Hello, may I ask if you need to add a model to the project yourself when using this project? How can I add it if needed?

[isamu-isozaki]

@zhouw1015 sorry, I was away for a bit. Currently, the implementation assumes you use a llama 3 70b on an api server. I'll add back support for local models tomorrow. But, as far as I tested, I don't think you can beat hackthebox boxes with purely smaller models like whiterabbitneo

[isamu-isozaki]

Now it should work with

python main.py --ip_address="whatever_ip_address_you_want" --llm_path=path_to_your model

[zhouw1015]

Are there any plans to implement web access?

现在它应该可以

python main.py --ip_address="whatever_ip_address_you_want" --llm_path=path_to_your model

Are there any plans to implement web access?

[zhouw1015]

@zhouw1015抱歉，我离开了一会儿。目前，实现假设您在api 提交使用 llama 3 70b。明天我会重新添加对本地模型的支持。 但是，据我所知，我认为您无法使用纯粹的小型企业模型（如whiterabbitneo）击败hackthebox盒子

Can the llama 3 8b successfully breach a basic cybersecurity testing platform?

[isamu-isozaki]

Are there any plans to implement web access?

现在它应该可以

python main.py --ip_address="whatever_ip_address_you_want" --llm_path=path_to_your model

Are there any plans to implement web access?

@zhouw1015 ah we are currently adding rag/elastic search over some scraped websites if that's what u mean! For automatic command execution not currently since I don't think it's practical yet.

For llama 3 8b, I think it depends on how basic. For an easy htb box I seriously doubt it. For the input parsing module(summarization) or generative module(explaining tasks) I think it can do but for updating the todo list on what's the next task etc, I've only seen very good models succeed in it like llama 3 70b since it requires a bit of reasoning