Hybrid search (keyword + semantic) in LangChain - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ component_behavior

intermediate

What is the output of this LangChain hybrid search snippet?

Consider this LangChain code combining keyword and semantic search. What does the results list contain after running?

LangChain

from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
from langchain.chains import RetrievalQA
from langchain.llms import OpenAI

embeddings = OpenAIEmbeddings()
vectorstore = FAISS.load_local('faiss_index', embeddings)

query = 'Explain hybrid search benefits'

results = vectorstore.hybrid_search(query, k=3, alpha=0.5)

print([doc.page_content for doc in results])

A A list of 3 documents ranked only by keyword matching to the query

B A list of 3 documents ranked by a mix of keyword and semantic similarity to the query

C A list of 3 documents ranked only by semantic similarity to the query

D An error because FAISS does not support hybrid_search method

💡 Hint

Think about what 'hybrid_search' means combining two search types.

The hybrid_search method in LangChain's FAISS vectorstore returns documents ranked by a weighted combination of keyword and semantic similarity, controlled by the alpha parameter.

📝 Syntax

intermediate

Which option correctly calls hybrid search with alpha parameter in LangChain?

You want to perform a hybrid search with 5 results and alpha 0.7. Which code snippet is correct?

LangChain

query = 'Find documents about AI'

# Choose the correct call to hybrid_search

A results = vectorstore.hybrid_search(query, k=5, alpha=0.7)

B results = vectorstore.hybrid_search(query, alpha=0.7, k=5)

C results = vectorstore.hybrid_search(k=5, query=query, alpha=0.7)

D results = vectorstore.hybrid_search(query, 5, 0.7)

💡 Hint

Check the method signature and argument order in LangChain docs.

The hybrid_search method expects the query first, then k, then alpha as keyword arguments. Option A matches this correctly.

🔧 Debug

advanced

Why does this hybrid search code raise an AttributeError?

This code snippet raises AttributeError: 'FAISS' object has no attribute 'hybrid_search'. What is the cause?

LangChain

from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings

vectorstore = FAISS.load_local('index_dir', OpenAIEmbeddings())
results = vectorstore.hybrid_search('query text', k=3, alpha=0.4)

A The FAISS class version used does not implement hybrid_search method

B The alpha parameter must be a string, not a float

C The load_local method returns None, so vectorstore is None

D The query argument must be passed as a keyword argument

💡 Hint

Check if your LangChain and FAISS versions support hybrid_search.

Older versions of LangChain's FAISS vectorstore do not have the hybrid_search method implemented, causing AttributeError.

❓ state_output

advanced

What is the effect of changing alpha in hybrid search?

Given this code snippet, what happens to the search results if alpha changes from 0.2 to 0.8?

LangChain

results_low_alpha = vectorstore.hybrid_search('machine learning', k=3, alpha=0.2)
results_high_alpha = vectorstore.hybrid_search('machine learning', k=3, alpha=0.8)

# Compare results_low_alpha and results_high_alpha

A Alpha controls the number of results returned, so 0.8 returns more results

B Higher alpha (0.8) makes results rely more on semantic similarity; lower alpha (0.2) relies more on keyword matching

C Higher alpha (0.8) makes results rely more on keyword matching; lower alpha (0.2) relies more on semantic similarity

D Alpha has no effect on results; it is ignored by hybrid_search

💡 Hint

Alpha balances keyword vs semantic importance in hybrid search.

In LangChain's hybrid_search, alpha closer to 1 means more weight on keyword matching; closer to 0 means more weight on semantic similarity.

🧠 Conceptual

expert

Which statement best describes hybrid search in LangChain?

Select the most accurate description of how hybrid search works in LangChain's vectorstores.

A Hybrid search indexes documents twice, once for keywords and once for embeddings, and merges results at query time without weighting

B Hybrid search runs keyword search first, then filters results by semantic similarity threshold

C Hybrid search uses only semantic similarity but boosts scores of documents containing exact query keywords

D Hybrid search combines keyword matching scores and semantic similarity scores using a weighted sum controlled by alpha to rank documents

💡 Hint

Think about how alpha parameter influences scoring in hybrid search.

LangChain's hybrid search calculates a combined score by weighting keyword and semantic similarity scores with alpha, then ranks documents accordingly.