I am unable to get search results with Atlas Vector Search Index

Kevin_Tsai · 2024-04-15T23:47:34.079Z

Hi, I’m testing MongoDB Atlas for the first time with the RAG with Atlas Vector Search, Langchain, and OpenAI tutorial: RAG with Atlas Vector Search, LangChain, and OpenAI.

I am able to import my documents with their embeddings into a collection and added a Vector Search Index, but do not get any results when performing a similarity search. However, I do get results when I add a regular Search Index. My Atlas version is 7.0.8.

Vector Search Index:
{
“fields”: [
{
“numDimensions”: 1536,
“path”: “embedding”,
“similarity”: “cosine”,
“type”: “vector”
},
{
“path”: “page”,
“type”: “filter”
}
]
}

Search Index:
{
“mappings”: {
“dynamic”: true,
“fields”: {
“embedding”: [
{
“dimensions”: 1536,
“similarity”: “cosine”,
“type”: “knnVector”
}
]
}
}
}

Prakul_Agarwal · 2024-04-16T00:06:48.670Z

Hi @Kevin_Tsai, Firstly, Welcome to the MongoDB forum!
Can you share what version of LangChain you are using?

Prakul_Agarwal · 2024-04-16T00:09:02.347Z

Also btw, that particular article you are following is dated and you can refer to the up to date documentation here - https://www.mongodb.com/docs/atlas/atlas-vector-search/ai-integrations/langchain/

Kevin_Tsai · 2024-04-16T21:37:14.915Z

load_data.py:

from langchain_community.document_loaders import PyPDFLoader
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_mongodb import MongoDBAtlasVectorSearch
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain.prompts import PromptTemplate
from langchain.text_splitter import RecursiveCharacterTextSplitter
from pymongo import MongoClient
import key_param

Connect to your Atlas cluster

client = MongoClient(key_param.MONGO_URI)

Define collection and index name

db_name = “langchain_db”
collection_name = “test”
atlas_collection = client[db_name][collection_name]
vector_search_index = “vector_index”

Load the PDF

loader = PyPDFLoader(
“https://query.prod.cms.rt.microsoft.com/cms/api/am/binary/RE4HkJP”)
data = loader.load()

Split PDF into documents

text_splitter = RecursiveCharacterTextSplitter(
chunk_size=200, chunk_overlap=20)
docs = text_splitter.split_documents(data)

Create the vector store

vector_search = MongoDBAtlasVectorSearch.from_documents(
documents=docs,
embedding=OpenAIEmbeddings(
openai_api_key=key_param.openai_api_key, disallowed_special=()),
collection=atlas_collection,
index_name=vector_search_index
)

extract_information.py:

import pprint
from langchain_community.document_loaders import PyPDFLoader
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_mongodb import MongoDBAtlasVectorSearch
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain.prompts import PromptTemplate
from langchain.text_splitter import RecursiveCharacterTextSplitter
from pymongo import MongoClient
import key_param

client = MongoClient(key_param.MONGO_URI)
db_name = “langchain_demo”
collection_name = “collection_of_text_blobs”
atlas_collection = client[db_name][collection_name]
vector_search_index = “vector_index”

Define the text embedding model

embedding = OpenAIEmbeddings(
openai_api_key=key_param.openai_api_key, disallowed_special=())

Initialize the Vector Store

vector_search = MongoDBAtlasVectorSearch(
collection=atlas_collection,
embedding=embedding,
)

query = “MongoDB Atlas security”
results = vector_search.similarity_search(query)
print(“\nvector_search.similarity_search(query):”)
pprint.pprint(results)

Instantiate Atlas Vector Search as a retriever

retriever = vector_search.as_retriever(
search_type=“similarity”,
search_kwargs={“k”: 10, “score_threshold”: 0.75}
)

Define a prompt template

template = “”"

Use the following pieces of context to answer the question at the end.
If you don’t know the answer, just say that you don’t know, don’t try to make up an answer.

{context}

Question: {question}
“”"
custom_rag_prompt = PromptTemplate.from_template(template)

llm = ChatOpenAI(openai_api_key=key_param.openai_api_key, temperature=0)

def format_docs(docs):
return “\n\n”.join(doc.page_content for doc in docs)

Construct a chain to answer questions on your data

rag_chain = (
{“context”: retriever | format_docs, “question”: RunnablePassthrough()}
| custom_rag_prompt
| llm
| StrOutputParser()
)

Prompt the chain

question = “How can I secure my MongoDB Atlas cluster?”
answer = rag_chain.invoke(question)

print("Question: " + question)
print("Answer: " + answer)

Return source documents

documents = retriever.get_relevant_documents(question)
print(“\nSource documents:”)
pprint.pprint(documents)

results:
vector_search.similarity_search(query):

Question: How can I secure my MongoDB Atlas cluster?
Answer: Some ways to secure your MongoDB Atlas cluster include enabling network encryption, setting up IP whitelisting, enabling authentication mechanisms like username/password or LDAP, enabling role-based access control, enabling auditing, and regularly updating your MongoDB version to the latest stable release with security patches. Additionally, you can consider implementing encryption at rest and using VPC peering for added security measures.

If you need more specific guidance on securing your MongoDB Atlas cluster, it would be best to refer to the official MongoDB documentation or consult with a database security expert.

Source documents:

Kevin_Tsai · 2024-04-16T21:37:19.640Z

I updated my code using the new docs and I get a response from the Q/A, but I do not see any relevant source documents. Also, vector_search.similarity_search(query) does not return any results.

Kevin_Tsai · 2024-04-16T21:37:23.612Z

Thanks, Prakul!

langchain==0.1.16
langchain-community==0.0.32
langchain-core==0.1.42
langchain-mongodb==0.1.3
langchain-openai==0.1.3
langchain-text-splitters==0.0.1