Question - Any plans to support local LLM models?

[0xm4v3rick]

Hi Team,

Thanks for making this tool. Are there any plans to support local LLM model for this as an alternative to GPT?
Pentest contains sensitive info and it would be good to not share this with a SaaS solution and keep it local. An configuration option to have connect to local model would be great to have.

Thanks!

[busraoguzoglu]

Hello, is there any progress on this? Is it possible to use the tool with other models like LLAMA? I saw that the latest update made this possible but will you add any examples for that?

[GreyDGL]

@busraoguzoglu Yes. Please check the latest commit. We have supported GPT4All models.
Note that during the testing, we notice that all the open-source models are not as good as GPT4.

[SATUNIX]

@busraoguzoglu Yes. Please check the latest commit. We have supported GPT4All models. Note that during the testing, we notice that all the open-source models are not as good as GPT4.

Good point, its best to stick with GPT4API For now until specific models can be trained I've ran some testing using Vicuna and LLaMa2 and the results were negligible..

[GreyDGL]

@SATUNIX fine-tuned Vicuna-13b looks fine, but other than that, all the open-source models below 13B are kinda not working at all.

[GreyDGL]

I'll merge this issue to #69 and implement further features.