LangChain: Question Answering with Qdrant and GPT-3.5 Turbo

This article describes a Python script that leverages LangChain, Qdrant, and the GPT-3.5 Turbo model to perform question-answering on text data. The script combines various libraries to create an end-to-end pipeline for text retrieval and question answering.


The provided Python script showcases how to integrate multiple libraries to build a powerful question-answering system:

  • LangChain: A library that facilitates seamless chaining and orchestration of various NLP components.
  • Qdrant: An open-source vector database that allows efficient storage and retrieval of high-dimensional vectors.
  • HuggingFace Transformers: A popular library for transformer-based language models and embeddings.
  • qdrant_client: A Python client for the Qdrant database.


Before running the script, ensure you have the following installed:

  1. Python 3.x
  2. Docker (for setting up the Qdrant database)

Additionally, install the required Python libraries by running:

pip install langchain.vectorstores langchain.embeddings langchain chat_models qdrant_client transformers

Setup Qdrant Database

To use the Qdrant database, follow these steps:

  1. Pull the Qdrant Docker image:
sudo docker pull qdrant/qdrant
  1. Run the Qdrant Docker container:
sudo docker run -p 6333:6333 qdrant/qdrant

This will set up the Qdrant database on your localhost, ready to be used.

How the Script Works

The Python script demonstrates an end-to-end question-answering pipeline using LangChain, Qdrant, and the GPT-3.5 Turbo model.

Initializing LangChain and HuggingFace Embeddings

The script starts by importing the required libraries, including LangChain and HuggingFace Embeddings. It also initializes the GPT-3.5 Turbo model for question-answering.

Connecting to Qdrant

The script creates a connection to the Qdrant database using the qdrant_client Python library. It sets up a collection named “pdfz” to store and retrieve text embeddings.

LangChain Setup

Next, the script configures the LangChain components:

  • ChatOpenAI: An OpenAI GPT-3.5 Turbo model used for generating answers to questions.
  • HuggingFaceEmbeddings: A Sentence Transformers model for generating text embeddings.

The code:

Creating the Qdrant Object

The Qdrant object is initialized with the Qdrant client, collection name, and embeddings. This object acts as a bridge between LangChain’s question-answering component and the Qdrant database.

Question Answering

The script defines a sample question and performs the question-answering process using the RetrievalQA class from LangChain. The retriever utilizes the Qdrant database for efficient retrieval of relevant embeddings, and the answer is generated using the GPT-3.5 Turbo model.

The result is printed to the console, displaying the answer to the sample question.

Running the Script

Before running the script, ensure you have completed the following steps:

  1. Installed all required libraries.
  2. Set up the Qdrant database using Docker.

After fulfilling the prerequisites, simply run the Python script. It will perform question-answering on the sample question and print the answer to the console.

Note: Modify the sample question and experiment with different input queries to obtain various answers.


The provided Python script demonstrates the power of combining LangChain, Qdrant, and the GPT-3.5 Turbo model for building an efficient and effective question-answering system. By utilizing these libraries, developers and data scientists can easily create complex NLP pipelines, enable vector storage and retrieval, and provide accurate answers to user queries. This integration showcases the potential for advanced NLP applications, including chatbots, knowledge bases, and more.

For more information about the Python libraries used in the script, refer to their respective documentation: