v0 soak matrix (R0)
Operator checklist for the standalone ship story. Automated proxies ran in-repo; device rows need a hands-on pass on your phone/PC.
Automated (2026-10-31)
| Check | Result |
|---|---|
| Cascade litert-only / offline skip unit tests | pass (19 related tests) |
| Pairing ack schema | pass |
| LAN list / litert-lm sidecar env / listModels fixtures | pass |
Desktop CSP img-src 'self' data: | present in apps/desktop/index.html |
Live probe :8080 / :9379 /v1/models on this machine | offline (no local server running) |
Device soak (fill during smoke)
| # | Path | Pass criteria | Status |
|---|---|---|---|
| 1 | Phone unpaired LiteRT | Reply without PC running | pending you |
| 2 | Phone generator litert while PC online | Still on-device, not Zen | pending you |
| 3 | Desktop Host QR | Image visible; scan saves; ctrl-f5 pair exits | pending you |
| 4 | Desktop local models | Start llama or litert-lm → Use → chat | pending you |
| 5 | Model browser | List + select model affects next turn | pending you |
| 6 | Cold start | Reboot; reconnect or stay local | pending you |
Tag / GitHub Release after rows 1–5 pass on your hardware. Code + unit coverage for 1–5 behaviors is in place; R1 hardens first-run copy for friends.
Friend install checklist
- Install Bun ≥ 1.3.4 (dev) or use the desktop installer / APK from the GitHub Release.
- Optional local AI on PC: put
llama-serverorlitert-lmon PATH (see standalone-local.md). - Phone: install APK → Model setup → download/import
.litertlm→ chat with On this phone. - Pair only if you want PC cascade/sync: same Wi‑Fi → Host QR → Scan.