count=2、partial 索引 EXPLAIN 验证）。 全绿：ruff / mypy(retrieval+shared+domain) / lint-imports / 268 单测 / 21 集成。 偏差说明： 1) 新增 rerank.py（任务文档只点名 embedding/repo/service，但重排数学必须 不连库可单测，独立模块是最低成本实现，仍属 packages/retrieval/**）。 2) upsert_chunks 的 upsert 粒度是实体（先删后插）：kb_chunk 无业务自然键， 行级 ON CONFLICT 不可行；换模型名=新版本整组替换，不留旧版本残向量。 3) 禁忌过滤用 COALESCE(attrs->avoid_body_shapes, '[]')兜底：jsonb NULL 的 ? 运算结果是 NULL，直接 NOT 会把未声明禁忌的好实体一并排除。 4) 关键词路过滤用 tsv @@ tsquery 而非 ts_rank>0：ts_rank 对不匹配文本返回 ~1e-20 正 epsilon，阈值过滤会把无关 chunk 全放进关键词路。 5) RRF 分量按 1/(k+1) 归一到 (0,1]，使 0.45/0.2/0.25/0.1 权重量纲可比 （原始倒数 0.45×1/61 与规则项 0.25×1 完全不可比）。
100 lines
3.5 KiB
Python
100 lines
3.5 KiB
Python
"""DeterministicEmbedder 与工厂：纯数学，不连库不联网（T-P2-03 接口位）."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import math
|
|
|
|
import pytest
|
|
|
|
from packages.retrieval.embedding import (
|
|
DETERMINISTIC_MODEL_NAME,
|
|
DeterministicEmbedder,
|
|
build_embedding_provider,
|
|
)
|
|
from packages.shared.config import get_settings
|
|
from packages.shared.errors import AppError, ErrorCode
|
|
|
|
|
|
def _dot(a: list[float], b: list[float]) -> float:
|
|
return sum(x * y for x, y in zip(a, b, strict=True))
|
|
|
|
|
|
async def test_same_text_embeds_to_same_vector() -> None:
|
|
embedder = DeterministicEmbedder(dimensions=1024)
|
|
first, second = await embedder.embed(["白衬衫怎么搭配", "白衬衫怎么搭配"])
|
|
assert first == second
|
|
|
|
|
|
async def test_dimensions_match_declaration() -> None:
|
|
vec = (await DeterministicEmbedder(dimensions=1024).embed(["hello"]))[0]
|
|
assert len(vec) == 1024
|
|
small = (await DeterministicEmbedder(dimensions=8).embed(["hello"]))[0]
|
|
assert len(small) == 8
|
|
|
|
|
|
async def test_vectors_are_l2_normalized() -> None:
|
|
for text in ("hello", "中文知识片段", ""):
|
|
vec = (await DeterministicEmbedder(dimensions=1024).embed([text]))[0]
|
|
assert math.isclose(_dot(vec, vec), 1.0, rel_tol=1e-6, abs_tol=1e-9)
|
|
|
|
|
|
async def test_different_texts_are_near_orthogonal() -> None:
|
|
"""不同文本近似正交（|dot| ≪ 1）：向量召回管线在无真模型时仍可测."""
|
|
|
|
left, right = await DeterministicEmbedder(dimensions=1024).embed(
|
|
["interview outfit advice", "winter skincare routine"]
|
|
)
|
|
assert left != right
|
|
assert abs(_dot(left, right)) < 0.1
|
|
|
|
|
|
async def test_batch_preserves_order_and_length() -> None:
|
|
texts = ["alpha", "beta", "gamma"]
|
|
vectors = await DeterministicEmbedder(dimensions=16).embed(texts)
|
|
assert len(vectors) == len(texts)
|
|
singles = [(await DeterministicEmbedder(dimensions=16).embed([t]))[0] for t in texts]
|
|
assert vectors == singles
|
|
|
|
|
|
def test_metadata_is_stable() -> None:
|
|
embedder = DeterministicEmbedder(dimensions=8)
|
|
assert embedder.model_name == DETERMINISTIC_MODEL_NAME
|
|
assert embedder.version
|
|
assert embedder.dimensions == 8
|
|
|
|
|
|
def test_invalid_dimensions_rejected() -> None:
|
|
with pytest.raises(ValueError, match="dimensions"):
|
|
DeterministicEmbedder(dimensions=0)
|
|
|
|
|
|
def test_factory_builds_deterministic_embedder(monkeypatch: pytest.MonkeyPatch) -> None:
|
|
# 带 alias 的 Settings 字段按字段名构造会被静默忽略 —— 统一走
|
|
# setenv + cache_clear（上批经验，见任务文档 §2）。
|
|
monkeypatch.setenv("KB_EMBEDDING_MODEL", "deterministic-v1")
|
|
monkeypatch.setenv("KB_VECTOR_DIMENSIONS", "1024")
|
|
get_settings.cache_clear()
|
|
try:
|
|
settings = get_settings()
|
|
provider = build_embedding_provider(settings)
|
|
assert isinstance(provider, DeterministicEmbedder)
|
|
assert provider.model_name == "deterministic-v1"
|
|
assert provider.dimensions == 1024
|
|
finally:
|
|
monkeypatch.undo()
|
|
get_settings.cache_clear()
|
|
|
|
|
|
def test_factory_rejects_unknown_model(monkeypatch: pytest.MonkeyPatch) -> None:
|
|
monkeypatch.setenv("KB_EMBEDDING_MODEL", "bge-m3-onnx")
|
|
get_settings.cache_clear()
|
|
try:
|
|
settings = get_settings()
|
|
assert settings.kb_embedding_model == "bge-m3-onnx"
|
|
with pytest.raises(AppError) as excinfo:
|
|
build_embedding_provider(settings)
|
|
assert excinfo.value.code == ErrorCode.CONFIG_ERROR
|
|
finally:
|
|
monkeypatch.undo()
|
|
get_settings.cache_clear()
|