Files
kima-hub/docker-compose.prod.yml
Your Name 9083835bfd chore: v1.7.0 -- vibe galaxy, CI pipeline, enrichment hardening, PWA, preprod sweep
- Bump frontend and backend to 1.7.0
- Update CHANGELOG with full 1.7.0 release notes
- Remove vibe-test dev prototype page and unused R3F components
  (VibeUniverse, TrackCloud, TrackTooltip, universeUtils)
- Fix stale audio.completed counter: flush live DB count at isFullyComplete
  transition -- counter was frozen at last audioQueued > 0 cycle value
- Add GitHub Actions CI pipeline: lint/typecheck, unit tests, security scan,
  E2E predeploy, nightly Docker build and push to Hub + GHCR
- Add E2E enrichment cycle spec with 55-min timeout and memory monitoring script
- Add E2E vibe spec covering map, song path, search, alchemy, similar tracks
- PWA hardening: offline fallback, update banner, WCO, manifest fixes
- Production readiness: OOM memory caps in both compose files, DoS/SSRF/auth fixes
- Remove double-auth in systemSettings (requireAdmin already enforces auth)
- Fix mobile vibe page full-height rendering, vibe map timer leak, abort signal wiring
- Fix E2E test helpers: graceful skip with waitFor + try/catch for empty-library CI
- Fix create-e2e-user.sh: admin role, bcrypt shell expansion, psql heredoc quoting
2026-03-16 18:25:08 -05:00

54 lines
1.8 KiB
YAML

# Kima - Premium Self-Hosted Music Server
# All-in-One container with AI-powered vibe matching
#
# Features included:
# - Smart playlists with mood detection
# - Enhanced vibe matching (Essentia AI)
# - High-quality audio streaming
#
# Usage:
# 1. Edit the MUSIC_PATH below
# 2. Run: docker compose -f docker-compose.prod.yml up -d
# 3. Open http://localhost:3030
services:
kima:
image: ${DOCKERHUB_USERNAME:-chevron7locked}/kima:${VERSION:-latest}
container_name: kima
ports:
- "${PORT:-3030}:3030"
volumes:
# REQUIRED: Path to your music library
- ${MUSIC_PATH:-/path/to/your/music}:/music
# Persistent data (database, cache, etc.)
- kima_data:/data
environment:
- TZ=${TZ:-UTC}
- SESSION_SECRET=${SESSION_SECRET:-}
# Lidarr webhook callback URL - how Lidarr reaches Kima when downloads complete
# Default uses host.docker.internal which works on most setups with extra_hosts below
# Override if using custom Docker networks: e.g., http://192.168.0.20:3030
- KIMA_CALLBACK_URL=${KIMA_CALLBACK_URL:-http://host.docker.internal:3030}
- DISABLE_CLAP=${DISABLE_CLAP:-}
# Makes host.docker.internal work on Linux (already works on Docker Desktop)
extra_hosts:
- "host.docker.internal:host-gateway"
# Cap memory to prevent host kernel OOM cascade from anon_vma_chain slab exhaustion.
# Without this, the host kernel OOM killer fires when Next.js VMA chains accumulate.
# 6g is enough for frontend + backend + embeddings + Redis + Postgres.
mem_limit: 6g
memswap_limit: 8g
# Fix Redis memory overcommit warning
sysctls:
- vm.overcommit_memory=1
restart: unless-stopped
healthcheck:
test: ["CMD", "node", "/app/healthcheck.js"]
interval: 30s
timeout: 10s
start_period: 60s
retries: 3
volumes:
kima_data: