I've created a NSFW chatbot using Mistral Le Platform and it was surprisingly easy.
The service is free to use and doesn't require any login. This however poses a big challenge:
How do you secure it against somebody using your LLM API credits?
I've built several layer of protection, including simple ones like IP rate-limiting, server side forced system prompt injections and more advanced ones like fully obfuscated and encrypted requests with a secret way of transferring access tokens. Curious to see how it will develop and if someone here could crack it.
I hope I don't break the rules of this community by posting this, but the technical challenge of securing the system was cool!
The service isn't my type of thing, but maybe you could post somewhere in detail about the technical challenges of securing this system. I'm sure that the some people who read HN would find that interesting.
The service is free to use and doesn't require any login. This however poses a big challenge: How do you secure it against somebody using your LLM API credits?
I've built several layer of protection, including simple ones like IP rate-limiting, server side forced system prompt injections and more advanced ones like fully obfuscated and encrypted requests with a secret way of transferring access tokens. Curious to see how it will develop and if someone here could crack it.
I hope I don't break the rules of this community by posting this, but the technical challenge of securing the system was cool!