TREK/Dockerfile

# ── Stage 0: gosu ────────────────────────────────────────────────────────────
# Rebuild gosu with a current Go toolchain so the runtime image ships no stale
# Go stdlib (Debian's apt gosu is built with an old Go that trips CVE scanners).
# The binary and its runtime behaviour are identical to the apt package.
FROM golang:1.26-alpine AS gosu-build
# Pinned to the commit of gosu's 1.19 tag: @latest made every build pick up whatever
# was pushed last, and the tag itself is not a semver version go install accepts.
RUN CGO_ENABLED=0 GOBIN=/out go install github.com/tianon/gosu@6456aaa0f3c854d199d0f037f068eb97515b7513

# ── Stage 1: shared ──────────────────────────────────────────────────────────
FROM node:24-alpine AS shared-builder
WORKDIR /app
COPY package.json package-lock.json ./
COPY shared/package.json ./shared/
RUN npm ci --workspace=shared
COPY shared/ ./shared/
RUN npm run build --workspace=shared

# ── Stage 2: client ──────────────────────────────────────────────────────────
FROM node:24-alpine AS client-builder
WORKDIR /app
COPY package.json package-lock.json ./
COPY shared/package.json ./shared/
COPY client/package.json ./client/
COPY client/scripts/patch-maplibre.mjs ./client/scripts/
RUN npm ci --workspace=client
COPY --from=shared-builder /app/shared/dist ./shared/dist
COPY client/ ./client/
# A prerelease build does not bump client/package.json, so the bundle takes its
# version from the same build argument the server's VERSION file comes from.
ARG APP_VERSION=dev
RUN npm run build --workspace=client

# ── Stage 3: server ──────────────────────────────────────────────────────────
# --ignore-scripts: nothing here needs a postinstall, tsc only reads types. The one
# native module, better-sqlite3, ships its binaries inside the npm tarball since 13.x.
FROM node:24-alpine AS server-builder
WORKDIR /app
COPY package.json package-lock.json ./
COPY shared/package.json ./shared/
COPY server/package.json ./server/
RUN npm ci --workspace=server --ignore-scripts
COPY --from=shared-builder /app/shared/dist ./shared/dist
COPY server/ ./server/
# tsc needs more than the 2 GB heap Node picks on a host with about 12 GB of RAM;
# on a bigger build machine the default is larger, so only small hosts failed.
RUN NODE_OPTIONS=--max-old-space-size=4096 npm run build --workspace=server

# ── Stage 4: production runtime ──────────────────────────────────────────────
FROM node:24-trixie-slim
WORKDIR /app

# Workspace manifests only — source never enters this stage.
COPY package.json package-lock.json ./
COPY shared/package.json ./shared/
COPY server/package.json ./server/

# The trailing chown runs in this layer on purpose: it covers the manifests and
# the freshly installed node_modules while they are already part of this layer's
# changeset, so it costs nothing. Everything copied after this point carries
# --chown=node:node for the same reason — a recursive chown in a later layer
# would copy up every inode it touches and duplicate the whole tree in the image.
#
# No compiler in this stage. better-sqlite3 13.x carries prebuilt binaries for
# amd64 and arm64 inside its tarball, so there is nothing left to build: the old
# python3 + build-essential fallback compiled it against whatever Node headers
# the base image had that day, and 24.19+ headers give an addon that aborts with
# "Assertion failed: (env) != nullptr" the first time a statement is collected
# (#2421). --ignore-scripts because a lockfile-driven npm ci still synthesises
# `node-gyp rebuild` for the binding.gyp it finds, blind to the package's
# gypfile:false (npm/cli#9837). The only other install scripts among the runtime
# deps are esbuild's binary check, redundant next to its optional platform
# package, and a telemetry ping.
RUN apt-get update && \
    apt-get install -y --no-install-recommends tzdata dumb-init wget ca-certificates \
    libkitinerary-bin libsqlite3-0 && \
    npm ci --workspace=server --omit=dev --ignore-scripts && \
    { ln -sf "$(find /usr/lib -name kitinerary-extractor -type f | head -1)" /usr/local/bin/kitinerary-extractor || true; } && \
    rm -rf /var/lib/apt/lists/* /usr/local/lib/node_modules/npm /usr/local/bin/npm /usr/local/bin/npx && \
    chown -R node:node /app

# gosu rebuilt with a current Go toolchain (stage 0) — the entrypoint script uses
# it to drop to node.
COPY --from=gosu-build /out/gosu /usr/local/bin/gosu

ENV XDG_CACHE_HOME=/tmp/kf6-cache
# Prevent Qt from probing for a display in headless containers.
ENV QT_QPA_PLATFORM=offscreen
# Fixed path for both amd64 (static binary) and arm64 (symlink to apt binary).
# Override with KITINERARY_EXTRACTOR_PATH if you install it elsewhere.
ENV KITINERARY_EXTRACTOR_PATH=/usr/local/bin/kitinerary-extractor

COPY --chown=node:node --from=server-builder /app/server/dist ./server/dist
# Runtime data assets read from server/assets at runtime: airports.json (flight
# transport search) and atlas/*.geojson.gz (Atlas country/region map). The build
# only emits dist, so these must be copied explicitly or the features silently
# degrade to empty in the image.
COPY --chown=node:node --from=server-builder /app/server/assets ./server/assets
# The in-app help pages (/help) read this straight from disk at runtime, so the
# docs always match the version running. Without it, wikiService falls back to
# fetching the GitHub wiki, which tracks main and needs network access.
COPY --chown=node:node wiki ./wiki
# tsconfig-paths/register reads this at runtime to resolve MCP SDK paths.
COPY --chown=node:node server/tsconfig.json ./server/
# Encryption-key rotation is run on demand via tsx (a prod dep) straight from the
# raw .ts source — it never enters dist, so it must be copied in explicitly or
# `node --import tsx scripts/migrate-encryption.ts` fails with module-not-found.
COPY --chown=node:node server/scripts/migrate-encryption.ts ./server/scripts/migrate-encryption.ts
# Admin recovery script (node server/reset-admin.js) for locked-out installs.
COPY --chown=node:node server/reset-admin.js ./server/reset-admin.js
COPY --chown=node:node --from=shared-builder /app/shared/dist ./shared/dist
COPY --chown=node:node --from=client-builder /app/client/dist ./server/public
COPY --chown=node:node --from=client-builder /app/client/public/fonts ./server/public/fonts

# journey/ and places/ must be listed here and in server/src/index.ts (#1762) —
# a dir created lazily on first upload needs write permission on the uploads
# mount point itself, which fails with EACCES when the bind-mounted host dir
# isn't writable by node. Only paths this layer creates are chowned; anything
# already in the image arrived node-owned via --chown above.
RUN mkdir -p /app/data/logs /app/uploads/files /app/uploads/covers /app/uploads/avatars \
      /app/uploads/photos /app/uploads/journey /app/uploads/places && \
    ln -s /app/uploads /app/server/uploads && \
    ln -s /app/data /app/server/data && \
    chown -R node:node /app/data /app/uploads && \
    chown -h node:node /app/server/uploads /app/server/data

ENV NODE_ENV=production
ENV NODE_USE_ENV_PROXY=1
ENV PORT=3000
# The version goes into a file, not into ENV. A container keeps its environment
# when it is recreated on a newer image (Portainer's recreate, a copied run
# config), so an APP_VERSION env var would keep announcing the old release and
# the clients would never pick up the new bundle. The server reads this file
# first (server/src/app-config/image-version.ts).
ARG APP_VERSION=dev
RUN if [ "$APP_VERSION" != "dev" ]; then printf '%s\n' "$APP_VERSION" > /app/server/VERSION; fi

# OCI metadata: Renovate, Watchtower and the registries read the source label to
# link an image update to its release notes (#1498). The release workflows add
# the commit and the build time; a local build carries everything else.
LABEL org.opencontainers.image.title="TREK" \
      org.opencontainers.image.description="Self-hosted collaborative travel planner" \
      org.opencontainers.image.url="https://github.com/liketrek/TREK" \
      org.opencontainers.image.source="https://github.com/liketrek/TREK" \
      org.opencontainers.image.documentation="https://github.com/liketrek/TREK/wiki" \
      org.opencontainers.image.licenses="AGPL-3.0-only" \
      org.opencontainers.image.version="${APP_VERSION}"

EXPOSE 3000

HEALTHCHECK --interval=30s --timeout=5s --start-period=15s --retries=3 \
  CMD wget -qO- "http://localhost:${PORT:-3000}/api/health" || exit 1

# Start-up lives in a script, not an inline `sh -c` string: container management
# UIs re-tokenise Config.Cmd when you edit a container in place, and a command
# carrying quotes does not survive that round-trip (the container then dies with
# a shell syntax error before Node starts). A CMD of one bare token does.
COPY server/scripts/entrypoint.sh /usr/local/bin/trek-entrypoint
# Root-owned on purpose — this is what calls gosu, so it runs before the drop to
# node. chmod is explicit because a checkout on Windows carries no exec bit.
RUN chmod 0755 /usr/local/bin/trek-entrypoint

ENTRYPOINT ["dumb-init", "--"]
CMD ["/usr/local/bin/trek-entrypoint"]