1

Hello MongoDB community,

Context

I am coding a Report Maker program using gemma-2b to summarize long audio transcription and formalize it into a report, i fine-tuned gemma for it to be consistent in that task. However I though that having a RAG over gemma would provide very cool features to let the user query specific part from the whole transcription.

Issue

I follow the video and the blogpost tutorial about RAG over gemma and yet, strictly use the same code but the vector search query is always null/empty. I wonder if I missed something or and update.

Below my actual code in visual studio:

Python Code

import pandas as pd

import json

from sentence_transformers import SentenceTransformer
import pymongo
import os
from dotenv import load_dotenv
from datasets import load_dataset
load_dotenv()

Get the MONGO URI key from the environment variables

MONGO_URI = os.getenv(‘MONGO_DB_URI’)

Load Dataset

from datasets import load_dataset
import pandas as pd

MongoDB/embedded_movies · Datasets at Hugging Face

dataset = load_dataset(“MongoDB/embedded_movies”)

Convert the dataset to a pandas DataFrame

dataset_df = pd.DataFrame(dataset[‘train’])

Remove data point where plot column is missing

dataset_df = dataset_df.dropna(subset=[‘fullplot’])
print(“\nNumber of missing values in each column after removal:”)
print(dataset_df.isnull().sum())

Remove the plot_embedding from each data point in the dataset as we are going to create new embeddings with an open-source embedding model from Hugging Face: gte-large

dataset_df = dataset_df.drop(columns=[‘plot_embedding’])

thenlper/gte-large · Hugging Face

embedding_model = SentenceTransformer(“thenlper/gte-large”)

def get_embedding(text: str) → list[float]:
if not text.strip():
print(“Attempted to get embedding for empty text.”)
return
embedding = embedding_model.encode(text)
return embedding.tolist()

dataset_df[“embedding”] = dataset_df[“fullplot”].apply(get_embedding)

def get_mongo_client(mongo_uri):
“”“Establish connection to the MongoDB.”“”
try:
client = pymongo.MongoClient(mongo_uri)
print(“Connection to MongoDB successful”)
return client
except pymongo.errors.ConnectionFailure as e:
print(f"Connection failed: {e}")
return None

mongo_uri = MONGO_URI
if not mongo_uri:
print(“MONGO_URI not set in environment variables”)

mongo_client = get_mongo_client(mongo_uri)

Ingest data into MongoDB

db = mongo_client[“movies”]
collection = db[“movie_collection_2”]

Delete any existing records in the collection

collection.delete_many({})

documents = dataset_df.to_dict(‘records’)
collection.insert_many(documents)
print(“Data ingestion into MongoDB completed”)

def vector_search(user_query, collection):
“”"
Perform a vector search in the MongoDB collection based on the user query.

Args:
user_query (str): The user's query string.
collection (MongoCollection): The MongoDB collection to search.

Returns:
list: A list of matching documents.
"""

# Generate embedding for the user query
query_embedding = get_embedding(user_query)

if query_embedding is None:
    return "Invalid query or embedding generation failed."

# Define the vector search pipeline
pipeline = [
    {
        "$vectorSearch": {
            "index": "vector_index",
            "queryVector": query_embedding,
            "path": "embedding",
            "numCandidates": 150,  # Number of candidate matches to consider
            "limit": 4,  # Return top 4 matches
        }
    },
    {
        "$project": {
            "_id": 0,  # Exclude the _id field
            "fullplot": 1,  # Include the plot field
            "title": 1,  # Include the title field
            "genres": 1,  # Include the genres field
            "score": {"$meta": "vectorSearchScore"},  # Include the search score
        }
    },
]

# Execute the search
results = collection.aggregate(pipeline)
return list(results)

def get_search_result(query, collection):

get_knowledge = vector_search(query, collection)

search_result = ""
for result in get_knowledge:
    search_result += f"Title: {result.get('title', 'N/A')}, Plot: {result.get('fullplot', 'N/A')}\n"

return search_result

Conduct query with retrieval of sources

query = “What is the best romantic movie to watch and why?”
source_information = get_search_result(query, collection)
combined_information = f"Query: {query}\nContinue to answer the query by using the Search Results:\n{source_information}."
print(combined_information)

1 Like
2

Hi William,

Thanks for making a post about your issue.

A common reason for having null results from the vector search operation can be due to not having the vector search index definition set up correctly.

Can you please verify that your vector search index definition, created using MongoDB Atlas looks as follow:

   {
      "fields": [
        {
          "numDimensions": 1024,
          "path": "embedding",
          "similarity": "cosine",
          "type": "vector"
        }
      ]
    }

Also please do check the name of the vector search index definition on MongoDB Atlas is: vector_index

Thanks,

Richmond

1 Like