search

GreyDGL / PentestGPT

A GPT-empowered penetration testing tool
MIT License
6.67k stars 796 forks source link

Why use multiple classes for GPT / LocalLLM? #143

Closed krrishdholakia closed 7 months ago

krrishdholakia commented 11 months ago

Hey @GreyDGL interesting approach to abstracting llm providers - by creating separate classes chatgpt_api.py and gpt4all_api.py

Why do it this way - vs. making the completion call inside the init.py?

GreyDGL commented 11 months ago

Hey @krrishdholakia thanks for asking. I also see your PR and will review it soon. So the idea is that we want users to wrap up their own LLM modules (APIs, proxies, local LLMs, etc.) and directly import from the API class. This is required in security domain as ppl may want to use their own models. Integrating everything together is also a good idea, especially for those standard modules. I'll test the litellm package first and see how to proceed.

krrishdholakia commented 11 months ago

Hey @GreyDGL sounds good - let me know if you think anything is missing, happy to make modifications (to PR and package)