API chatgpt deploy on local environment? Is it possible? i have had no luck in finding any answer

API
1

Hello, not sure if it was asked before or not, but i was unable to find anything regarding this.

Is there any way to use the API, to make a similar chatgpt 4.5 with data analysis and code interpreter, basically same functionalities, to implement on a local environment, say either in coding environment, or most important, which is the main object of my interest that is on a linux (debian 11) VPS to handle site files, add/modify/remove, improve and basically interact and adapt frontend/backend, and basically handle the server after it is deployed and surveyed on the functionalities, logic etc, so that i simply ask via a chat input request so that he can edit, and basically alter files and code in a certain environment? Maybe im too futuristic delusional but since it can do so on the site with limited context page and own environment to handle code for test and debugging, i assume, api playground (pay/use) etc, could apply in the same way, if the environment is local and is given access to a certain part that he can handle and via api, to run, test debug, implement, adjust files on a pay/use ratio.

Any reply in any direction would be helpful, thank you!

1 Like
2

You want it to run fully locally? Or call the API and use it on your own things?

1 Like
3

Hi and welcome to the community!

I am assuming you want to run the model locally.

Short answer: No. Not possible.

Expanding a little bit: unless you can make OpenAI an offer they can’t refuse (money wise) there is no way to deploy Chat GPT locally.
You would have to dig into the open source community projects and build a solution that fits all your needs. If you are looking for somewhat GPT 3.5 like performance I’d assume it’s possible building on top of existing solutions.

5

no, obviously not, just call the api and use it on my own environment.

6

Have you tried GitHub - SamurAIGPT/Open-Custom-GPT: Create Custom GPT and add/embed on your site using Assistants api . It is built for the same purpose, allows you to run all the functionalities with an Openai key using Assistants api

7

That sounds like a good way to crash or corrupt a server.

However, if foolhardy, it can be done.

Code interpreter is having the AI emit a “python” function that, instead of a JSON for a RESTful API, is Python code. OpenAI runs this in a Python Jupyter notebook sandbox virtual environment, giving a file system that can’t be damaged, and prohibiting access to networks.

With the correct system prompting and function specification, one can get the code interpreter AI to emit its code-writing to you, and then you could exec() that python and catch its return. And let the fun begin.