Crema

A native client for the Wendougee LITA-BA / LITA-BR / DATA-S espresso machines, built for coffee nerds who want to author, share, and run pressure / flow / weight profiles.

Driven by the reverse-engineered protocol in LitaLite.

Status: iOS / iPadOS / macOS in active development. Live BLE connection, real-time pressure / flow / volume charting, multi-stage profile authoring, auto-tune for new beans, and community profile sharing with cross-platform Sign-In-with-Apple + Google. Android port planned (the backend already speaks Google identity).

Features

The hero — live brew visualization

• Real-time chart of pressure, flow, and pumped volume against the target profile (ghost setpoint lines for each stage).
• Stage labels with live setpoints float over the curve at full size (iPad / macOS).
• Metric HUD with elapsed time, pressure, flow, and volume.
• Demo mode: replay a captured shot against its real-shape profile when no machine's connected.

Profile authoring

• Visual node editor — drag node-per-stage to reshape duration + setpoint live. Each stage can be pressure- or flow-priority.
• List editor — classic form view with collapsible per-stage cards, reorder, duplicate, delete.
• Multi-stage (more than 3, with pauses / waits between stages).
• Optional grinder settings saved on the profile (size + RPM + single-dose).

Auto-tune — the USP

A 7-step wizard that dials in a new bean from scratch in ≤3 guided shots:

1. Bean — name, roast, drink type (espresso vs milk).
2. Recipe — dose, target yield, target brew time, ratio.
3. Prep — recommended grind + dose + checklist. "Set Grinder" pushes the recipe's grind size to the paired grinder via FF55.
4. Brewing — live chart + "Stop when cup hits Xg" CTA (the machine's volume reading is pump volume, pre-puck — it's not yield).
5. Measure — confirm cup weight (manual input or live BLE scale via ScaleTransport protocol).
6. Suggest — verdict + recommendation card for the next shot (finer / coarser grind, pressure delta), with shot-grade reveal.
7. Complete — save the best-scoring shot's profile, trimmed to the user's target brew time, to the library.

Community sharing

• Sign in with Apple or Google (PKCE OAuth via ASWebAuthenticationSession — no SDK dependency).
• Upload any profile to the community backend, get a shareable URL + QR code.
• crema://profile/<id> deep links and https://<host>/p/<id> Universal Links open the app and show a preview before adding to library.
• .crema files (AirDrop, Messages, Files app) decode locally — no network needed.
• Browse the community feed, like, follow other coffee nerds.

Shot history + feedback loop

• Every brew can be saved with a tasting note (sour / balanced / bitter) and yield.
• ProfileTweakSuggester rule engine surfaces targeted adjustments from the last shot's outcome.

Repo layout

Crema/
├── Sources/
│   ├── CremaKit/                Engine — pure Swift, no UI deps. Cross-platform.
│   │   ├── Modbus.swift, FF55.swift, BLEMachineTransport.swift, …
│   │   ├── BrewProfile.swift    Domain + on-the-wire encoder
│   │   ├── ProfileLibrary.swift Observable library + Codable persistence
│   │   ├── AutoTune/            Pure-functional recommender + types + ScaleTransport
│   │   └── Sharing/             ProfileShareCodec, ShareableProfileURL, ShareAPIClient
│   └── CremaApp/                SwiftUI apps (iOS + iPadOS + macOS share 95%+)
│       ├── Views/               Live chart, profile editors, settings, sharing UI
│       ├── Views/AutoTune/      Wizard sheet + step views + confetti
│       ├── Views/Sharing/       Sign-in, share-out, browse, import, settings
│       └── Data/                Stores + drivers (LiveDriver, AutoTuneSession, etc.)
├── Tests/CremaKitTests/         71 tests across 14 suites
├── backend/                     Hono on Node 22 + Drizzle + Postgres (self-hostable)
│   ├── src/                     Routes, schema, auth (Apple + Google verify)
│   ├── db-init/                 Initial schema SQL (Postgres-init hook)
│   ├── Dockerfile, docker-compose.yml
│   └── README.md                Deploy + env vars
├── project.yml                  xcodegen multi-platform project
└── Crema-iOS.entitlements / Crema-macOS.entitlements

Running

macOS

xcodegen generate
xcodebuild -project Crema.xcodeproj -scheme Crema-macOS -configuration Release \
  -destination 'platform=macOS' -allowProvisioningUpdates build
cp -R "$(xcodebuild -showBuildSettings -project Crema.xcodeproj -scheme Crema-macOS \
  -configuration Release 2>/dev/null | awk '/ CONFIGURATION_BUILD_DIR /{print $3}')/Crema.app" /Applications/
open /Applications/Crema.app

(Or just swift run for a console build; you'll lose the system bundle ID + permission prompt nicety.)

iOS / iPadOS

xcodegen generate
open Crema.xcodeproj
# Pick a device or simulator, ⌘R

For TestFlight: build the App Store archive (xcodebuild ... archive), export with App Store Connect method, upload with xcrun altool --upload-app. There's a working pipeline documented in the recent commits.

Backend

cd backend
docker compose up --build
# API on :8080, Postgres on :5432, schema applied on first boot

See backend/README.md for env vars (DATABASE_URL, JWT_SECRET, APPLE_CLIENT_ID, GOOGLE_CLIENT_IDS, Universal Links discovery, …).

The iOS app points at http://localhost:8080 by default; change via Settings → Backend → Base URL in-app, or CREMA_BACKEND_URL env at launch.

Tests

swift test

71 tests across 14 suites — Modbus framing byte-exact vs captured fixtures, FF55 checksum, BrewProfile encoder, machine registry, ShotHistory + TipPreferences, AutoTuneRecommender, profile sharing codec + URL parsing, stub transport lifecycle.

Roadmap

• Android client (Kotlin / Jetpack Compose) — shares the backend; Google sign-in already wired.
• BLE scale integration — ScaleTransport protocol is defined; Bookoo Themis and Acaia drivers next.
• Public TestFlight + App Store submission.
• Community v2 — comments, ratings, equipment filters, account merging (Apple ⇄ Google by email).

License

To be decided — currently all rights reserved.