Building a Shop Support Chatbot with OpenAI and Python

Introduction

Chatbots have become an increasingly popular way for businesses to provide customer support. They’re available 24/7, can handle a large volume of queries, and can provide personalized responses. In this article, we’ll show you how to create a shop support chatbot using OpenAI and Python. Specifically, we’ll focus on supporting item pricing queries, but you can use the same techniques to support other types of queries.

Prerequisites

Before we get started, you’ll need to sign up for an OpenAI API key and store it in an environment variable. You’ll also need to install the OpenAI API client for Python using pip.

pip install openai

Retrieving item prices

To create a shop support chatbot, we’ll need to be able to retrieve item prices. We can do this by storing a list of items and their prices in a dictionary. For example:

items = {
    "apple": 0.5,
    "banana": 0.25,
    "orange": 0.75,
    "pear": 0.6
}

We can then create a function that takes an item name as input and returns the price of that item. For example:

def get_item_price(item):
    return items.get(item, None)

This function returns the price of the item if it exists in the dictionary, or None if it doesn’t exist.

Generating responses using OpenAI:

Now that we can retrieve item prices, we can create a chatbot that supports item pricing queries. We can use OpenAI’s GPT-3 language model to generate responses to user queries.

First, we’ll create an OpenAI API client:

import openai_secret_manager

assert "openai" in openai_secret_manager.get_services()
secrets = openai_secret_manager.get_secret("openai")

import openai
openai.api_key = secrets["api_key"]

We can then define a function that takes a user query as input and uses GPT-3 to generate a response. For example:

def generate_response(query):
    prompt = f"What is the price of {query}?"
    response = openai.Completion.create(
        engine="davinci",
        prompt=prompt,
        max_tokens=1024,
        n=1,
        stop=None,
        temperature=0.5,
    )

    message = response.choices[0].text.strip()
    if message == "None":
        message = "I'm sorry, I don't know the price of that item."

    return message

This function takes a user query, such as “What is the price of an apple?”, and generates a response using GPT-3. We’re using the Davinci engine, which is the most powerful OpenAI language model, and setting the max_tokens parameter to 1024, which limits the length of the response. We’re also setting the temperature parameter to 0.5, which controls the creativity of the response.

Putting it all together:

Finally, we can create a loop that prompts the user for a query, retrieves the item price using the get_item_price function, and generates a response using the generate_response function. For example:

while True:
    query = input("What do you want to know the price of? ")
    price = get_item_price(query)
    if price is not None:

We can use string formatting to generate a response that includes both the item name and price. For example:

message = f"The price of {query} is {price}."

We can then print the response to the user:

print(message)

If the item doesn’t exist in the dictionary, we can use the generate_response function to generate a response using GPT-3:

else:
    message = generate_response(query)
    print(message)

Putting it all together, our final code looks like this:

import openai_secret_manager

assert "openai" in openai_secret_manager.get_services()
secrets = openai_secret_manager.get_secret("openai")

import openai
openai.api_key = secrets["api_key"]

items = {
    "apple": 0.5,
    "banana": 0.25,
    "orange": 0.75,
    "pear": 0.6
}

def get_item_price(item):
    return items.get(item, None)

def generate_response(query):
    prompt = f"What is the price of {query}?"
    response = openai.Completion.create(
        engine="davinci",
        prompt=prompt,
        max_tokens=1024,
        n=1,
        stop=None,
        temperature=0.5,
    )

    message = response.choices[0].text.strip()
    if message == "None":
        message = "I'm sorry, I don't know the price of that item."

    return message

while True:
    query = input("What do you want to know the price of? ")
    price = get_item_price(query)
    if price is not None:
        message = f"The price of {query} is {price}."
        print(message)
    else:
        message = generate_response(query)
        print(message)

Conclusion:

In this blog, we’ve shown you how to create a shop support chatbot using OpenAI and Python. We’ve demonstrated how to retrieve item prices using a dictionary, how to generate responses using OpenAI’s GPT-3 language model, and how to put it all together in a loop that prompts the user for a query. You can use these techniques to create other types of chatbots that support different types of queries.

Suggestions

Here are some suggestions to improve the shop support chatbot:

1. Add more items: The current version of the chatbot only supports a few items. You can add more items to the items dictionary to make it more useful.
2. Handle synonyms: The chatbot only recognizes items by their exact names. You can use a synonym library like WordNet or spaCy to recognize synonyms and map them to the correct item.
3. Improve response generation: The generate_response function is a simple implementation that just sends a prompt to OpenAI’s API and returns the response. You can improve the response generation by using more sophisticated models and techniques like sentiment analysis, intent recognition, and dialogue management.
4. Add a feedback loop: You can add a feedback loop that collects user feedback on the bot’s responses and uses it to improve the bot’s performance over time.
5. Add user context: You can add user context to the chatbot by storing information about the user’s preferences and previous interactions. This can help the bot provide more personalized and relevant responses.
6. Add a GUI: You can add a graphical user interface (GUI) to the chatbot using a framework like tkinter or PyQt. This can make the chatbot more user-friendly and accessible to non-technical users.

By implementing these improvements, you can make the shop support chatbot more useful, intelligent, and user-friendly.