mirror of
https://github.com/lancedb/lancedb.git
synced 2026-10-16 08:42:29 +00:00
Brendan Clement
758f738b8c
This PR is part of the documentation revamp. Once the full set of PRs is merged, we’ll publish the combined site to staging for review before updating docs.lancedb.com. Ingestion, search and integration examples are incomplete or inconsistent across languages, and some guidance misstates SDK behavior. This brings the guides and their example sources together. It includes complete Python, TypeScript and Rust workflows, isolated provider and reranker examples, the Superlinked corrections, and guidance on retries that can duplicate writes. Stacked on #4491. Each MDX file has one PR owner. Rebase onto main after #4491 is squash-merged. Local assembly, Mint validation and link checks pass. Example code is unchanged from the previously validated package. Publication stays held through integration.
97 lines
3.6 KiB
Python
97 lines
3.6 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright The LanceDB Authors
|
|
|
|
"""The published reranker examples in test_integrations.py write only to a directory they create.
|
|
|
|
Each example runs as readers see it: its marked region is extracted the way the snippet
|
|
generator extracts it, saved as a script and run offline. A stand-in embedding function answers for the
|
|
sentence-transformers model, stand-in rerankers replace the model-backed AnswerDotAI, ColBERT
|
|
and cross-encoder rerankers, and outbound connections are refused. The RRF, linear-combination
|
|
and MRR rerankers need no model and run unchanged. Sentinel tables in the owned home directory,
|
|
where these examples used to write, and in the temporary directory must be left unchanged.
|
|
"""
|
|
|
|
import importlib.util
|
|
import runpy
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
import lancedb.rerankers
|
|
import pytest
|
|
import test_provider_examples as provider
|
|
|
|
owned = provider.owned
|
|
SOURCE = Path(__file__).with_name("test_integrations.py")
|
|
GENERATOR = Path(__file__).parents[1] / "mdx_snippets_gen.py"
|
|
MODEL_RERANKERS = {
|
|
"reranking_answerdotai_usage": "AnswerdotaiRerankers",
|
|
"reranking_colbert_usage": "ColbertReranker",
|
|
"reranking_cross_encoder_usage": "CrossEncoderReranker",
|
|
}
|
|
LOCAL_RERANKERS = [
|
|
"reranking_linear_combination_usage",
|
|
"reranking_mrr_usage",
|
|
"reranking_rrf_usage",
|
|
]
|
|
|
|
|
|
def load_generator():
|
|
if "mdx_snippets_gen" not in sys.modules:
|
|
spec = importlib.util.spec_from_file_location("mdx_snippets_gen", GENERATOR)
|
|
module = importlib.util.module_from_spec(spec)
|
|
sys.modules[spec.name] = module # its dataclasses resolve annotations by module
|
|
spec.loader.exec_module(module)
|
|
return sys.modules["mdx_snippets_gen"]
|
|
|
|
|
|
def reader_code(name):
|
|
"""The example exactly as the snippet generator publishes it."""
|
|
generator = load_generator()
|
|
start, end = generator.markers_for_lang("py")
|
|
lines = SOURCE.read_text().splitlines(keepends=True)
|
|
snippet = generator.parse_snippets(lines, start, end)[name]
|
|
return generator.snippet_text(generator.dedent(snippet))
|
|
|
|
|
|
def run_example(owned, tmp_path, name):
|
|
script = tmp_path / "scripts" / f"{name}.py"
|
|
script.parent.mkdir(exist_ok=True)
|
|
script.write_text(reader_code(name))
|
|
existing = set(owned.temp_root.iterdir())
|
|
namespace = runpy.run_path(str(script))
|
|
|
|
after = [provider.snapshot(owned.connect, uri) for uri in owned.sentinels]
|
|
assert after == owned.before
|
|
created = set(owned.temp_root.iterdir()) - existing
|
|
assert len(created) == 1
|
|
(own_directory,) = created
|
|
assert len(owned.connected) == 1
|
|
assert owned.connected[0].is_relative_to(own_directory.resolve())
|
|
assert owned.connect(str(owned.connected[0])).list_tables().tables == ["test"]
|
|
return namespace
|
|
|
|
|
|
@pytest.mark.parametrize("name", sorted(MODEL_RERANKERS))
|
|
def test_model_reranker_example_writes_only_to_its_own_directory(
|
|
owned, monkeypatch, tmp_path, name
|
|
):
|
|
monkeypatch.setattr(
|
|
lancedb.rerankers, MODEL_RERANKERS[name], provider.OfflineReranker
|
|
)
|
|
|
|
namespace = run_example(owned, tmp_path, name)
|
|
|
|
(reranker,) = provider.OfflineReranker.instances
|
|
assert reranker.searches == ["vector", "fts", "hybrid"]
|
|
assert namespace["result"]
|
|
|
|
|
|
@pytest.mark.parametrize("name", LOCAL_RERANKERS)
|
|
def test_local_reranker_example_writes_only_to_its_own_directory(owned, tmp_path, name):
|
|
namespace = run_example(owned, tmp_path, name)
|
|
|
|
assert namespace["result"]
|
|
assert all("_relevance_score" in row for row in namespace["result"])
|
|
if name == "reranking_mrr_usage":
|
|
assert namespace["combined"].num_rows > 0
|