[Python] LangChain FAISS similarity_search returns empty list despite populated index

I’m trying to use LangChain with FAISS to build a simple document retriever. I’ve indexed several documents, but when I call similarity_search, I always get an empty list.

from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import FAISS
from langchain.schema import Document

docs = [
Document(page_content="Paris is the capital of France."),
Document(page_content="Berlin is the capital of Germany."),
Document(page_content="Tokyo is the capital of Japan.")
]

embeddings = OpenAIEmbeddings(openai_api_key="<API_KEY>")
vectorstore = FAISS.from_documents(docs, embeddings)

query = "What is the capital of Germany?"
results = vectorstore.similarity_search(query, k=2)

print(results) # Expected: at least one Document about Berlin, but this prints []


Environment:

• langchain==0.0.297
• faiss-cpu==1.7.4
• openai==0.27.4
• Python 3.9.16

Expected behavior: similarity_search should return the Document about Berlin, or at least a non-empty list.

What I’ve tried:

• Verified that docs contains the correct content.
• Checked that vectorstore.index has vectors (no errors when building).

Actual behavior: An empty list is returned every time, even for very simple queries.

Continue reading...