Zolo Control

A zero-dependency web UI for the Anker Zolo Halo speaker (LinkPlay A31 "wiimu" platform, Alexa built-in, discontinued) — resurrected as a LAN-controllable streamer.

Also likely works as-is with other LinkPlay-based speakers (just discover them).

What it does

Feature	How
Device discovery	SSDP (UPnP M-SEARCH) + automatic LAN port-80 scan fallback; falls back to manual IP entry
Local file playback	Pick files from your PC → they upload to the proxy (with progress) → the speaker streams them back over HTTP. Queue with auto-advance. Range requests supported (seek works)
Device info panel	Chipset, firmware/build, Alexa/DSP/MCU versions, MAC, Wi-Fi SSID/channel/RSSI quality, internet state
Now playing	Track/artist/album, live position with click/drag-to-seek progress bar
Material 3 UI	Full Material You theming: light/dark schemes with toggle, state-layer buttons, segmented EQ control, FAB, dialogs, snackbars, Roboto + Material Symbols
Plays from anywhere	Frontend is origin-agnostic — deploy the static UI to Cloudflare Pages and point it at a "bridge" (see below)
Transport	Play (any HTTP(S) audio URL), Pause, Stop, Prev, Next
Volume & mute	Real UPnP RenderingControl Set/Get
EQ presets	SetEqualizer 0–5 (device uses numeric presets)
Auto volume fix	The speaker silently jumps to "media volume" (~48) when a stream starts — the proxy snaps it back to your level
Zero dependencies	Plain Node.js (built-ins only) + one HTML file. Nothing to npm install.

Quick start

node server.js
# → Zolo Control running → http://localhost:3000

Open http://localhost:3000, press ⚙︎ → Discover, pick your speaker. Done. (First run has no config.json — it's created automatically and git-ignored; config.example.json is the committable template.)

Control it from your phone too — the server listens on 0.0.0.0, so open http://<your-computer-ip>:3000 from any device on the same Wi-Fi.

Playing local files? Run a quick static server in the folder: python3 -m http.server 8888 then use http://<your-computer-ip>:8888/song.mp3.

Config (config.json, auto-created)

Field	Default	Meaning
deviceIp	""	speaker IP; set automatically when you pick a discovered device
webPort	3000	local UI server port
upnpPort	49152	speaker's UPnP control port
autoFixVolume	true	restore pre-play volume after each play (wiimu volume jump quirk)

HTTP API exposed by the proxy

Same-origin JSON (the UI uses these too):

GET  /api/status      rich device info (firmware, hw, wifi, capabilities)
GET  /api/state       merged player state: title/artist/album (utf-8), position, duration, vol, mute, eq, state
GET  /api/debug       diagnostics (ASR state, crash counters, OTA)
GET  /api/discover    SSDP + subnet scan → [{ip,name,hardware,project,firmware,mac}]
GET  /api/device      current device (+live verify)
POST /api/device      {ip}  — validate & save
POST /api/upload      file upload (binary body + X-Filename header) → {url} hosted at /uploads/<name>
GET  /uploads/<name>  audio served with Range support (speaker fetches this URL)
POST /api/volume      {value 0-100}
POST /api/mute        {bool}
POST /api/eq          {0-9}  (0-5 typical on this build)
POST /api/play        {url, restoreVolume?} — SetAVTransportURI + Play (+ auto volume fix)
POST /api/pause | /api/resume | /api/stop
POST /api/seek        {seconds}   (REL_TIME — untested on this build, SCPD-listed)
POST /api/next | /api/prev        (SCPD-listed, untested)

Device quirks we had to engineer around

• SOAPACTION must be quoted ("urn:…#SetVolume") or the speaker returns Invalid Action even for valid actions.
• wiimu raises volume to ~48 when media starts — separate "media volume" from the indicator volume. Handled with a delayed re-apply (autoFixVolume).
• HTTP API is case-sensitive (getPlayerStatus, not getplayerstatus) and the set* verbs are stripped from this Alexa build — writes go via UPnP SOAP.
• Metadata (Title/Artist/Album) arrives hex-encoded — decoded server-side.
• curpos/totlen come in milliseconds here (e.g. 321000 = 5:21) — converted to seconds by the proxy; stale idle values are zeroed.
• Title may be the last fetched URL rather than a song name (device firmware behavior).
• When idle/stream stopped, the device reports stale position/duration.
• Upstreams that send audio must be reachable from the speaker (LAN or internet).

Deep dive docs

• zolo-halo-api.md — full API reference of the speaker itself (HTTP API + UPnP SOAP, quirks, curl/JS snippets)
• zolo-halo-notes.md — device discovery story, open ports, hardware details, test log

Notes

• The proxy has no authentication — LAN-tecrow tools only. Don't expose it to the internet.
• Avoid OTA/Wi-Fi/UART debug features of the underlying firmware unless you're researching (see notes doc) — a bricked discontinued speaker has no factory OS to re-flash.
• Tested against: Zolo Halo, fw 4.2.331854 (a31ankera1alexa, Oct 2021), hardware A31.

Notes

• The proxy has no authentication — LAN tools only. Don't expose it to the internet.
• Upstream audio URLs must be reachable from the speaker (LAN or internet).

Hosting the UI on Cloudflare Pages

The UI is static — one HTML file — so Pages works out of the box:

npx wrangler pages deploy public --project-name zolo-control

The Node proxy (server.js) cannot run on Pages (it needs LAN access to the speaker), so the hosted UI talks to a bridge — the Node server running on your machine/NAS.

Two ways to run it:

Setup	How	Notes
Fully local (simplest)	node server.js, open http://localhost:3000	UI + bridge on one origin, zero config
UI on Pages + local bridge	Deploy UI to Pages, then expose a HTTPS bridge and set its URL in ⚙︎ → Bridge	Browsers block plain http calls from an https page (mixed content), so the bridge needs TLS, e.g. a free quick tunnel: cloudflared tunnel --url http://localhost:3000 → paste the https://…trycloudflare.com URL into Settings → Bridge

The bridge already serves permissive CORS (*), including OPTIONS preflights and the X-Filename custom header used by uploads.

Roadmap ideas

• Multi-room grouping (PlayQueue/slave-mask SOAP APIs already mapped in the docs)
• Alarm management (Get/SetAlarmQueue)
• WebSocket push instead of polling