Lune AI

Expert LLMs trained per-topic that outperform top AI models.

150 followers

Expert LLMs trained per-topic that outperform top AI models.

150 followers

•

•

Lune AI is a replacement to standard LLMs for coding tasks - minimize hallucinations with a developer-driven marketplace of expert LLMs trained on hundreds of technical topics that outperform standalone models. Create and contribute to Lunes and get paid.

Free Options

Launch tags:API•Developer Tools•Artificial Intelligence

Launch Team / Built With

Framer AI AgentsDesign and publish professional sites with AI

Promoted

Lune AI

Maker

📌

Hey Product Hunters! I'm Sam, CEO and co-founder of Lune AI and I'm super stoked to introduce Lune AI, a community driven marketplace of expert LLMs for code that I'm building with my brother @josh_park12 Lune AI is a platform for any developer who wants to use accurate LLMs with up to date knowledge on the latest libraries and packages. Individual Lunes are up to 37% more accurate than traditional LLMs. We also offer an OpenAI hot-swappable API that can easily be used inside Cursor, or in any codegen application. Here are two examples of how our product is used today: 1. If you are a developer working with a new Langchain feature, and you want an LLM to implement it for you, you can use the Langchain Lune that another dev created on our platform or directly inside Cursor to get accurate outputs, instead of wasting time working with a hallucinated version that Claude or GPT might give you. 2. If you are building a codegen product using OpenAI's API, you could edit a single line of code to swap your OpenAI chat/completion calls with the Lune API to get more accurate codegen output for your specific use cases. One creative use case we’ve seen so far is training Lunes on component libraries to generate front end code! Lunes started out as an internal tool that gained quick adoption from our friends and coworkers, so we are so excited for the world to try it out. You can also receive payouts as your Lunes gain popularity and by contributing your own knowledge and expertise directly to the outputs of Lunes! Want to start chatting with our public repository of user-created Lunes? Visit our website and create a free account! If there are bugs or you find that some Lunes aren’t performing well, please ping us! We’re working every day to improve the product. https://www.lune.dev

Report

2yr ago

Hi, I was building something similar (RAG on top of an OpenAI-compatible API) for my personal use as I was working with some large libraries! I am very happy to see a YC startup solving this; hallucination is such a pain for coding. Could you elaborate on the technical implementation? How should library authors structure documentation the give best results? How do you manage files—are you chunking them? What is the context window of the underlying model that you are using (I find Gemini Pro with 2M tokens very handy for this)? How do you manage documentation inside a Monorepo (like https://github.com/vuetifyjs/vue...) ? Thanks!

Report

2yr ago