[hellofrequencylab]

Requested by Daniel

What

The CyberUnicorn console can now use a Flipper Zero as its tag writer. It recognizes whether the board on the cable is the Unicorn or the Flipper and switches modes. In Flipper mode there is no screen feed and no boot wait, the Scanner and Update tabs are hidden, and the copy says to hold the tag on the back of the Flipper.

Why

Daniel wants to write placards from the console with the Flipper. The Flipper side is the Frequency app (hellofrequencylab/projects#17). While that app is open, the Flipper exposes a second USB serial port that answers this console's protocol: system.status, system.caps, hat.probe, frequency.write_tag and frequency.read_tag. The console's picker only listed the CH340, so the Flipper never showed up.

How to review (owner)

1. Wait for CI to go green.
2. On the Vercel Preview, in Chrome on the Mac with the Flipper plugged in and Apps → NFC → Frequency open:
   • Go to /admin/cyberunicorn and click Connect USB. The picker lists the Flipper. Pick the port whose name ends in 3. The header should read FLIPPER 0.1.0, with no screen feed, and the Reader should show "Flipper NFC".
   • Choose a Space and Write placard, then hold a blank NTAG on the back of the Flipper. The placard should appear in the list with its UID, and a phone tap should check in.
3. Plug the CyberUnicorn back in and connect. Nothing should change for it: same boot wait, same screen feed, all tabs.

How / files

File	Change
lib/device/web-serial.ts	FLIPPER_VENDOR_ID, usbDeviceOf(port), rememberedConsolePorts() (CyberUnicorn first, then Flipper ports) and pickConsolePort() (CH340 + Flipper filters). The updater keeps pickPort/rememberedPort, which are CH340 only, so a Flipper is never offered for an OS flash. Open errors and the transport label name the Flipper.
components/spaces/loyalty/cyberunicorn-console.tsx	connect() now picks the candidate ports and attach() opens one. A Flipper port gets 3 short system.status asks instead of the 40 s boot wait. Remembered Flipper ports are tried in turn, because the first one is the Flipper CLI and never answers. A device state drives the copy, the FLIPPER header, no mirror, and hiding the Scanner and Update tabs.

Verification

• pnpm exec tsc --noEmit: exit 0. It needs NODE_OPTIONS=--max-old-space-size=8192 locally, because the default heap ran out of memory.
• pnpm lint on the two files (eslint --max-warnings=0): clean.
• pnpm exec vitest run lib/device: 9 files, 80 tests passed. The protocol, relay, cable-scan and screen-mirror suites are unaffected.
• Checked the Vercel preview
• On hardware, the Flipper app answered this exact protocol from the Mac over /dev/cu.usbmodemflip_I5had0ha3: status, caps, hat.probe, unknown_command, bad_args, and a clean "No card" timeout. That is noted in the Flipper PR.

Notes / follow-ups

• A real tag write through the console still needs Daniel at the Flipper.
• tagProblem() keeps its CyberUnicorn wording for busy, no_reader and a plain timeout. The Flipper app never sends the first two. The console demo thread's in-flight PR also edits tagProblem(), so I left that text alone to avoid a conflict.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
frequency-web	Ready	Preview	Oct 21, 2026 5:55pm UTC

[hellofrequencylab]

The test check failed on one test: scripts/backlog-contract.test.ts, "the full --probes=all sweep is green, and stays inside its cost budget". It hit vitest's 90 s timeout. It did not fail an assertion.

On this exact head (59b3808), node scripts/check-backlog.mjs --probes=all exits 0 locally in about 60 s. All 1653 probes agree, including the four cu5k probes this branch touches. So the timeout was the sweep running slow while sharing a loaded CI runner, not a broken probe. I've re-run the failed job once.

---

Generated by Claude Code