feat: add FAQs for nitro documentation

[hiro-v]

@vuonghoainam on tutorials for ‘npm-openai’ and ‘python-openai’ integration code with nitro

[tikikun]

• Hey, isn’t Nitro just Llama.cpp with an api server
• Hey, isn’t Nitro just llama-cpp-python without Python bindings
• Why should I use nitro instead of llama-cpp-python

[tikikun]

• Hey, ollama offer same thing, why switching to nitro

[tikikun]

Hey, isn’t Nitro just Llama.cpp with an api serve

• Yes it is, not just llama.cpp nitro will soon come with multiple other models (whisper, bark, stable diffusion) everything all in one single binary. There is no need to develop your own api server now on top of ai models.

  Hey, isn’t Nitro just llama-cpp-python without Python bindings? Why should i use nitro instead of llama cpp python

  • Yes it is, because nitro is not binded to Python you can leverage very performant software that leverage your system to the fullest. You do not need to learn fast api or how to deploy a python webserver, because nitro webserver is fully optimized by default (lead docs to how nitro is implemented on drogon webserver)

  Hey, ollama offer same thing, why switching to nitro

  • Ollama design was meant for other purpose it has bigger size (~200mb) compared to 3mb distribution of nitro, you can embed nitro into a subprocess and never need to worry too much about package size for your app.

[tikikun]

• Why model name is "chat-gpt-3.5"?

Many applications implement openai chatgpt API, we want you to be able to use nitro on any AI clients that you want therefore the openai adapters have to be 100% compatible, you can put any model name there, but in case you already implement chatgpt api, we make sure that you can just replace api.openai.com with localhost:3928 and your chat clients will just work (chatbox, sillytavern, oobaboga,etc... )