How to beginner · 3 min read

How to use retriever in LlamaIndex

Q: How to use retriever in LlamaIndex

Use the Retriever interface in LlamaIndex to query your indexed documents efficiently. Initialize a retriever from your index (e.g., VectorStoreIndex) and call retriever.retrieve(query) to get relevant documents for your query.

Quick answer

Use the Retriever interface in LlamaIndex to query your indexed documents efficiently. Initialize a retriever from your index (e.g., VectorStoreIndex) and call retriever.retrieve(query) to get relevant documents for your query.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install llama-index openai

Setup

Install llama-index and set your OpenAI API key as an environment variable.

Run pip install llama-index openai
Set environment variable OPENAI_API_KEY with your OpenAI API key

bash

pip install llama-index openai

Step by step

This example shows how to create a VectorStoreIndex from documents, get its retriever, and query it for relevant documents.

python

import os
from llama_index import VectorStoreIndex, SimpleDirectoryReader

# Load documents from a directory
documents = SimpleDirectoryReader('data').load_data()

# Create a vector index from documents
index = VectorStoreIndex.from_documents(documents)

# Get the retriever from the index
retriever = index.as_retriever()

# Query string
query = "What is the main topic of the documents?"

# Retrieve relevant documents
results = retriever.retrieve(query)

# Print retrieved documents content
for i, doc in enumerate(results):
    print(f"Document {i+1}:\n{doc.get_text()}\n")

output

Document 1:
The main topic of the documents is about AI and machine learning.

Document 2:
These documents discuss the applications of LlamaIndex in data retrieval.

Common variations

You can customize the retriever behavior by passing parameters like search_kwargs to control the number of results or similarity thresholds. Also, you can use other index types like GPTVectorStoreIndex or async retrieval if supported.

python

retriever = index.as_retriever(search_kwargs={'k': 3})  # Retrieve top 3 documents
results = retriever.retrieve(query)

# Async example (if supported by your environment)
# import asyncio
# async def async_retrieve():
#     results = await retriever.aretrieve(query)
#     for doc in results:
#         print(doc.get_text())
# asyncio.run(async_retrieve())

Troubleshooting

If retrieval returns empty results, ensure your documents are properly loaded and indexed.
Check your OpenAI API key environment variable is set correctly.
For slow retrieval, consider reducing the number of documents or tuning search_kwargs.

✅

Key Takeaways

Use index.as_retriever() to get a retriever for querying documents efficiently.
Customize retrieval with search_kwargs to control result count and similarity.
Ensure documents are loaded and indexed before retrieval to avoid empty results.

Verified 2026-04 · gpt-4o, llama-3.2

Verify ↗