rNet is just a dev tool that helps developers build AI apps, while for users, it works as a credit wallet.
I kept seeing the same problem in AI apps. The developer pays for all the AI usage and then tries to get the money back with subscription model or the user has to paste their own API key, which is hard for people who are not technical + different key for every AI provider.
ChatGPT already fixed this for its own tools(Codex). if you have a ChatGPT plan, you can sign in to codex extension uses that same plan. No New key needed just sign in with same account. but this only works inside ChatGPT's own tools, not for developer's AI apps.
So i built the same idea, but open for any apps(Developer apps) to use. every user as one AI credit wallet, where they add credit , then signs in to any app that uses rNet and spends credit on app. The developer just registers their app. No billing system to build , no usage tracking. the users pay only one time because AI credit are shared like ChatGPT's plan.
Note : rNet does not read, store, modify or add fallback providers to developer's requests. They are passed through as-is to the AI provider you're(Developer) using. Formatting the request correctly for that provider is the developer's responsibility.
One more detail on how the credit system works under the hood. AI token pricing is usually a tiny fraction of a dollar per token, so tracking use four units: dollar, millidollar, microdollar, and nanodollar. Each one is a thousand times smaller than the one before it, same as milli, micro, and nano work everywhere else.
A millidollar is what shows up as "credit" in the wallet. A nanodollar is used behind the scenes to calculate the exact cost of each AI call.
For example, if a model costs $0.125 per 1 million tokens, then one token costs 125 nanodollars.
The goal is for developers to build AI apps freely, without worrying about AI tokens and to make buying and using AI credits as easy, efficient, and low-waste as possible for users.
And of course the client libraries are MIT licensed and support Java, Node.js and Python, developers can easily add them to their projects and extend them.
Nice. Do you also provide an API or middleware version of TokenAdvisor? Would be great to plug it in before LLM calls so prompts get optimized automatically in production.
They prefer AI music because it’s cheaper and more profitable than paying real artists. The concern is that real musicians may get pushed aside instead of being fairly paid.