Buckets:

download
raw
3.48 kB
import os
os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
import asyncio
from pathlib import Path
import pytest
from unittest.mock import patch
from rag_core.search import Search
from rag_core.service import RAGService
class FakeEmbeddings:
def __init__(self, results):
self._results = results
def search(self, query, limit):
# ignore query, limit, just return preset
return self._results
@pytest.mark.asyncio
async def test_search_extension_and_model_weights(tmp_path):
"""Search._process_and_rank_results applies extension + model weights correctly."""
# Prevent actual embedding loading
with patch.object(Search, "_load_embeddings", return_value=None):
# Config extension weights
ext_cfg = {"extensions": {".py": 0.9, ".md": 0.98}}
model_weights = {"file1.py": 2.0, "docs/readme.md": 0.5}
s = Search(
tmp_path / "emb",
dual=False,
extension_weights=ext_cfg,
model_weights=model_weights,
)
# Inject fake embeddings returning base scores
s.general_embeddings = FakeEmbeddings(
[
{"id": "file1.py", "text": "code file", "score": 1.0},
{"id": "docs/readme.md", "text": "readme file", "score": 1.0},
]
)
results = s.search("query", limit=2)
# Expect order: file1.py first due to higher adjusted score
assert results[0]["id"] == "file1.py"
# Extract adjusted scores
adj = {r["id"]: r["adjusted_score"] for r in results}
# file1: 0.9 * 2.0 * 1.0 = 1.8
# readme: 0.98 * 0.5 * 1.0 = 0.49
assert pytest.approx(adj["file1.py"], rel=1e-6) == 1.8
assert pytest.approx(adj["docs/readme.md"], rel=1e-6) == 0.49
@pytest.mark.asyncio
async def test_runtime_weight_updates_results(tmp_path):
"""RAGService.set_weight immediately influences adjusted_score in subsequent searches."""
embeddings_path = tmp_path / "emb"
embeddings_path.mkdir()
weights_path = tmp_path / "weights.yaml"
weights_path.write_text("extensions: {}")
config_path = tmp_path / "repositories.yml"
config_path.write_text("{}")
# Patch embedding load in both Search and Registry usage
with patch.object(Search, "_load_embeddings", return_value=None):
rag = RAGService(
embeddings_path=embeddings_path,
config_path=config_path,
weights_path=weights_path,
use_dual_embedding=False,
)
# Provide fake embeddings
rag.search.general_embeddings = FakeEmbeddings([{"id": "foo.py", "text": "foo code", "score": 1.0}])
# First search (no weight)
first = await rag.search_docs("x", limit=1)
assert first[0]["id"] == "foo.py"
base_adjusted = first[0]["adjusted_score"]
assert pytest.approx(base_adjusted, rel=1e-6) == 1.0 # default adjusted = 1
# Set weight
await rag.set_weight("foo.py", 2.0)
# Second search
second = await rag.search_docs("x", limit=1)
assert second[0]["id"] == "foo.py"
new_adjusted = second[0]["adjusted_score"]
# Expect doubling (extension weight default 1.0)
assert pytest.approx(new_adjusted, rel=1e-6) == 2.0
# Model score field should reflect updated multiplier (clamped logic inside search not applied to adjusted_score beyond 0.5-2.0 range)
assert second[0]["model_score"] == 2.0

Xet Storage Details

Size:
3.48 kB
·
Xet hash:
ae97bd664179d83c34df5b2144086710c0d47df3e337af10f7788e4c113c951d

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.