hey HN, I built OpenGranola — a macOS app that sits next to your calls, transcribes both sides of the conversation locally, and surfaces talking points from your own notes in real time.
The idea came from having too many calls where I knew I had the perfect data point or quote somewhere in my notes, but couldn't find it fast enough. I wanted something that would do the retrieval for me, while the conversation is still happening.
How it works:
- Point it at a folder of markdown/text files (meeting prep, research, customer briefs, whatever) - Start a call and hit "Start" - It transcribes both speakers on-device using Parakeet TDT (no audio leaves your Mac) - When the conversation hits a decision point or question, it searches your knowledge base and suggests relevant talking points
The whole thing can run 100% locally — pair it with Ollama/llama.cpp for LLM suggestions and embeddings, and nothing touches the network. Or use OpenRouter + Voyage AI if you prefer cloud models.
A few things I'm happy with:
- The app window is invisible to screen sharing by default, so the other side never sees it - Transcription is fully offline via Apple Silicon neural engine (~600MB one-time model download) - Sessions auto-save as plain text transcripts - MIT licensed, Swift/SwiftUI, ~3k LOC
Tech stack: Swift 6.2, SwiftUI, FluidAudio (on-device ASR), RAG over local files with vector embeddings.
Requires Apple Silicon, macOS 26+. DMG on the releases page or build from source with one script.
Would love feedback — especially on the suggestion relevance and what kind of knowledge base content works best for you.
