evinjohnn/natively-cluely-ai-assistant

Natively hit 650 stars and 9 releases in 30 days by solving the one problem every AI meeting tool ignores: your audio never leaves your machine. While Cluely charges a premium to transcribe your meetings on their servers, Natively runs the entire pipeline locally — real-time transcription, AI-generated answers, and a searchable knowledge base of every conversation you've ever had, all stored in a local SQLite vector database. Here's what makes developers lose their minds the first time they try it: you're in a technical interview, someone asks about your experience with distributed systems, and Natively silently pulls context from three previous meetings where you discussed exactly that — then surfaces talking points in real-time. No cloud round-trip. No API latency. Just instant, context-aware answers from your own meeting history. The architecture is surprisingly clean for a desktop AI app. TypeScript and React handle the UI with a glassmorphism design that's genuinely pleasant. Rust powers the native audio capture — dual-channel, so it grabs system audio and your microphone separately. The BYOK (Bring Your Own Keys) model means you plug in whatever LLM you want: Gemini, GPT-5.x, Claude, Groq, or go fully offline with Ollama. Same flexibility for speech-to-text — eight providers including Soniox for ultra-fast streaming accuracy. The killer feature nobody talks about is the screen-share invisibility. Natively is undetectable during video calls and recordings. Your interviewer, client, or manager never sees it. Combined with the premium profile system — where you feed it your resume, job description, and company research — it becomes a silent copilot that knows your background better than you do. Requires Node.js v20+, a Rust toolchain for native audio, and API keys for your chosen providers. The npm install flow is straightforward, but configuring speech providers adds intermediate-level complexity. Works on macOS 12+ and Windows 10/11.