On this episode, we had absolutely the pleasure of sitting down with Carmine Paolino — an AI innovator, Ruby fanatic, and all-around tech wizard. From his early days automating PC video games at age 5 to constructing cutting-edge AI instruments in Berlin, Carmine’s journey is as inspiring as it’s spectacular.
We dove deep into his newest creation: RubyLLM, a Ruby gem that simplifies working with giant language fashions (LLMs) like GPT-4, Claude, and Gemini. Consider it as an intuitive, plug-and-play toolkit that lets Ruby builders faucet into highly effective AI options — chat, picture technology, embedding, instruments, and even multi-model help — with just some traces of code. And sure, it’s as superior because it sounds.
Key Takeaways:
- RubyLLM is constructed for simplicity and energy. Carmine wished a instrument that “simply works” — one unified interface for chatting, streaming, instrument use, picture technology, and extra. It abstracts away the API mess and retains issues Ruby-friendly.
- Tooling help is next-level. RubyLLM permits for agentic AI by letting devs outline instruments (like checking the climate or sending a calendar invite). The gem handles when and the way to use them — magic!
- Help for a number of fashions and suppliers. OpenAI, Anthropic, Google — RubyLLM makes it straightforward to change between them seamlessly, even mid-conversation. Carmine additionally teased a future integration with a better mannequin registry through an AI-powered API referred to as Parsera.
- Streaming and efficiency? Lined. Carmine shares intelligent structure tips utilizing Turbo Streams and async Ruby for blazing-fast, light-weight responses — even when dealing with many concurrent customers.
- Actual-world use case: ChatWithWork. Carmine’s app lets customers “chat” with their docs from Google Drive, Notion, and Slack. RubyLLM is the spine, and it’s obtained actual startup traction. (Oh, and he DJed the night time it went viral on Hacker Information.)
- Embeddings and picture technology are simply as straightforward. Want vector search or auto-generated podcast artwork? Simply name .embed or .paint — severely, that’s it.
