CellarKeep

Fermentation apps are built for beer or for commercial wineries. Small-batch wine, mead and cider makers get long timelines, staggered nutrient additions, rackings between a handful of shared vessels, and bottles that need years, and they end up in spreadsheets. CellarKeep is an open-source, self-host-first production tracker built around how those batches actually get made.

Everything in a batch's life is one event journal: measurements, additions, rackings, stabilizing, sweetening, bottling, notes. Each event is planned, done or skipped, so the guided schedule and the freeform log are the same feature. The Today view is simply the planned events that are due.

Recipes carry steps anchored to the batch start or to the previous step. A pure schedule engine stamps them onto a new batch and scales the ingredient bill to the actual volume. Vessel occupancy is derived from racking and bottling events rather than stored. Measurement fields live in JSONB typed by per-event zod schemas, shared by server actions and client forms, so a new measurement kind needs no migration.

Single Next.js app, Postgres with Drizzle, Better Auth. A docker-compose file brings up the app and database, migrations run on container start, and the first visit creates the admin account. GitHub Actions runs typecheck, lint, unit, database and Playwright suites.

Late steps reflow the chain. Anchors encode drift intent at authoring time: completing a step late shifts everything anchored to it and leaves start-anchored steps alone. No cascade prompts.

Inventory deducts on confirmed reality. An addition marked done deducts stock through an explicit, editable confirmation, never silently and never on plan. That is the drift bug other tools have.

Editing a recipe never touches in-flight batches. Stamped events belong to the batch.

Calculators live where the numbers are needed. TOSNA, ABV, sulfite dosing by pH, backsweetening and chaptalization are pure, exhaustively tested functions that surface inside the forms that use them.