Build a Custom Trained Chat GPT in 5 Minutes

This article will teach you how to run a local chat GPT trained on your custom data in 5 minutes.
A custom-trained LLM bot can be great for scanning large amounts of data and having a domain-specific conversation. For example, you can train your bot on a company knowledge base, educational materials, financial\ legal, or even sharing your trip ideas for planning your perfect route 🙂

This can be done in 5 minutes; let’s get to work.

Photo by Volodymyr Hryshchenko on Unsplash

Prerequisites:

• This article assumes you have Python 3.*.* and Pip installed (can be downloaded here). To verify this, please open your terminal and type: python --version OR python3 --version
  And if you have it correctly set up, you should get back something like the following:

(Make sure to use the correct Python reference throughout this guide consistently)

• pip --version OR pip3 --version
  And if you have it correctly set up, you should get back something like the following:

(Make sure to use the correct Pip reference throughout this guide consistently)

• This article assumes you already have an OpenAi account (which can be created here)
• Prepare some data in English; I suggest starting with less than 100MB and slowly adding more. For example, data can be in PDF, CSV, or TXT formats.

Let’s get started

(1) In the terminal:

• Upgrade Pip (Python package manager):
  python -m pip install -U pip
• Install the following libraries:
  pip install openai gpt_index==0.4.24 langchain==0.0.118 PyPDF2 PyCryptodome gradio

OpenAI — Large Language Model (LLM) to create and train the AI chatbot.
GPT Index — for connecting with your data.
PyPDF2 — for phrasing PDFs.
Gradio — chatbot UI.
Lang Chain — a framework for developing applications powered by language models.

• Now let’s create your bot. Navigate to the folder you want to create the bot in:
  mkdir chatbot
cd chatbot
mkdir docs
touch app.py

(2) In your favorite code or text editor, open app.py, and paste the following code into it, and save it afterward:

from gpt_index import SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper
from langchain.chat_models import ChatOpenAI
import gradio as gr
import sys
import os

os.environ["OPENAI_API_KEY"] = 'Your Secret API Key'

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=ChatOpenAI(temperature=0.7, model_name="gpt-3.5-turbo", max_tokens=num_outputs))
  documents = SimpleDirectoryReader(directory_path).load_data()
  index = GPTSimpleVectorIndex(documents, llm_predictor=llm_predictor, prompt_helper=prompt_helper)
  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.components.Textbox(lines=7, label="Enter your text"),
                     outputs="text",
                     title="Custom-trained AI Chatbot")

index = construct_index("docs")
iface.launch(share=True)

(3) Now move your data files to the newly created /chatbot/docs folder.

(4) In the browser:

1. Get a free API key for OpenAi, by navigating to: https://platform.openai.com/account/api-keys
2. Sign in if necessary.
3. Click the “+ Create new secret key.”
4. Give your key a name and click “Create secret Key.”
5. Copy the key and save it for the next steps.
6. Paste the key in app.py, replacing ‘Your Secret API Key.’

(5) Back in the terminal on the /chatbot/ folder run: python app.py

(6) Open your favorite browser and open this URL: http://127.0.0.1:7860

That’s it. You are done! And should see something like this:

Now it’s up to you to find valuable information to teach your bot and learn how you can benefit from it.