A Web-based SDR client

Fullstack TypeScript web SDR receiver. Bun backend manages rtl_tcp and does the DSP (FFT spectrum + demodulation); React/Tailwind/shadcn frontend renders the spectrum/waterfall and plays demodulated audio.

Prerequisites

• Bun ≥ 1.3
• librtlsdr (provides rtl_tcp): brew install librtlsdr
• An RTL-SDR Blog V3 dongle plugged in (rtl_test -t should list it)

Optional (features disable themselves when missing):

• rtl_433 for ISM-band sensor decode: brew install rtl_433

• whisper.cpp for live speech-to-text of the tuned station: brew install whisper-cpp, then download a ggml model into ~/.cache/whisper.cpp (or models/), e.g.

  curl -L -o ~/.cache/whisper.cpp/ggml-small.en.bin \
    https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.en.bin

  Pick between downloaded models in the transcript panel (the largest is the default; WHISPER_MODEL=/path/to.bin overrides, WHISPER_LANG=auto for non-English stations). Dropping a Silero VAD model (ggml-silero-*.bin) alongside it makes whisper skip music instead of hallucinating lyrics over it.

Run

bun install
bun run dev        # starts server (:8787) and web (:5173) together

Open http://localhost:5173.

Layout

• packages/shared — WS protocol types, modes, constants, rtl_tcp command codes
• packages/server — Bun: rtl_tcp manager + TCP client, DSP pipeline, WebSocket server
• packages/web — Vite + React + Tailwind + shadcn UI

See packages/server/README.md for the DSP pipeline and the rtl_tcp protocol notes.