# radio-gateway

> Ham radio & GMRS gateway, repeater and packet radio — bridges two-way radios to Mumble, Broadcastify, and the internet. AIOC USB, RSPduo dual SDR, TH-9800/D75/KV4P CAT control, AI announcements, ADS-B

- **URL**: https://www.freshcrate.ai/projects/radio-gateway
- **Author**: ukbodypilot
- **Category**: MCP Servers
- **Latest version**: `v4.0.0` (2026-06-01)
- **License**: Unknown
- **Source**: https://github.com/ukbodypilot/radio-gateway
- **Language**: Python
- **GitHub**: 5 stars, 2 forks
- **Registry**: github
- **Tags**: `adsb`, `aioc`, `broadcastify`, `gmrs`, `ham-radio`, `kv4p`, `linux`, `mcp`, `python`

## Description

Ham radio & GMRS gateway, repeater and packet radio — bridges two-way radios to Mumble, Broadcastify, and the internet. AIOC USB, RSPduo dual SDR, TH-9800/D75/KV4P CAT control, AI announcements, ADS-B tracking, MCP server for AI control, live web dashboard. Raspberry Pi & Linux.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v4.0.0` | 2026-06-01 | High | AllStarLink integration. The gateway can now bridge into the AllStar network as a first-class audio source/sink and connect to any node on demand — without running a radio node on the gateway box. A headless ASL3 "bridge node" (no RF hardware) speaks the USRP (DVSwitch) protocol to a new in-gateway plugin; everything else (Mumble, web player, other radios) routes to/from AllStar like any other bus endpoint. Bringing this up also turned the plugin system from "auto-discovered but only half-wired |
| `v3.8.0` | 2026-05-25 | High | # IC-7100 GUI v2 — Real Radio Replacement  This release promotes the `/ic7100` web panel from "remote-control demo" to a control surface designed for **live operating**. Knobs feel like the physical radio. CI-V latency is in the human-imperceptible range. The squelch LED keeps up with the noise floor. The frequency knob doesn't pile up minute-long backlogs. The AF knob drives the radio's actual speaker volume, not just a software gain.  ## Headline IC-7100 GUI changes  ### Performance - **40× CI |
| `v3.7.0` | 2026-05-18 | High | The transcription system is now multi-machine, and the Fleet Manager finally has the docs it deserves.  ## 🚀 Headline features  ### Distributed transcription pool  VAD runs on the gateway; ASR inference can be local, remote, or routed between both **by clip length**.  - New module `transcribe_engine.py` — `LocalInferenceEngine` (Moonshine or Whisper via faster-whisper, in-process) + `RemoteEngine` (HTTP client with background `/status` polling) + `_pick_worker()` dispatcher - New `tools/transcr |
| `v3.5.0` | 2026-05-03 | High | Two big things this release: a **persistent transcription log with natural-language search**, and a **bus tick refactor** that pushes every sink off the audio hot path. Plus an installer overhaul, several UI redesigns, and a handful of fixes for bugs that surfaced along the way.  ### Added — Transcription log + AI search  - **Persistent transcription log** (`transcription_log.py`). SQLite database with FTS5 full-text search. Every transcript Moonshine produces is stored with its timestamp, sour |
| `v3.4.0` | 2026-04-22 | High | First version aimed at fresh users. You should be able to go from a clean Arch or Debian box to a running gateway using only the repo, without a support session.  ## Installer & onboarding  - **New [INSTALL.md](INSTALL.md)** — full fresh-install walkthrough: prereqs, AUR helper setup, credential acquisition (Mumble / Broadcastify / Telegram / GDrive), pre-flight checklist, expected first-start output, runtime-state map, troubleshooting, manual uninstall, companion-repo pointers. - **`requirement |
| `v3.3.0` | 2026-04-20 | High | ## [3.3.0] -- 2026-04-19  ### Added - **DeepFilterNet 3 denoise engine** — second neural denoiser alongside RNNoise, selectable per bus.   - `_DFN3Stream` in `audio_util.py`: stateful streaming ONNX (16 MB) via existing onnxruntime. No new Python deps, no numpy conflict. ~40 dB cut on white noise. Model vendored in-repo at `tools/models/dfn3/denoiser_model.onnx` — no runtime download.   - Engine abstraction: `DenoiseStream` duck-type, `make_denoise_stream(engine)` factory, shared `_mix_with_dry_ |
| `v3.2.0` | 2026-04-19 | High | ## Highlights  ### Transcription overhaul - **Whisper replaced with Moonshine** (`useful-moonshine-onnx`) — faster, CPU-efficient, no torch, no CUDA. - **Silero v5 VAD** replaces the dBFS envelope. Threshold is now a probability 0.0–1.0; speech detection is actually speech detection, not loudness. - **Polyphase 48k → 16k resample** (scipy) replaces bare decimation — no more aliasing artifacts feeding the ASR. - **RNNoise neural denoise** (`pyrnnoise`), available two ways:   - Per-bus "D" filter |
| `v3.1.0` | 2026-04-09 | High | ## Highlights  - **SDR single-tuner mode** — RSPduo runs one tuner with multi-channel demodulation. CPU reduced 57% (31% → 13%). Mode selector on `/sdr` page. - **SDR1/SDR2 as separate routing nodes** — each channel independently routable to any bus - **D75 legacy cleanup** — 1,136 lines of dead code removed; D75 is now link-endpoint-only - **`audio_util.py`** — shared level metering module replaces 55 duplicate inline patterns - **Web UI shared code** — `common.js` with `postJson`, `createPolle |
| `v3.0` | 2026-04-08 | Medium | ## What's New in 3.0  ### Unified Audio Engine All buses (listen, solo, duplex, simplex) now run in a single BusManager thread. The main loop is simplified to SDR rebroadcast TX and WebSocket push. One code path for all audio — cleaner, more maintainable, fewer bugs.  ### Loop Recorder Per-bus continuous recording with visual waveform review. Enable the **R** button on any bus in the routing UI to start recording.  - Canvas-based waveform viewer with zoom, pan, click-to-play - Right-click drag t |
| `v2.0.0` | 2026-04-01 | Medium | # Radio Gateway v2.0.0  **The biggest release yet.** A ground-up rewrite of the audio architecture — from a monolithic mixer to a visual, bus-based routing system. 119 commits, 28 bugs fixed, 7,200+ lines of dead code removed.  ![Routing UI](https://raw.githubusercontent.com/ukbodypilot/radio-gateway/main/docs/screenshots/routing.png)  ## Highlights  ### Visual Audio Routing - **Drag-and-drop node editor** — wire sources, busses, and sinks visually - **4 bus types:** Listen (monitor mix), Solo ( |

## Citation

- HTML: https://www.freshcrate.ai/projects/radio-gateway
- Markdown: https://www.freshcrate.ai/projects/radio-gateway.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/radio-gateway/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._