Skip to content

The agent engineering platform.

community: BM25Retriever preservation of document id (#27019)

Asi Greenholtsd34bf78f3bef
diff --git a/libs/community/langchain_community/retrievers/bm25.py b/libs/community/langchain_community/retrievers/bm25.py
index 543058c131..70910ce170 100644
--- a/libs/community/langchain_community/retrievers/bm25.py
+++ b/libs/community/langchain_community/retrievers/bm25.py
@@ -33,6 +33,7 @@ class BM25Retriever(BaseRetriever):
cls,
texts: Iterable[str],
metadatas: Optional[Iterable[dict]] = None,
+ ids: Optional[Iterable[str]] = None,
bm25_params: Optional[Dict[str, Any]] = None,
preprocess_func: Callable[[str], List[str]] = default_preprocessing_func,
**kwargs: Any,
@@ -42,6 +43,7 @@ class BM25Retriever(BaseRetriever):
Args:
texts: A list of texts to vectorize.
metadatas: A list of metadata dicts to associate with each text.
+ ids: A list of ids to associate with each text.
bm25_params: Parameters to pass to the BM25 vectorizer.
preprocess_func: A function to preprocess each text before vectorization.
**kwargs: Any other arguments to pass to the retriever.
@@ -61,7 +63,15 @@ class BM25Retriever(BaseRetriever):
bm25_params = bm25_params or {}
vectorizer = BM25Okapi(texts_processed, **bm25_params)
metadatas = metadatas or ({} for _ in texts)
- docs = [Document(page_content=t, metadata=m) for t, m in zip(texts, metadatas)]
+ if ids:
+ docs = [
+ Document(page_content=t, metadata=m, id=i)
+ for t, m, i in zip(texts, metadatas, ids)
+ ]
+ else:
+ docs = [
+ Document(page_content=t, metadata=m) for t, m in zip(texts, metadatas)
+ ]
return cls(
vectorizer=vectorizer, docs=docs, preprocess_func=preprocess_func, **kwargs
)
@@ -86,11 +96,14 @@ class BM25Retriever(BaseRetriever):
Returns:
A BM25Retriever instance.
"""
- texts, metadatas = zip(*((d.page_content, d.metadata) for d in documents))
+ texts, metadatas, ids = zip(
+ *((d.page_content, d.metadata, d.id) for d in documents)
+ )
return cls.from_texts(
texts=texts,
bm25_params=bm25_params,
metadatas=metadatas,
+ ids=ids,
preprocess_func=preprocess_func,
**kwargs,
)
diff --git a/libs/community/tests/unit_tests/retrievers/test_bm25.py b/libs/community/tests/unit_tests/retrievers/test_bm25.py
index ef40b25ba7..81afae84c7 100644
--- a/libs/community/tests/unit_tests/retrievers/test_bm25.py
+++ b/libs/community/tests/unit_tests/retrievers/test_bm25.py
@@ -43,3 +43,42 @@ def test_repr() -> None:
]
bm25_retriever = BM25Retriever.from_documents(documents=input_docs)
assert "I have a pen" not in repr(bm25_retriever)
+
+
[email protected]("rank_bm25")
+def test_doc_id() -> None:
+ docs_with_ids = [
+ Document(page_content="I have a pen.", id="1"),
+ Document(page_content="Do you have a pen?", id="2"),
+ Document(page_content="I have a bag.", id="3"),
+ ]
+ docs_without_ids = [
+ Document(page_content="I have a pen."),
+ Document(page_content="Do you have a pen?"),
+ Document(page_content="I have a bag."),
+ ]
+ docs_with_some_ids = [
+ Document(page_content="I have a pen.", id="1"),
+ Document(page_content="Do you have a pen?"),
+ Document(page_content="I have a bag.", id="3"),
+ ]
+ bm25_retriever_with_ids = BM25Retriever.from_documents(documents=docs_with_ids)
+ bm25_retriever_without_ids = BM25Retriever.from_documents(
+ documents=docs_without_ids
+ )
+ bm25_retriever_with_some_ids = BM25Retriever.from_documents(
+ documents=docs_with_some_ids
+ )
+ for doc in bm25_retriever_with_ids.docs:
+ assert doc.id is not None
+ for doc in bm25_retriever_without_ids.docs:
+ assert doc.id is None
+ for doc in bm25_retriever_with_some_ids.docs:
+ if doc.page_content == "I have a pen.":
+ assert doc.id == "1"
+ elif doc.page_content == "Do you have a pen?":
+ assert doc.id is None
+ elif doc.page_content == "I have a bag.":
+ assert doc.id == "3"
+ else:
+ raise ValueError("Unexpected document")