qalarc.com / projects / rfai-monitor

RFAI Monitor

A project by Qalarc — AI systems studio, Sydney, Australia. Built, shipped and maintained by Qalarc.

Software-defined radio + local AI, v0.8.0: Airspy/HackRF capture, Whisper transcription, intent classification, transmitter fingerprinting, named sessions & editions. No cloud.

Airspy MiniHackRF OneSoapySDRnumpy/scipy DSPfaster-whisper (ROCm)Kokoro-82M TTSOllama (phi4, qwen, gemma)Online clustering (CFO/spectral)FastAPI + SSE/WebSocketSQLite + FTS5Svelte 5 + SvelteKitTauri 2 (Rust shell)systemd
RFAI Monitor screenshot

RFAI is a radio-frequency intelligence platform, all on-device.

An Airspy Mini (with HackRF One for transmit) captures live RF across UHF CB, marine, aviation and amateur bands — all 80 Australian CB channels in one 6 Msps capture; faster-whisper transcribes voice locally with streaming partials; entities and intent are classified with local LLMs; and a hardware fingerprinting engine (carrier frequency offset, spectral envelope, key-up transient) identifies the individual physical radio — field-verified separating two identical radios.

0 adds the commercial foundation: named capture sessions with JSON/CSV export, consumer (listen-only) vs operator editions, SSB demod and digital-signal labelling. 9-page Svelte 5 UI (Dashboard, Log Room, Chat, Scanner, Sessions, Monitor, Analysis, TX ID, About) + Tauri 2 desktop shell; FastAPI backends with SSE/WebSockets; SQLite + FTS5 archive; Comms API with Discord/Slack webhooks.

Research lanes: SPECTRE all-domain SIGINT grid, psylink data-over-audio modem, GPS/APRS, morse/SSTV.

Capabilities

Tags

#ai#airspy#editions#fastapi#fingerprinting#hackrf#kokoro#local#ollama#python#radio#rf

Status: active · First built: 2026-05-04 · Last updated: 2026-08-28