How to Add Your Own Data in GPT to Create a Customized Chatbot

Are you looking to create a customized chatbot with your own data? GPT (Generative Pre-trained Transformer) is a powerful language model that can be fine-tuned on your specific dataset. This tutorial will show you how to add your data to GPT and create a customized chatbot using Python.

We will be using the following code to demonstrate the process step-by-step:

from gpt_index import SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper, ServiceContext
from langchain import OpenAI
import gradio as gr
import sys
import os

os.environ["OPENAI_API_KEY"] = 'sk-*********************'

def construct_index(directory_path):
    max_input_size = 4096
    num_outputs = 512
    max_chunk_overlap = 20
    chunk_size_limit = 600
    
    prompt_helper = PromptHelper(max_input_size, num_outputs, max_chunk_overlap, chunk_size_limit=chunk_size_limit)
    llm_predictor = LLMPredictor(llm=OpenAI(temperature=0.7, model_name="text-davinci-003", max_tokens=num_outputs))
        documents = SimpleDirectoryReader(directory_path).load_data()
        print("-------------")
        print(documents)
        service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor, prompt_helper=prompt_helper)
        index = GPTSimpleVectorIndex.from_documents(documents, service_context=service_context)
        index.save_to_disk('index.json')
        return index
    def chatbot(input_text):
        index = GPTSimpleVectorIndex.load_from_disk('index.json')
        response = index.query(input_text, response_mode="compact")
        return response.response
    iface = gr.Interface(fn=chatbot,
                         inputs=gr.inputs.Textbox(lines=7, label="Enter your text"),
                         outputs="text",
                         title="My AI Chatbot")
    index = construct_index("gdrive/MyDrive/GPTDATA")
    iface.launch(share=True)

Step 1: Prepare your data To create a customized chatbot, you need to prepare your own dataset. You can create your dataset in a simple text file, (I have added my data as PDF and uploaded it to my google drive and connect it in code) where each line represents a new message or input for the chatbot. The dataset can be as large or as small as you want, depending on the complexity of the chatbot you want to create.

Step 2: Install the required libraries To create a customized chatbot using GPT, you need to install the following libraries: gpt_index, langchain, gradio.

You can install these libraries using the pip package manager with the following command:

pip install gpt_index langchain gradio

Step 3: Import the required libraries In the first line of the code, we import all the required libraries for creating a customized chatbot. These libraries include SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper, ServiceContext, OpenAI, gradio, sys, and os.

from gpt_index import SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper, ServiceContext
from langchain import OpenAI
import gradio as gr
import sys
import os

Step 4: Set up the OpenAI API

To use OpenAI’s GPT model, you need to set up your API key. You can get your API key from the OpenAI website. Once you have your API key, you can set it up in the following line of code:

os.environ["OPENAI_API_KEY"] = 'sk-****************************'

Step 5: Construct the GPT index The construct_index function is used to construct the GPT index. This function takes the directory path of the text file containing your dataset as input. In this function, we set the parameters for the GPT model, including the maximum input size, the number of outputs, and the chunk size limit. We also initialize the PromptHelper and LLMPredictor objects using the OpenAI API. Finally, we load the data from the text file using SimpleDirectoryReader, initialize the ServiceContext, and construct the GPT index using GPTSimpleVectorIndex.

def construct_index(directory_path):
    max_input_size = 4096
    num_outputs = 512
    max_chunk_overlap = 20
    chunk_size_limit = 600

    prompt_helper = PromptHelper(max_input_size, num_outputs, max_chunk_overlap, chunk_size_limit=chunk_size_limit)
    
    llm_predictor = LLMPredictor(llm=OpenAI(temperature=0.7, model_name="text-davinci-003", max_tokens=num_outputs))
    documents = SimpleDirectoryReader(directory_path).load_data()
    service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor, prompt_helper=prompt_helper)
    
    index = GPTSimpleVectorIndex.from_documents(documents, service_context=service_context)
    index.save_to_disk('index.json')
    
    return index

Step 6: Create the chatbot function The chatbot function is used to generate a response from the GPT model. This function takes an input text message as input and returns a response generated by the GPT model. In this function, we load the GPT index from the disk using GPTSimpleVectorIndex, and then use the query method to generate a response from the GPT model.

def chatbot(input_text):
    index = GPTSimpleVectorIndex.load_from_disk('index.json')
    response = index.query(input_text, response_mode="compact")
    return response.response

Step 7: Set up the Gradio interface The iface object is used to set up the Gradio interface for the chatbot. We use the gr.Interface function to set up the interface. This function takes the chatbot function as input, along with the input and output types of the interface.

iface = gr.Interface(fn=chatbot,
                     inputs=gr.inputs.Textbox(lines=7, label="Enter your text"),
                     outputs="text",
                     title="My AI Chatbot")

Step 8: Launch the chatbot In the final step, we call the construct_index function to create the GPT index from the text file containing our dataset. We then launch the Gradio interface using the iface.launch method.

index = construct_index("gdrive/MyDrive/GPTDATA")
iface.launch(share=True)

In this tutorial, we have shown you how to create a customized chatbot using your data and GPT. Following these steps, you can easily create and customize your own chatbot. Note: this is not the exact code I have been using. I have removed some complexity and unnecessary code from actually what I am working with, so you might face bugs in the above code, but you can fix it very easily. Let me know in the comment section if you need any help.