You can build your own Chat-GPT Clone : Here's how

Amidst all the hype of the GPT-4 which aint free ; What if I said , 'Yo, You can spin up your own clone of Open AI's chat-gpt on your own dear computer !' Interesting ?

To set the right expectation,

• This article doesn't explain how things work in any detail
• This might not work for all the machines.

I personally tried on mac book air m2 (8/ 128G) . Isn't the right host computer, for running the jacked model.

Before, we move ahead ,Here's a outcome , one can expect out of this article .

Pre-Requisites

• Docker Desktop
• Docker CLI
• Ollama

Ollama

Ollama ,in short, is a framework to run machine learning models (mostly it's LLM).

Ollama is natively supported on Mac & linux . Windows users can use WSL and install the linux version

Step 1 : Download Ollama

Dowload from here https://ollama.com/

Step 2 : Pulling the LLM

In this article, we'll use Meta's llama as our model . However , you can choose the model from the library

ollama run llama3        

You can start asking your llm alr

Once you're done, you can verify if ollama is installed properly by running the curl

curl https://localhost:11434        

Step 3: Let's beautify with an UI

It's all 'bout beauty , folks!

We'll be using open-webui ! Run the command below, to spin up a container running the client.

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main        

Now, go to the web browser and hit localhost:3000

Do note that the first thing you'll see is a sign-in /sign-up page.

Create an account if not already.

The first account created , will reserve the admin rights!

Boom! That's all 'bout it folks! You can download more models and select the model of your choice.