100 points · 1 submission
with turbopuffer
Most code stays silent. But what if you could talk to it? That is the idea behind ElevenNotch: a lightweight app that lives right in your MacBook’s notch and gives you an AI coding partner in seconds. It understands your codebase, helps you brainstorm new features, think through refactors, and untangle confusing parts of a project. Working with code snippets is often tedious and error-prone. Copying and pasting rarely captures the full context, and important dependencies are easy to miss. ElevenNotch removes that friction. Just ask about any part of your code, and the assistant responds with clear, concise answers grounded in the bigger picture of your project. ElevenNotch is powered by two core technologies working in tandem. Your codebase is indexed using Mistral's Codestral Embed model — a state-of-the-art embedding model — and stored in TurboPuffer, a serverless vector database. When you ask a question, TurboPuffer performs a semantic vector search across your entire project in milliseconds, retrieving the most relevant code chunks regardless of how you phrased the question. This means asking "how does this part handle errors?" finds the right code even if the word "error" never appears in it. Retrieved chunks are then passed as context to a language model, which synthesizes a focused, grounded answer — one that reflects the actual shape of your project rather than generic assumptions about how the code might work. The answer is then delivered entirely through ElevenLabs' Text-to-Speech API, so the response comes back as natural, spoken audio rather than text on a screen. The result is a truly conversational experience: you speak a question, your codebase is searched, and a voice answers you back — without ever leaving your flow state or touching your keyboard. Instead of wrestling with isolated snippets, you get a fast, conversational way to explore and understand your codebase.
Submitted 15 Apr 2026