hark — pipe your Mac's audio (meetings, podcasts, mic) straight into Fabric

[PhantomYdn]

Fabric turns text into insight — but a lot of the most valuable text is locked in audio: meetings, calls, podcasts, voice notes. I built hark (https://github.com/PhantomYdn/hark) to be the front half of that pipe on macOS: it captures your mic, system audio, or a specific app and transcribes it on-device, then writes the transcript to stdout. So it drops straight into your favorite patterns:

# Summarize a live meeting (system audio + your mic) when you stop it
hark --system --mix -t - | fabric-ai -p summarize_meeting

# Mine a podcast or interview file for the good stuff
hark -i interview.mp3 -t - | fabric-ai -p extract_wisdom

# Think out loud, get structured notes
hark --engine apple -t - | fabric-ai -p extract_ideas

Why it might be interesting for you:

• Unix to the core — one input → stdout, composes with fabric, ffmpeg, jq, your shell. hark is the verb.
• On-device & private — no network calls by default; engines are local (whisper.cpp, Apple Speech, WhisperKit, Parakeet).
• One native binary — pure Swift + Core Audio, no BlackHole/virtual devices, brew install and go.
• Bonus: a remote-control agent + a Tampermonkey userscript that auto-records Google Meet → transcript → straight into a Fabric pattern.

Heads-up: it's macOS only (14.4+; prebuilt binary is Apple Silicon, Intel builds from source). MIT licensed. It's pre-1.0 but the capture/transcribe/pipe path is solid (tapped brew install).

Repo + examples: https://github.com/PhantomYdn/hark — would love feedback from Fabric folks!