What it is
AI Desk Meter is a visibility layer for local AI work and the open foundation for the planned MuseMeter 3.0 commercial package. It reports provider state, confidence, backend warnings, and display-safe usage indicators without scraping private dashboards or pretending estimates are exact.
Python provider layer for mock, manual, and Arc-RAR state-file payloads.
Wi-Fi/stdout transport today, with BLE and serial companion bridges planned.
Arc-RAR first, then Omnibinary event spine and Neural Synth visualization once backend contracts are stable.
Open corridor: v0.x through the stable pre-3.0 foundation remain open source. MuseMeter 3.0: planned commercial full package with second-brain / Neural Synth / AI buddy features.
Backend-first architecture
Provider source → validated payload → transport → dashboard/device endpoint
The dashboard is not the authority. It renders state from providers and fails closed when data is missing or invalid.
DIY / Cost Specs
The full printable hardware spec is bundled at DIY_Claude_Code_Desk_Usage_Meter_Spec_Guide.html. The build-ready buying sheet is Parts & Sourcing. The runtime dashboard also includes a DIY / Cost Specs panel so the bill of materials is visible inside the app shell.
- Recommended board: Waveshare ESP32-S3-Touch-AMOLED-2.16.
- Search term:
ESP32-S3-Touch-AMOLED-2.16. - Where to start: Waveshare official store/docs, Waveshare GitHub sample, Espressif Component Registry, or Amazon/regional marketplace for convenience.
- Required extras: data-capable USB-C cable and a desktop/Pi runtime host.
| Tier | Estimated budget | Notes |
|---|---|---|
| Cheap Prototype | $25–$60 | ESP32-S3 board plus small LCD/OLED for protocol/UI validation. |
| Best Match | $50–$120 all-in | Waveshare ESP32-S3-Touch-AMOLED-2.16, USB-C power, case, and stand. |
| Product-style build | $90–$160 one-off | AMOLED board, enclosure, OTA/pairing polish, and sellable desk-meter finish. |
Hardware targets
| Device | Role |
|---|---|
| Desktop/laptop | Full host daemon and dashboard |
| Raspberry Pi / Linux SBC | Full host daemon, kiosk, local bridge |
| ESP32-S3 | Wi-Fi/BLE display endpoint |
| Arduino-class MCU | Companion telemetry or simplified display node |
Arc-RAR integration path
The first bridge reads a local Arc-RAR state JSON file. The live bridge consumes stable Arc-RAR CLI JSON output: status, receipts latest, archive verify, and session inspect.
No-server default mode
AI Desk Meter does not require a local server. Generate a payload file with ai-meter write-status --provider mock --out runtime/status.json or keep it updated with ai-meter watch --provider mock --out runtime/status.json --interval 0.5, then import it below.
No local file imported yet.
Optional local API preview
The local API is optional development/debug tooling. Run ai-meter serve --host 127.0.0.1 --port 8787 only when you want browser polling. If the API is not running, the page stays in safe demo/offline mode.
Local API not connected yet.
Raspberry Pi / Linux SBC ready path
v0.5 adds install, smoke-test, systemd, kiosk, and network-safety docs for Raspberry Pi-class systems. Arduino-class devices remain companion display/telemetry nodes, not backend authority.
bash scripts/install_linux_sbc.sh bash scripts/run_smoke_test.sh ai-meter write-status --provider mock --out runtime/status.json ai-meter watch --provider mock --out runtime/status.json --interval 0.5 # optional preview only: ai-meter serve --host 127.0.0.1 --port 8787
ESP32 / Arduino companion bridge
v0.6 adds a compact companion endpoint for small displays and serial telemetry nodes. The host remains the source of truth; companion devices render state only.
ai-meter companion-status --provider mock ai-meter write-companion --provider mock --out runtime/companion.json ai-meter watch-companion --provider mock --out runtime/companion.json --interval 0.5 # optional preview only: GET /companion/status?provider=mock
The pixel buddy and ✶ Musing... state are preserved as the baseline display identity. For now, Musing means the agent is responding to prompt input or an action is loading.
Open-source corridor and MuseMeter 3.0
The current repo remains the open foundation. The planned MuseMeter 3.0 package is the later commercial-licensed second-brain / Neural Synth / AI buddy product. The baseline character and ✶ Musing... state are intentional and preserved.
Roadmap
- Clean dashboard and mock/manual providers.
- Arc-RAR state-file provider.
- Arc-RAR CLI contract provider.
- Provider contract tests and diagnostics export.
- Raspberry Pi / Linux SBC kiosk validation. ✅
- ESP32-S3 display endpoint hardening. ✅
- Arduino-class companion telemetry bridge. ✅
- Native dashboard shell.
- Omnibinary adapter, ARC-Core hardwire integration, and Neural Synth toggle page.
- MuseMeter 3.0 commercial full package: second-brain / Neural Synth / AI buddy.
Native shell path
v0.8 adds a Tauri-oriented desktop shell plan for wrapping the local API and dashboard while keeping provider truth in the backend.
Open track: AI Desk Meter v0.x-v2.x. Future track: MuseMeter 3.0 commercial second-brain / Neural Synth / AI buddy.
v1.2.0 Stable Runtime App Shell
Buttons, tabs, docs/specs navigation, and layout containers are now stable while runtime/status.json streams. The refresh loop updates values only, preserving the runtime-vs-Muse truth model and reducing flicker.