xinyin025/LangBot
1
0
Fork 0
mirror of https://github.com/langbot-app/LangBot.git synced 2026-10-22 08:25:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #2692 from langbot-app/test/quality-valkey-search-oracles-20261017

Browse Source 
test(vector): verify real Valkey search and deletion results
This commit is contained in:
huanghuoguoguo
2026-10-17 20:12:15 +08:00
committed by GitHub
parent 64cdb7095d 4a2780b841
commit c9f886a203
1 changed files with 18 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/integration/vector/test_valkey_search.py
+18 -5
View File
@@ -169,6 +169,7 @@ async def test_full_text_search(backend):
async def test_hybrid_filter_then_knn(backend):
db, collection = backend
ids, embeddings, metadatas, documents = _sample_docs()
documents = ['cats nearby', 'cats elsewhere', 'cats farther']
await db.add_embeddings(collection, ids, embeddings, metadatas, documents)
result = await _poll_until(
@@ -180,10 +181,10 @@ async def test_hybrid_filter_then_knn(backend):
query_text='cats',
filter={'file_id': 'fileA'},
),
lambda r: len(r['ids'][0]) >= 1,
lambda r: len(r['ids'][0]) == 2,
)
# Only fileA docs (d1, d3) should be candidates.
assert set(result['ids'][0]).issubset({'d1', 'd3'})
assert result['ids'][0] == ['d1', 'd3']
assert [item['file_id'] for item in result['metadatas'][0]] == ['fileA', 'fileA']
@pytest.mark.asyncio
@@ -191,16 +192,17 @@ async def test_vector_weight_not_honored(backend):
"""Passing different vector_weight values must NOT change ranking."""
db, collection = backend
ids, embeddings, metadatas, documents = _sample_docs()
documents = ['cats nearby', 'cats elsewhere', 'cats farther']
await db.add_embeddings(collection, ids, embeddings, metadatas, documents)
common = dict(
collection=collection, query_embedding=[1.0, 0.0, 0.0, 0.0], k=3, search_type='hybrid', query_text='cats'
)
await _poll_until(lambda: db.search(**common), lambda r: len(r['ids'][0]) >= 1)
await _poll_until(lambda: db.search(**common), lambda r: len(r['ids'][0]) == 3)
r_low = await db.search(**common, vector_weight=0.1)
r_high = await db.search(**common, vector_weight=0.9)
assert r_low['ids'][0] == r_high['ids'][0]
assert r_low['ids'][0] == r_high['ids'][0] == ['d1', 'd3', 'd2']
@pytest.mark.asyncio
@@ -257,6 +259,9 @@ async def test_delete_by_filter_returns_count(backend):
deleted = await db.delete_by_filter(collection, filter={'file_id': 'fileA'})
assert deleted == 2
items, total = await _poll_until(lambda: db.list_by_filter(collection, limit=10), lambda result: result[1] == 1)
assert total == 1
assert [item['id'] for item in items] == ['d2']
@pytest.mark.asyncio
@@ -273,6 +278,10 @@ async def test_list_by_filter_pagination(backend):
page2, total = await db.list_by_filter(collection, limit=2, offset=2)
assert total == 3
assert len(page2) == 1
first_ids = {item['id'] for item in page1}
second_ids = {item['id'] for item in page2}
assert first_ids.isdisjoint(second_ids)
assert first_ids | second_ids == {'d1', 'd2', 'd3'}
@pytest.mark.asyncio
@@ -282,7 +291,11 @@ async def test_delete_collection(backend):
await db.add_embeddings(collection, ids, embeddings, metadatas, documents)
await _poll_until(lambda: db.list_by_filter(collection, limit=10), lambda r: r[1] >= 3)
client = await db._ensure_client()
keys = [f'kb:{collection}:{document_id}' for document_id in ids]
assert await client.exists(keys) == 3
await db.delete_collection(collection)
assert await client.exists(keys) == 0
# After dropping, search on a missing index returns empty.
result = await db.search(collection, [1.0, 0.0, 0.0, 0.0], k=3, search_type='vector')