The agent engineering platform.
Python agentsaiai-agentsanthropicchatgptdeepagentsenterpriseframeworkgeminigenerative-ailangchainlanggraphllmmultiagentopen-sourceopenaipydanticpythonragtypescript
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.pyindex 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.pyindex 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")