Inspiration
I dont know how to code, and i heard of someone winning a hackathon with a no-code project using replit, so i tried to take the same route
What it does
its very simply a pre-trained model by google, with 55k tokens worth of text to keep the conversation on topic and laser-focused on teaching you NMAP
How we built it
combination of replit.com, and google gemini. Google Ai studio was what i landed on as the final demo. i converted a 200page pdf document that is a crash course on NMAP, into plaintext, which i fed into gemini as a document, and saved the chat. i can now share the saved chat and users can use nmapAssist in more narrow context
Challenges we ran into
i couldnt figure out how to "prime" the chat well. training a new llm from scratch would be inefficient and retraining only a small part of the model requires more technical knowledge than i have, so i settled on finding the llm with the highest parameter input so i can input my context along with the first chat of the user without the user needing to do anything on their end. if i were to try this again i would try to get an openai key instead of a google api key because i think this would be smoother to integrate on the GPT store.
Accomplishments that we're proud of
finishing and getting a working mvp (i think)
What we learned
ai is hard. ai is fun.
What's next for nmapAssist
create an ai application for kali linux. give an ai agent (auto-gpt) access to the command-line and root access. ask it to learn the documentation on all the tools. Accelerated learning begins
Log in or sign up for Devpost to join the conversation.