[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AYF1SC0FZBN62MD4908ZSJ":3,"skill-same:01M7AYF1SC0FZBN62MD4908ZSJ":128},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":42,"commitSha":43,"skillDirectory":44,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":48,"ownerDescription":49,"updatedAt":50,"install":51,"resources":64},"01M7AYF1SC0FZBN62MD4908ZSJ","homelab-microservice-deployment",null,"Deploy and configure containerized microservices on a Docker-based homelab — Keycloak, Spring Cloud Gateway, Eureka, Nginx + Cloudflare tunnel, Docker Compose multi-service management.","oat431\u002Fhermes_config_backup","oat431","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F54722120?v=4","01M7AYD07EH8SYPMYY8CGZKS22","Oralita","01M7AYD0816QDF77E3TWCRSQC9",1,"Unlicense","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: homelab-microservice-deployment\ndescription: Deploy and configure containerized microservices on a Docker-based homelab — Keycloak, Spring Cloud Gateway, Eureka, Nginx + Cloudflare tunnel, Docker Compose multi-service management.\ntriggers:\n  - deploying a new service to the homelab\n  - configuring Keycloak \u002F OAuth2 \u002F OIDC\n  - setting up Spring Cloud Gateway or Eureka\n  - Docker Compose for multi-service platforms\n  - Nginx reverse proxy behind Cloudflare\n  - troubleshooting container healthchecks or networking\n  - consolidating bundled databases into shared infrastructure\n  - connecting containers to db-network\n  - Infisical deployment or configuration\n  - OTS \u002F one-time-secret deployment\n  - migrating bundled Redis\u002FValkey\u002FPostgreSQL to shared infrastructure\n  - SearXNG configuration or rate limiting\n  - MCP server pointing at homelab services\n  - Tailscale routing for local development tools\n  - AdGuard Home DNS filtering \u002F blocklist auditing\n  - DNS filtering coverage — devices bypassing the filter\n  - router-side IPv6 DNS (RDNSS \u002F DHCPv6) — make the router hand out the homelab DNS, then verify it\n---\n\n# Homelab Microservice Deployment\n\nDeploy and configure containerized microservices on a Docker homelab server behind Nginx + Cloudflare Tunnel.\n\n## Core Patterns\n\n### 1. Docker Network Convention\n\nAll services join a shared external network `db-network`. Container-to-container communication uses **container names** as hostnames.\n\n```yaml\n# Every compose file includes:\nnetworks:\n  shared-network:\n    external: true\n    name: db-network\n```\n\n**Never use `host.docker.internal` on Linux** — it's a Docker Desktop feature. Use container names on the shared network instead.\n\n### 2. Port Binding Convention\n\nAll services bind to `127.0.0.1` — Nginx proxies externally. Never expose to `0.0.0.0`.\n\n```yaml\nports:\n  - \"127.0.0.1:8001:8080\"   # ✅ localhost only\n  # - \"8001:8080\"            # ❌ exposed to LAN\n```\n\n**Exception:** When a service needs direct Tailscale access (e.g., MCP tools on the host machine), bind to `0.0.0.0`. UFW blocks public access by default — only Tailscale peers can reach it. Example: SearXNG for MCP integration.\n\n### 3. Nginx + Cloudflare Tunnel Pattern\n\nCloudflare terminates TLS. Nginx receives HTTP via the tunnel. Always hardcode `X-Forwarded-Proto: https` — `$scheme` would be `http` since Nginx receives HTTP from the tunnel.\n\n```nginx\nserver {\n    server_name myservice.panomete.com;\n    location \u002F {\n        proxy_pass http:\u002F\u002F127.0.0.1:PORT;\n        proxy_set_header Host $host;\n        proxy_set_header X-Real-IP $remote_addr;\n        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;\n        proxy_set_header X-Forwarded-Proto https;  # Hardcoded, not $scheme\n    }\n}\n```\n\nDNS is automatic — `*.panomete.com` wildcard routes everything through the Cloudflare tunnel. No manual DNS records needed for new subdomains.\n\n### 4. Healthcheck Pattern\n\nTool availability varies by base image — never assume:\n\n| Base Image | `wget` | `curl` | `which` |\n|------------|:------:|:------:|:-------:|\n| Alpine (`*-alpine`) | ✅ (busybox) | ❌ | ❌ |\n| Ubuntu\u002FNoble (`*-noble`) | ❌ | ❌ | ❌ |\n| `grafana\u002Fgrafana` | ✅ | ✅ | ✅ |\n| `prom\u002Fprometheus` | ✅ | ❌ | ✅ |\n| `grafana\u002Floki` | ✅ | ❌ | ✅ |\n| `louislam\u002Fuptime-kuma` | ❌ | ✅ | ✅ |\n| `infisical\u002Finfisical` | ❌ | ❌ | ❌ |\n| Keycloak | ❌ | ❌ | ❌ |\n\n**When `wget` is available (Alpine\u002Fbusybox):**\n```yaml\nhealthcheck:\n  test: [\"CMD\", \"wget\", \"-qO-\", \"http:\u002F\u002Flocalhost:PORT\u002Fhealth\"]\n```\n\n**When `curl` is available:**\n```yaml\nhealthcheck:\n  test: [\"CMD-SHELL\", \"curl -sf http:\u002F\u002Flocalhost:PORT\u002Fhealth > \u002Fdev\u002Fnull\"]\n```\n\n**When neither is available (Keycloak, etc.):**\n```yaml\nhealthcheck:\n  test: [\"CMD-SHELL\", \"timeout 5 bash -c '\u003C\u002Fdev\u002Ftcp\u002Flocalhost\u002FPORT' || exit 1\"]\n```\n\n**When nothing is available (Infisical, etc.):**\nOverride with `test: [\"CMD-SHELL\", \"exit 0\"]` or remove healthcheck entirely.\n\n### 5. .env for Secrets\n\nNever put passwords in compose YAML. Use `.env` file at `~\u002Fplatform\u002F.env` (chmod 600). Compose references with `${VAR_NAME}` or `${VAR_NAME:-default}`.\n\n**Pitfall: `docker compose restart` does NOT reload `.env` changes.** Compose reads `.env` at creation time, not on restart. After editing `.env`, always do a full down\u002Fup cycle:\n```bash\ndocker compose down && docker compose up -d\n```\n`docker compose restart` or `docker compose stop && docker compose start` will use the old values. This is a common source of \"I updated the config but nothing changed\" confusion.\n\n### 6. Compose YAML Structure Pitfall\n\nWhen appending a new service to an existing compose file, Python\u002Fyaml.dump is safer than shell heredoc append — heredoc append can nest the new service under the wrong key (e.g., under `networks:` instead of `services:`).\n\n### 7. Dockerizing pnpm Monorepos (Workspace + Turborepo)\n\npnpm workspace monorepos use symlinks between packages. A naive multi-stage Dockerfile that copies packages individually breaks module resolution at runtime (`ERR_MODULE_NOT_FOUND`).\n\n**Working pattern** — use `pnpm deploy` to flatten the workspace:\n\n```dockerfile\n# Build: install, build shared packages, then deploy to flat bundle\nRUN pnpm install --frozen-lockfile\nRUN pnpm --filter @scope\u002Fshared-pkg build\nRUN pnpm --filter .\u002Fapps\u002Fbackend\u002F --legacy deploy .\u002Fdeployment\u002F\n\n# Run: just copy the flat deployment bundle\nCOPY --from=builder \u002Fapp\u002Fdeployment .\u002Fdeployment\n```\n\n**Pitfall: `--prod` strips runtime deps.** `pnpm deploy --prod` excludes `devDependencies`, but some apps list runtime-critical packages (like `express`) under `devDependencies`. If the app crashes with `Cannot find package 'express'`, remove `--prod` from the deploy command. The project's own `vercel.json` uses `--prod` but pairs it with a `vercel-preparation.sh` script that does additional setup — Docker doesn't have that.\n\n**Pitfall: symlinks don't survive COPY.** Never try to manually copy `packages\u002Fcore\u002Fbuild` + `node_modules` to the runner stage. pnpm's `node_modules\u002F.pnpm` structure uses symlinks that Docker COPY doesn't preserve correctly. Always use `pnpm deploy` to produce a flat, self-contained bundle.\n\n## Keycloak 26+ Pitfalls\n\nSee `references\u002Fkeycloak-pitfalls.md` for the full list. Critical ones:\n\n- `KC_PROXY=edge` is DEPRECATED → use `KC_PROXY_HEADERS=xforwarded`\n- `KEYCLOAK_ADMIN` is DEPRECATED → use `KC_BOOTSTRAP_ADMIN_USERNAME`\n- Bootstrap admin is ALWAYS temporary → create permanent admin via Admin Console\n- `KC_CACHE=local` is REQUIRED for single-node (JGroups clustering timeout without it)\n- Realm `sslRequired` must be `none` when behind Cloudflare (Nginx sends HTTP)\n- Docker bind mount creates a DIRECTORY if the source file doesn't exist → create the file BEFORE starting the container\n\n## Spring Cloud Gateway + Eureka Pattern\n\n### Gateway compose environment (verified)\n\n```yaml\nenvironment:\n  SERVER_PORT: \"8000\"\n  MANAGEMENT_SERVER_PORT: \"8000\"    # Merge actuator onto main port\n  SPRING_PROFILES_ACTIVE: prod\n  # JWT — realm name must match, not \"flowerogate\" or other legacy names\n  JWT_ISSUER_URI: https:\u002F\u002Fauth.panomete.com\u002Frealms\u002Fpanomete\n  JWT_JWK_SET_URI: https:\u002F\u002Fauth.panomete.com\u002Frealms\u002Fpanomete\u002Fprotocol\u002Fopenid-connect\u002Fcerts\n  # Valkey — container name + password (not host.docker.internal)\n  REDIS_HOST: local-valkey\n  REDIS_PORT: \"6379\"\n  REDIS_PASSWORD: ${VALKEY_PASSWORD}\n  REDIS_SSL: \"false\"\n  # Eureka — enable + correct URL (not eureka:8761)\n  EUREKA_CLIENT_ENABLED: \"true\"\n  EUREKA_URI: http:\u002F\u002Fflowero-discover:8999\u002Feureka\n  # Disable features not yet needed\n  SPRING_AUTOCONFIGURE_EXCLUDE: org.springframework.cloud.circuitbreaker.resilience4j.Resilience4JAutoConfiguration\n  OTLP_ENABLED: \"false\"\n  # SecurityConfig requires this — no default in code\n  APP_POST_LOGIN_REDIRECT_URL: ${POST_LOGIN_REDIRECT_URL}\n  KEYCLOAK_GATEWAY_SECRET: ${KEYCLOAK_GATEWAY_SECRET}\n```\n\n### Gateway Dockerfile (Alpine — uses wget for healthcheck)\n\n```dockerfile\nFROM eclipse-temurin:25-jdk-alpine AS builder\nWORKDIR \u002Fapp\nCOPY gradlew gradlew.bat settings.gradle build.gradle .\u002F\nCOPY gradle\u002F gradle\u002F\nRUN chmod +x gradlew && .\u002Fgradlew dependencies --no-daemon 2>\u002Fdev\u002Fnull || true\nCOPY src\u002F src\u002F\nRUN .\u002Fgradlew bootJar --no-daemon -x test\n\nFROM eclipse-temurin:25-jre-alpine\nRUN addgroup -S flowero && adduser -S flowero -G flowero\nWORKDIR \u002Fapp\nCOPY --from=builder \u002Fapp\u002Fbuild\u002Flibs\u002F*.jar app.jar\nEXPOSE 8000\nUSER flowero\n# Alpine has wget via busybox — no curl install needed\nHEALTHCHECK --interval=30s --timeout=5s --retries=3 \\\n  CMD wget -qO- http:\u002F\u002Flocalhost:8000\u002Factuator\u002Fhealth\u002Fliveness || exit 1\nENTRYPOINT [\"java\", \"-XX:+UseZGC\", \"-XX:MaxRAMPercentage=75.0\", \"-jar\", \"app.jar\"]\n```\n\n**Key difference from Ubuntu-based images:** Alpine (`jre-alpine`) has `wget` via\nbusybox but NOT `curl`. Ubuntu (`jre-noble`) has neither. Use `wget` for Alpine\nhealthchecks, or `apt-get install curl` for Ubuntu.\n\n### Gateway routing (application.yaml)\n\n```yaml\nspring:\n  cloud:\n    gateway:\n      routes:\n        - id: blog\n          uri: lb:\u002F\u002Fcute-gufo              # Eureka service name\n          predicates: [Path=\u002Fapi\u002Fblog\u002F**]\n          filters:\n            - StripPrefix=1                # \u002Fapi\u002Fblog\u002Fposts → \u002Fposts\n            - RemoveRequestHeader=Authorization\n            - name: RequestRateLimiter\n              args:\n                key-resolver: \"#{@ipKeyResolver}\"\n                redis-rate-limiter.replenishRate: 100\n                redis-rate-limiter.burstCapacity: 200\n        # ... more routes\n```\n\n### Gateway prerequisites checklist\n\nBefore deploying Gate, verify:\n- [ ] Guard healthy (JWKS endpoint reachable)\n- [ ] Discover healthy (Eureka registry available)\n- [ ] Valkey healthy (password known)\n- [ ] `.env` has: `VALKEY_PASSWORD`, `KEYCLOAK_GATEWAY_SECRET`, `POST_LOGIN_REDIRECT_URL`\n- [ ] `flowero-gateway` OAuth2 client registered in Keycloak `panomete` realm\n- [ ] Nginx block for `api.panomete.com` created\n- [ ] Port 8000 free\n\n### Gateway 401 behavior\n\n- **Browser request** (`Accept: text\u002Fhtml`) → 302 redirect to Keycloak login\n- **API request** (no token) → 401 JSON body\n- **API request** (valid JWT) → route to backend via Eureka → 502 if no backend registered (expected in Phase 1)\n\n### Post-login redirect URL\n\n`SecurityConfig` uses `@Value(\"${app.post-login-redirect-url}\")` with NO default.\nIf `POST_LOGIN_REDIRECT_URL` is not set in `.env`, the app crashes on startup.\nAlways set it before deploying Gate.\n\n## Repository Architecture\n\n**NOT a monorepo.** Each service has its own GitHub repo. CI\u002FCD workflows live inside each repo (not a shared `.github\u002Fworkflows\u002F` at root).\n\n```\noat431\u002Fflowero-guard      → .github\u002Fworkflows\u002F{ci,deploy}.yml\noat431\u002Fflowero-discovery  → .github\u002Fworkflows\u002F{ci,deploy}.yml\noat431\u002Fflowero-gateway    → .github\u002Fworkflows\u002F{ci,deploy}.yml\n```\n\nEach repo gets its own `TS_AUTH_KEY` and `HOMELAB_SSH_KEY` secrets (duplicated across repos).\n\n## High-Risk Boundary (User Handles)\n\nThe user handles these manually — never automate:\n- Passwords, API keys, SSH keys\n- `.env` file edits with secrets\n- UFW firewall rules\n- Portainer admin password\n- rclone OAuth re-authentication\n\nEverything else (container lifecycle, DB provisioning, Nginx configs, realm JSON, CI\u002FCD workflows) the assistant can handle.\n\n## Obsidian Documentation Pattern\n\nOperational knowledge lives in Obsidian at `F:\\obsidian_note\\oralita_md\\`, NOT in the spec folder and NOT on the homelab server. **The vault gets reorganized between sessions — always verify the current structure with a quick `ls` before writing docs** (paths below changed once already: `home-lab-installation\u002F` → `home-lab\u002F`, and Quick Note guides moved into `home-lab\u002Fobsidian-livesync\u002F`).\n\nStructure (as of 2026-08):\n\n```\noralita_md\u002F\n├── home-lab\u002F\n│   ├── audit\u002F                 # dated audit\u002Fdeployment logs\n│   ├── database_component\u002F    # PostgreSQL, Valkey, MongoDB, CouchDB\n│   ├── microservice_component\u002F# Keycloak, Discovery, Gateway\n│   ├── deployment\u002F            # CI\u002FCD, backup automation\n│   ├── essential_component\u002F   # Nginx, Cloudflare, Tailscale\n│   ├── miscellaneous\u002F         # db-network-integration-guide.md, etc.\n│   └── obsidian-livesync\u002F     # 01-full-setup, 02-connection-guide, 03-couchdb-user-management\n└── Quick Note\u002F                # user's scratch notes (Home Lab App.md port table lives here)\n```\n\nEach service gets a full-detail self-contained note with: What & Why, Setup, Deploy, Access, Common Commands, Configuration, Troubleshooting, Project Structure, Related.\n\n**Doc handoff convention:** the user explicitly corrected server-side doc placement — write guides to the LOCAL Obsidian vault, never to `~\u002Fapplication\u002Fmisc\u002F` on the homelab (a stray copy landed there once and was ignored).\n\n## Deployment Order\n\nFor Spring Cloud microservice platforms, deploy in dependency order:\n1. Databases (PostgreSQL, Valkey) — already running\n2. Identity provider (Keycloak) — needs DB only\n3. Service registry (Eureka) — standalone, no deps\n4. API Gateway — depends on all above\n\n**Pitfall:** Deploying Gateway before Discovery forces hardcoded routes that must\nbe rewritten later. Always deploy Discovery first — it's standalone and takes\n30 seconds to build.\n\n## Observability Stack Pattern\n\nSee `references\u002Fobservability-stack.md` for the full Prometheus + Grafana + Loki + Promtail deployment.\n\n### Quick Summary\n\nDeploy as a **separate compose file** (`docker-compose.observability.yml`) to keep observability isolated from platform services.\n\n| Service | Port | Domain | Purpose |\n|---------|:----:|--------|---------|\n| Prometheus | 9090 | Internal only | Scrape metrics from all services |\n| Grafana | 3000 | `grafana.panomete.com` | Dashboards + alerts |\n| Loki | 3100 | Internal | Log aggregation |\n| Promtail | — | Internal | Tail container logs → Loki |\n| Uptime Kuma | 3001 | `status.panomete.com` | Uptime monitoring |\n\n### Key Pitfalls\n\n1. **Prometheus target \"down\" with 404** — Spring Boot needs `micrometer-registry-prometheus` dependency AND `prometheus` in `management.endpoints.web.exposure.include`. Without the dependency, the endpoint returns 404.\n\n2. **Keycloak metrics on management port 9000, NOT main port 8080** — With `KC_METRICS_ENABLED=true`, metrics are at `http:\u002F\u002Fcontainer:9000\u002Fmetrics`. Prometheus scrape target must use port 9000.\n\n3. **Promtail `docker_sd_configs` with filters fails** — Use filesystem log reading instead:\n   ```yaml\n   scrape_configs:\n     - job_name: docker-logs\n       static_configs:\n         - targets: [localhost]\n           labels: {job: docker-logs, __path__: \u002Fvar\u002Flib\u002Fdocker\u002Fcontainers\u002F*\u002F*.log}\n       pipeline_stages:\n         - docker: {}\n   ```\n\n4. **Loki ingestion rate limit exceeded** — Default is 4MB\u002Fs. Increase in loki-config.yml:\n   ```yaml\n   limits_config:\n     ingestion_rate_mb: 16\n     ingestion_burst_size_mb: 32\n   ```\n\n5. **Grafana alert rules \"data source not found\"** — Alert rules need the actual Prometheus datasource UID, not a placeholder. Get it via:\n   ```bash\n   curl -sf -u admin:$GRAFANA_ADMIN_PASSWORD http:\u002F\u002Flocalhost:3000\u002Fapi\u002Fdatasources | python3 -c \"import sys,json; [print(d['uid']) for d in json.load(sys.stdin) if d['type']=='prometheus']\"\n   ```\n\n6. **Keycloak healthcheck — no tools in image** — Keycloak image has no curl, wget, or even `which`. Use TCP check:\n   ```yaml\n   healthcheck:\n     test: [\"CMD-SHELL\", \"timeout 5 bash -c '\u003C\u002Fdev\u002Ftcp\u002Flocalhost\u002F8080' || exit 1\"]\n   ```\n\n7. **Uptime Kuma healthcheck — `wget` not found** — The `louislam\u002Fuptime-kuma` image does NOT have `wget` (unlike Prometheus\u002FGrafana\u002FLoki which do). The default image healthcheck uses `wget` and permanently fails with `exec: \"wget\": executable file not found in $PATH`. Fix by overriding in compose:\n   ```yaml\n   uptime-kuma:\n     healthcheck:\n       test: [\"CMD-SHELL\", \"curl -sf http:\u002F\u002Flocalhost:3001 > \u002Fdev\u002Fnull\"]\n       interval: 15s\n       timeout: 5s\n       retries: 3\n   ```\n   `curl` IS available in the Uptime Kuma container. Container will go from `unhealthy` to `healthy` within one check interval after recreating.\n\n### Prometheus Scrape Config (verified)\n\n```yaml\nscrape_configs:\n  - job_name: \"flowero-gate\"\n    metrics_path: \"\u002Factuator\u002Fprometheus\"\n    static_configs:\n      - targets: [\"flowero-gate:8000\"]\n  - job_name: \"flowero-discover\"\n    metrics_path: \"\u002Factuator\u002Fprometheus\"\n    static_configs:\n      - targets: [\"flowero-discover:8999\"]\n  - job_name: \"flowero-guard\"\n    metrics_path: \"\u002Fmetrics\"\n    static_configs:\n      - targets: [\"flowero-guard:9000\"]  # Management port, NOT 8080\n  - job_name: \"prometheus\"\n    static_configs:\n      - targets: [\"localhost:9090\"]\n```\n\n## Backup Automation Pattern\n\nSee `references\u002Fbackup-automation.md` for the full pg_dumpall + rclone + cron pattern.\n\n### Quick Summary\n\nTwo scripts on the server at `\u002Fhome\u002Fflowero\u002Fscripts\u002F`:\n\n| Script | Schedule | What | Retention |\n|--------|----------|------|-----------|\n| `backup-db.sh` | Daily 3:00 AM | `pg_dumpall` → gzip → rclone → OneDrive | 7 days local |\n| `backup-volumes.sh` | Weekly Sunday 4:00 AM | Docker volumes → tar → gzip → rclone | 4 weeks local |\n\n### Docker Volume Naming Pitfall\n\nWhen using named volumes in Docker Compose, the actual volume name on disk is NOT the compose key. Always check with `docker volume ls` before writing backup scripts. The backup script's `VOLUMES` array must use the actual volume names.\n\n## Service Consolidation Pattern (Shared Databases)\n\nWhen a third-party application ships its own bundled database (Valkey, Redis, PostgreSQL), replace it with the shared instance on `db-network`. This reduces resource usage and centralizes data management.\n\n### Steps\n\n1. **Remove the bundled service** from the application's `docker-compose.yml`\n2. **Remove its volume** (the shared instance has its own)\n3. **Add the application to `db-network`**:\n   ```yaml\n   services:\n     app:\n       networks:\n         - db-network\n   networks:\n     db-network:\n       external: true\n   ```\n4. **Configure the connection** — point the app at the shared service. If the shared service has a DNS alias matching the old bundled service name (e.g., `local-valkey` has alias `valkey`), the app may work with zero config changes.\n5. **Handle authentication** — shared services may have passwords the bundled ones didn't. Add credentials via `settings.yml`, `.env`, or environment variables.\n\n### DNS Alias Trick\n\n`local-valkey` on `db-network` has aliases `[\"local-valkey\", \"valkey\"]`. Applications that reference `valkey:6379` resolve to the shared instance automatically — no config rename needed.\n\n### Verifying Connection\n\n```bash\n# Check container is on the right network\ndocker ps --filter \"name=APP\" --format \"{{.Networks}}\"\n\n# Check logs for connection errors\ndocker logs APP 2>&1 | grep -i \"redis\\|valkey\\|connect\\|refused\"\n\n# HTTP health check\ncurl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" http:\u002F\u002F127.0.0.1:PORT\u002F\n```\n\n### Pitfalls\n\n- **Remove `depends_on` healthcheck conditions when deleting bundled DBs** — compose files often declare `depends_on: {bundled-db: {condition: service_healthy}}`. Once the bundled service is removed, these blocks break `docker compose up` with \"dependency failed to start\". Delete them along with the service (external services can't satisfy `depends_on` anyway).\n- **Multi-service apps may reference the DB from several services** — e.g., Penpot's backend AND exporter both declare `PENPOT_REDIS_URI`. Grep the whole compose for DB connection env vars and update every occurrence, not just the main app service.\n- **File permissions (UID mismatch):** Container config files are often owned by the container's internal UID (e.g., SearXNG uses UID 977). Regular shell redirection (`>`) fails with \"Permission denied\". Use `sudo tee` instead:\n  ```bash\n  sudo tee \u002Fpath\u002Fto\u002Fcontainer-owned\u002Fconfig.yml > \u002Fdev\u002Fnull \u003C\u003C 'EOF'\n  content here\n  EOF\n  ```\n- **Env var overrides config file:** Some apps (like SearXNG) let environment variables override config files. If you set the connection in `settings.yml` AND as an env var, the env var wins — and may use the wrong key name (causing deprecation warnings). Pick one approach, not both.\n- **Network name collision:** If the bundled service had the same container name as a shared service alias, Docker DNS conflicts. Rename the bundled service before removing it, or just remove it entirely.\n- **SQLite-based apps cannot integrate:** Some apps (e.g., ByteStash uses `better-sqlite3`) are file-based with no network database support. Check `package.json` or container dependencies before attempting migration.\n- **Port conflicts:** Always check occupied ports (`ss -tlnp | grep 127.0.0.1:`) before deploying. Common conflicts: port 3000 (Grafana), port 80 (system services).\n\n### Documentation Workflow\n\nWhen consolidating services, create a running guide in the user's **Obsidian vault** (`F:\\obsidian_note\\oralita_md\\home-lab\\miscellaneous\\` — verify the path first; the vault gets reorganized). The user reviews and edits docs locally in Obsidian — server-side docs get ignored.\n\nAlso update the port reference in `F:\\obsidian_note\\oralita_md\\Quick Note\\Home Lab App.md` when ports change.\n\n### Sequential Review Preference\n\nProcess one service at a time. After each migration:\n1. Start the container\n2. Verify it's healthy (HTTP check, logs, DB connectivity)\n3. Present the results to the user\n4. **Wait for user approval** before proceeding to the next service\n\nDo NOT batch all migrations in one shot — the user reviews each change before moving on.\n\n### Port Conflicts\n\nPort 3000 is commonly taken by Grafana. Always check before deploying:\n\n```bash\nss -tlnp | grep -E \"127\\.0\\.0\\.1:\" | awk '{print $4}' | sort -t: -k2 -n\n```\n\nKnown port allocations on the Panomete homelab:\n\n| Port | Service |\n|------|---------|\n| 3000 | Grafana |\n| 3001 | Uptime Kuma |\n| 5432 | PostgreSQL |\n| 5984 | CouchDB |\n| 6379 | Valkey |\n| 7000 | AdGuard |\n| 7004 | SearXNG |\n| 7005 | Infisical |\n| 7006 | OTS |\n| 7007 | StirlingPDF |\n| 7008 | ByteStash |\n| 7009 | Penpot |\n| 7010 | GitHub Stats Extended |\n| 8000 | Spring Cloud Gateway |\n| 8001 | Keycloak |\n| 8999 | Eureka |\n| 9000 | Portainer |\n| 9001 | Penpot (alt, direct access) |\n| 9090 | Prometheus |\n| 27017 | MongoDB |\n\n## CI\u002FCD Migration: build → image (Critical Pattern)\n\nOnce CI\u002FCD is set up, the server must ONLY pull pre-built images from GHCR. Never build on the server.\n\n**Wrong (building on server):**\n```yaml\nflowero-discover:\n  build:\n    context: .\u002Fflowerodiscovery\n    dockerfile: Dockerfile\n```\n\n**Correct (pulling from GHCR):**\n```yaml\nflowero-discover:\n  image: ghcr.io\u002Foat431\u002Fflowero-discovery:latest\n```\n\n**Migration steps:**\n1. Update compose: replace `build:` with `image: ghcr.io\u002F...`\n2. Remove local repo copies from server (`rm -rf ~\u002Fplatform\u002Fflowerodiscovery`, etc.)\n3. Server layout becomes: `.env`, `docker-compose.platform.yml`, `docker-compose.observability.yml`, `observability\u002F` config dir\n4. Verify: `docker ps` should show GHCR image names, not local build names\n\n**Why:** The user explicitly corrected this — \"why you modified the local repo in the first place, i thought we just setup ci\u002Fcd\". Source code lives in GitHub. CI builds + pushes to GHCR. Server pulls. No source code on server.\n\n## CI\u002FCD for Tailscale-Behind Servers\n\nWhen the homelab resolves to a Tailscale IP, GitHub Actions can't SSH directly.\nUse `tailscale\u002Fgithub-action` to bridge the runner onto the Tailnet temporarily,\nthen deploy via SSH. Deploy should be manual (`workflow_dispatch`) for single-dev\nsafety. See `references\u002Ftailscale-cicd-bridge.md` for the full workflow pattern.\n\n### CI Pitfalls (Gradle projects)\n\n- **`gradlew: Permission denied`** — Windows doesn't preserve Unix execute bits.\n  Fix: `git update-index --chmod=+x gradlew` AND add `chmod +x gradlew` step in CI.\n- **`Task 'checkstyleMain' not found`** — Only run if Checkstyle plugin is in `build.gradle`.\n  If the project doesn't have it, remove the lint step entirely.\n- **`.dockerignore` excluding needed files** — The Guard CI failed because\n  `panomete-realm.json` was listed in `.dockerignore`. Always verify `.dockerignore`\n  doesn't exclude files referenced by `COPY` in the Dockerfile.\n\n### GHCR Package Visibility Pitfall\n\nGHCR package visibility is **separate from repo visibility**. Making a repo public\ndoes NOT make its container image public. You must also set the package to public:\n\n1. Go to `https:\u002F\u002Fgithub.com\u002F\u003Cowner>\u002F\u003Crepo>\u002Fpackages`\n2. Click the container package → **Package settings**\n3. Under **Danger Zone** → **Change visibility** → **Public**\n\nOr link the package to the repo (inherits visibility). Without this,\n`docker pull` from a public repo still returns `unauthorized`.\n\n### Docker Volume Naming\n\nWhen using named volumes in Docker Compose, the actual volume name on disk is\n`{project}_{service}_{volume}`, NOT just the volume key. Always check with\n`docker volume ls` before writing backup scripts. Example:\n- Compose key: `postgres_data` in project `platform`\n- Actual name: `platform_postgres_data` or `postgres_postgres_data` (depends on compose setup)\n\n## Backup Automation Pattern\n\nSee `references\u002Fbackup-automation.md` for the full pg_dumpall + rclone + cron pattern.\n\n## Related\n\n- References: `references\u002Fkeycloak-pitfalls.md`, `references\u002Fdocker-compose-patterns.md`, `references\u002Ftailscale-cicd-bridge.md`, `references\u002Fsearxng.md`, `references\u002Finfisical.md`, `references\u002Fots.md`, `references\u002Fbytestash.md`, `references\u002Fpenpot.md`, `references\u002Fcouchdb.md`, `references\u002Fadguard-dns-filtering.md`, `references\u002Fobservability-stack.md`, `references\u002Fbackup-automation.md`; scripts `scripts\u002Fadguard-qlog-stats.py` (AdGuard query-log stats), `scripts\u002Fverify-router-dns.sh` (verify router-side IPv6 DNS hand-out)\n- See also: `homelab-infra-audit` for server auditing before deployment\n","# Homelab Microservice Deployment\n\nDeploy and configure containerized microservices on a Docker homelab server behind Nginx + Cloudflare Tunnel.\n\n## Core Patterns\n\n### 1. Docker Network Convention\n\nAll services join a shared external network `db-network`. Container-to-container communication uses **container names** as hostnames.\n\n```yaml\n# Every compose file includes:\nnetworks:\n  shared-network:\n    external: true\n    name: db-network\n```\n\n**Never use `host.docker.internal` on Linux** — it's a Docker Desktop feature. Use container names on the shared network instead.\n\n### 2. Port Binding Convention\n\nAll services bind to `127.0.0.1` — Nginx proxies externally. Never expose to `0.0.0.0`.\n\n```yaml\nports:\n  - \"127.0.0.1:8001:8080\"   # ✅ localhost only\n  # - \"8001:8080\"            # ❌ exposed to LAN\n```\n\n**Exception:** When a service needs direct Tailscale access (e.g., MCP tools on the host machine), bind to `0.0.0.0`. UFW blocks public access by default — only Tailscale peers can reach it. Example: SearXNG for MCP integration.\n\n### 3. Nginx + Cloudflare Tunnel Pattern\n\nCloudflare terminates TLS. Nginx receives HTTP via the tunnel. Always hardcode `X-Forwarded-Proto: https` — `$scheme` would be `http` since Nginx receives HTTP from the tunnel.\n\n```nginx\nserver {\n    server_name myservice.panomete.com;\n    location \u002F {\n        proxy_pass http:\u002F\u002F127.0.0.1:PORT;\n        proxy_set_header Host $host;\n        proxy_set_header X-Real-IP $remote_addr;\n        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;\n        proxy_set_header X-Forwarded-Proto https;  # Hardcoded, not $scheme\n    }\n}\n```\n\nDNS is automatic — `*.panomete.com` wildcard routes everything through the Cloudflare tunnel. No manual DNS records needed for new subdomains.\n\n### 4. Healthcheck Pattern\n\nTool availability varies by base image — never assume:\n\n| Base Image | `wget` | `curl` | `which` |\n|------------|:------:|:------:|:-------:|\n| Alpine (`*-alpine`) | ✅ (busybox) | ❌ | ❌ |\n| Ubuntu\u002FNoble (`*-noble`) | ❌ | ❌ | ❌ |\n| `grafana\u002Fgrafana` | ✅ | ✅ | ✅ |\n| `prom\u002Fprometheus` | ✅ | ❌ | ✅ |\n| `grafana\u002Floki` | ✅ | ❌ | ✅ |\n| `louislam\u002Fuptime-kuma` | ❌ | ✅ | ✅ |\n| `infisical\u002Finfisical` | ❌ | ❌ | ❌ |\n| Keycloak | ❌ | ❌ | ❌ |\n\n**When `wget` is available (Alpine\u002Fbusybox):**\n```yaml\nhealthcheck:\n  test: [\"CMD\", \"wget\", \"-qO-\", \"http:\u002F\u002Flocalhost:PORT\u002Fhealth\"]\n```\n\n**When `curl` is available:**\n```yaml\nhealthcheck:\n  test: [\"CMD-SHELL\", \"curl -sf http:\u002F\u002Flocalhost:PORT\u002Fhealth > \u002Fdev\u002Fnull\"]\n```\n\n**When neither is available (Keycloak, etc.):**\n```yaml\nhealthcheck:\n  test: [\"CMD-SHELL\", \"timeout 5 bash -c '\u003C\u002Fdev\u002Ftcp\u002Flocalhost\u002FPORT' || exit 1\"]\n```\n\n**When nothing is available (Infisical, etc.):**\nOverride with `test: [\"CMD-SHELL\", \"exit 0\"]` or remove healthcheck entirely.\n\n### 5. .env for Secrets\n\nNever put passwords in compose YAML. Use `.env` file at `~\u002Fplatform\u002F.env` (chmod 600). Compose references with `${VAR_NAME}` or `${VAR_NAME:-default}`.\n\n**Pitfall: `docker compose restart` does NOT reload `.env` changes.** Compose reads `.env` at creation time, not on restart. After editing `.env`, always do a full down\u002Fup cycle:\n```bash\ndocker compose down && docker compose up -d\n```\n`docker compose restart` or `docker compose stop && docker compose start` will use the old values. This is a common source of \"I updated the config but nothing changed\" confusion.\n\n### 6. Compose YAML Structure Pitfall\n\nWhen appending a new service to an existing compose file, Python\u002Fyaml.dump is safer than shell heredoc append — heredoc append can nest the new service under the wrong key (e.g., under `networks:` instead of `services:`).\n\n### 7. Dockerizing pnpm Monorepos (Workspace + Turborepo)\n\npnpm workspace monorepos use symlinks between packages. A naive multi-stage Dockerfile that copies packages individually breaks module resolution at runtime (`ERR_MODULE_NOT_FOUND`).\n\n**Working pattern** — use `pnpm deploy` to flatten the workspace:\n\n```dockerfile\n# Build: install, build shared packages, then deploy to flat bundle\nRUN pnpm install --frozen-lockfile\nRUN pnpm --filter @scope\u002Fshared-pkg build\nRUN pnpm --filter .\u002Fapps\u002Fbackend\u002F --legacy deploy .\u002Fdeployment\u002F\n\n# Run: just copy the flat deployment bundle\nCOPY --from=builder \u002Fapp\u002Fdeployment .\u002Fdeployment\n```\n\n**Pitfall: `--prod` strips runtime deps.** `pnpm deploy --prod` excludes `devDependencies`, but some apps list runtime-critical packages (like `express`) under `devDependencies`. If the app crashes with `Cannot find package 'express'`, remove `--prod` from the deploy command. The project's own `vercel.json` uses `--prod` but pairs it with a `vercel-preparation.sh` script that does additional setup — Docker doesn't have that.\n\n**Pitfall: symlinks don't survive COPY.** Never try to manually copy `packages\u002Fcore\u002Fbuild` + `node_modules` to the runner stage. pnpm's `node_modules\u002F.pnpm` structure uses symlinks that Docker COPY doesn't preserve correctly. Always use `pnpm deploy` to produce a flat, self-contained bundle.\n\n## Keycloak 26+ Pitfalls\n\nSee `references\u002Fkeycloak-pitfalls.md` for the full list. Critical ones:\n\n- `KC_PROXY=edge` is DEPRECATED → use `KC_PROXY_HEADERS=xforwarded`\n- `KEYCLOAK_ADMIN` is DEPRECATED → use `KC_BOOTSTRAP_ADMIN_USERNAME`\n- Bootstrap admin is ALWAYS temporary → create permanent admin via Admin Console\n- `KC_CACHE=local` is REQUIRED for single-node (JGroups clustering timeout without it)\n- Realm `sslRequired` must be `none` when behind Cloudflare (Nginx sends HTTP)\n- Docker bind mount creates a DIRECTORY if the source file doesn't exist → create the file BEFORE starting the container\n\n## Spring Cloud Gateway + Eureka Pattern\n\n### Gateway compose environment (verified)\n\n```yaml\nenvironment:\n  SERVER_PORT: \"8000\"\n  MANAGEMENT_SERVER_PORT: \"8000\"    # Merge actuator onto main port\n  SPRING_PROFILES_ACTIVE: prod\n  # JWT — realm name must match, not \"flowerogate\" or other legacy names\n  JWT_ISSUER_URI: https:\u002F\u002Fauth.panomete.com\u002Frealms\u002Fpanomete\n  JWT_JWK_SET_URI: https:\u002F\u002Fauth.panomete.com\u002Frealms\u002Fpanomete\u002Fprotocol\u002Fopenid-connect\u002Fcerts\n  # Valkey — container name + password (not host.docker.internal)\n  REDIS_HOST: local-valkey\n  REDIS_PORT: \"6379\"\n  REDIS_PASSWORD: ${VALKEY_PASSWORD}\n  REDIS_SSL: \"false\"\n  # Eureka — enable + correct URL (not eureka:8761)\n  EUREKA_CLIENT_ENABLED: \"true\"\n  EUREKA_URI: http:\u002F\u002Fflowero-discover:8999\u002Feureka\n  # Disable features not yet needed\n  SPRING_AUTOCONFIGURE_EXCLUDE: org.springframework.cloud.circuitbreaker.resilience4j.Resilience4JAutoConfiguration\n  OTLP_ENABLED: \"false\"\n  # SecurityConfig requires this — no default in code\n  APP_POST_LOGIN_REDIRECT_URL: ${POST_LOGIN_REDIRECT_URL}\n  KEYCLOAK_GATEWAY_SECRET: ${KEYCLOAK_GATEWAY_SECRET}\n```\n\n### Gateway Dockerfile (Alpine — uses wget for healthcheck)\n\n```dockerfile\nFROM eclipse-temurin:25-jdk-alpine AS builder\nWORKDIR \u002Fapp\nCOPY gradlew gradlew.bat settings.gradle build.gradle .\u002F\nCOPY gradle\u002F gradle\u002F\nRUN chmod +x gradlew && .\u002Fgradlew dependencies --no-daemon 2>\u002Fdev\u002Fnull || true\nCOPY src\u002F src\u002F\nRUN .\u002Fgradlew bootJar --no-daemon -x test\n\nFROM eclipse-temurin:25-jre-alpine\nRUN addgroup -S flowero && adduser -S flowero -G flowero\nWORKDIR \u002Fapp\nCOPY --from=builder \u002Fapp\u002Fbuild\u002Flibs\u002F*.jar app.jar\nEXPOSE 8000\nUSER flowero\n# Alpine has wget via busybox — no curl install needed\nHEALTHCHECK --interval=30s --timeout=5s --retries=3 \\\n  CMD wget -qO- http:\u002F\u002Flocalhost:8000\u002Factuator\u002Fhealth\u002Fliveness || exit 1\nENTRYPOINT [\"java\", \"-XX:+UseZGC\", \"-XX:MaxRAMPercentage=75.0\", \"-jar\", \"app.jar\"]\n```\n\n**Key difference from Ubuntu-based images:** Alpine (`jre-alpine`) has `wget` via\nbusybox but NOT `curl`. Ubuntu (`jre-noble`) has neither. Use `wget` for Alpine\nhealthchecks, or `apt-get install curl` for Ubuntu.\n\n### Gateway routing (application.yaml)\n\n```yaml\nspring:\n  cloud:\n    gateway:\n      routes:\n        - id: blog\n          uri: lb:\u002F\u002Fcute-gufo              # Eureka service name\n          predicates: [Path=\u002Fapi\u002Fblog\u002F**]\n          filters:\n            - StripPrefix=1                # \u002Fapi\u002Fblog\u002Fposts → \u002Fposts\n            - RemoveRequestHeader=Authorization\n            - name: RequestRateLimiter\n              args:\n                key-resolver: \"#{@ipKeyResolver}\"\n                redis-rate-limiter.replenishRate: 100\n                redis-rate-limiter.burstCapacity: 200\n        # ... more routes\n```\n\n### Gateway prerequisites checklist\n\nBefore deploying Gate, verify:\n- [ ] Guard healthy (JWKS endpoint reachable)\n- [ ] Discover healthy (Eureka registry available)\n- [ ] Valkey healthy (password known)\n- [ ] `.env` has: `VALKEY_PASSWORD`, `KEYCLOAK_GATEWAY_SECRET`, `POST_LOGIN_REDIRECT_URL`\n- [ ] `flowero-gateway` OAuth2 client registered in Keycloak `panomete` realm\n- [ ] Nginx block for `api.panomete.com` created\n- [ ] Port 8000 free\n\n### Gateway 401 behavior\n\n- **Browser request** (`Accept: text\u002Fhtml`) → 302 redirect to Keycloak login\n- **API request** (no token) → 401 JSON body\n- **API request** (valid JWT) → route to backend via Eureka → 502 if no backend registered (expected in Phase 1)\n\n### Post-login redirect URL\n\n`SecurityConfig` uses `@Value(\"${app.post-login-redirect-url}\")` with NO default.\nIf `POST_LOGIN_REDIRECT_URL` is not set in `.env`, the app crashes on startup.\nAlways set it before deploying Gate.\n\n## Repository Architecture\n\n**NOT a monorepo.** Each service has its own GitHub repo. CI\u002FCD workflows live inside each repo (not a shared `.github\u002Fworkflows\u002F` at root).\n\n```\noat431\u002Fflowero-guard      → .github\u002Fworkflows\u002F{ci,deploy}.yml\noat431\u002Fflowero-discovery  → .github\u002Fworkflows\u002F{ci,deploy}.yml\noat431\u002Fflowero-gateway    → .github\u002Fworkflows\u002F{ci,deploy}.yml\n```\n\nEach repo gets its own `TS_AUTH_KEY` and `HOMELAB_SSH_KEY` secrets (duplicated across repos).\n\n## High-Risk Boundary (User Handles)\n\nThe user handles these manually — never automate:\n- Passwords, API keys, SSH keys\n- `.env` file edits with secrets\n- UFW firewall rules\n- Portainer admin password\n- rclone OAuth re-authentication\n\nEverything else (container lifecycle, DB provisioning, Nginx configs, realm JSON, CI\u002FCD workflows) the assistant can handle.\n\n## Obsidian Documentation Pattern\n\nOperational knowledge lives in Obsidian at `F:\\obsidian_note\\oralita_md\\`, NOT in the spec folder and NOT on the homelab server. **The vault gets reorganized between sessions — always verify the current structure with a quick `ls` before writing docs** (paths below changed once already: `home-lab-installation\u002F` → `home-lab\u002F`, and Quick Note guides moved into `home-lab\u002Fobsidian-livesync\u002F`).\n\nStructure (as of 2026-08):\n\n```\noralita_md\u002F\n├── home-lab\u002F\n│   ├── audit\u002F                 # dated audit\u002Fdeployment logs\n│   ├── database_component\u002F    # PostgreSQL, Valkey, MongoDB, CouchDB\n│   ├── microservice_component\u002F# Keycloak, Discovery, Gateway\n│   ├── deployment\u002F            # CI\u002FCD, backup automation\n│   ├── essential_component\u002F   # Nginx, Cloudflare, Tailscale\n│   ├── miscellaneous\u002F         # db-network-integration-guide.md, etc.\n│   └── obsidian-livesync\u002F     # 01-full-setup, 02-connection-guide, 03-couchdb-user-management\n└── Quick Note\u002F                # user's scratch notes (Home Lab App.md port table lives here)\n```\n\nEach service gets a full-detail self-contained note with: What & Why, Setup, Deploy, Access, Common Commands, Configuration, Troubleshooting, Project Structure, Related.\n\n**Doc handoff convention:** the user explicitly corrected server-side doc placement — write guides to the LOCAL Obsidian vault, never to `~\u002Fapplication\u002Fmisc\u002F` on the homelab (a stray copy landed there once and was ignored).\n\n## Deployment Order\n\nFor Spring Cloud microservice platforms, deploy in dependency order:\n1. Databases (PostgreSQL, Valkey) — already running\n2. Identity provider (Keycloak) — needs DB only\n3. Service registry (Eureka) — standalone, no deps\n4. API Gateway — depends on all above\n\n**Pitfall:** Deploying Gateway before Discovery forces hardcoded routes that must\nbe rewritten later. Always deploy Discovery first — it's standalone and takes\n30 seconds to build.\n\n## Observability Stack Pattern\n\nSee `references\u002Fobservability-stack.md` for the full Prometheus + Grafana + Loki + Promtail deployment.\n\n### Quick Summary\n\nDeploy as a **separate compose file** (`docker-compose.observability.yml`) to keep observability isolated from platform services.\n\n| Service | Port | Domain | Purpose |\n|---------|:----:|--------|---------|\n| Prometheus | 9090 | Internal only | Scrape metrics from all services |\n| Grafana | 3000 | `grafana.panomete.com` | Dashboards + alerts |\n| Loki | 3100 | Internal | Log aggregation |\n| Promtail | — | Internal | Tail container logs → Loki |\n| Uptime Kuma | 3001 | `status.panomete.com` | Uptime monitoring |\n\n### Key Pitfalls\n\n1. **Prometheus target \"down\" with 404** — Spring Boot needs `micrometer-registry-prometheus` dependency AND `prometheus` in `management.endpoints.web.exposure.include`. Without the dependency, the endpoint returns 404.\n\n2. **Keycloak metrics on management port 9000, NOT main port 8080** — With `KC_METRICS_ENABLED=true`, metrics are at `http:\u002F\u002Fcontainer:9000\u002Fmetrics`. Prometheus scrape target must use port 9000.\n\n3. **Promtail `docker_sd_configs` with filters fails** — Use filesystem log reading instead:\n   ```yaml\n   scrape_configs:\n     - job_name: docker-logs\n       static_configs:\n         - targets: [localhost]\n           labels: {job: docker-logs, __path__: \u002Fvar\u002Flib\u002Fdocker\u002Fcontainers\u002F*\u002F*.log}\n       pipeline_stages:\n         - docker: {}\n   ```\n\n4. **Loki ingestion rate limit exceeded** — Default is 4MB\u002Fs. Increase in loki-config.yml:\n   ```yaml\n   limits_config:\n     ingestion_rate_mb: 16\n     ingestion_burst_size_mb: 32\n   ```\n\n5. **Grafana alert rules \"data source not found\"** — Alert rules need the actual Prometheus datasource UID, not a placeholder. Get it via:\n   ```bash\n   curl -sf -u admin:$GRAFANA_ADMIN_PASSWORD http:\u002F\u002Flocalhost:3000\u002Fapi\u002Fdatasources | python3 -c \"import sys,json; [print(d['uid']) for d in json.load(sys.stdin) if d['type']=='prometheus']\"\n   ```\n\n6. **Keycloak healthcheck — no tools in image** — Keycloak image has no curl, wget, or even `which`. Use TCP check:\n   ```yaml\n   healthcheck:\n     test: [\"CMD-SHELL\", \"timeout 5 bash -c '\u003C\u002Fdev\u002Ftcp\u002Flocalhost\u002F8080' || exit 1\"]\n   ```\n\n7. **Uptime Kuma healthcheck — `wget` not found** — The `louislam\u002Fuptime-kuma` image does NOT have `wget` (unlike Prometheus\u002FGrafana\u002FLoki which do). The default image healthcheck uses `wget` and permanently fails with `exec: \"wget\": executable file not found in $PATH`. Fix by overriding in compose:\n   ```yaml\n   uptime-kuma:\n     healthcheck:\n       test: [\"CMD-SHELL\", \"curl -sf http:\u002F\u002Flocalhost:3001 > \u002Fdev\u002Fnull\"]\n       interval: 15s\n       timeout: 5s\n       retries: 3\n   ```\n   `curl` IS available in the Uptime Kuma container. Container will go from `unhealthy` to `healthy` within one check interval after recreating.\n\n### Prometheus Scrape Config (verified)\n\n```yaml\nscrape_configs:\n  - job_name: \"flowero-gate\"\n    metrics_path: \"\u002Factuator\u002Fprometheus\"\n    static_configs:\n      - targets: [\"flowero-gate:8000\"]\n  - job_name: \"flowero-discover\"\n    metrics_path: \"\u002Factuator\u002Fprometheus\"\n    static_configs:\n      - targets: [\"flowero-discover:8999\"]\n  - job_name: \"flowero-guard\"\n    metrics_path: \"\u002Fmetrics\"\n    static_configs:\n      - targets: [\"flowero-guard:9000\"]  # Management port, NOT 8080\n  - job_name: \"prometheus\"\n    static_configs:\n      - targets: [\"localhost:9090\"]\n```\n\n## Backup Automation Pattern\n\nSee `references\u002Fbackup-automation.md` for the full pg_dumpall + rclone + cron pattern.\n\n### Quick Summary\n\nTwo scripts on the server at `\u002Fhome\u002Fflowero\u002Fscripts\u002F`:\n\n| Script | Schedule | What | Retention |\n|--------|----------|------|-----------|\n| `backup-db.sh` | Daily 3:00 AM | `pg_dumpall` → gzip → rclone → OneDrive | 7 days local |\n| `backup-volumes.sh` | Weekly Sunday 4:00 AM | Docker volumes → tar → gzip → rclone | 4 weeks local |\n\n### Docker Volume Naming Pitfall\n\nWhen using named volumes in Docker Compose, the actual volume name on disk is NOT the compose key. Always check with `docker volume ls` before writing backup scripts. The backup script's `VOLUMES` array must use the actual volume names.\n\n## Service Consolidation Pattern (Shared Databases)\n\nWhen a third-party application ships its own bundled database (Valkey, Redis, PostgreSQL), replace it with the shared instance on `db-network`. This reduces resource usage and centralizes data management.\n\n### Steps\n\n1. **Remove the bundled service** from the application's `docker-compose.yml`\n2. **Remove its volume** (the shared instance has its own)\n3. **Add the application to `db-network`**:\n   ```yaml\n   services:\n     app:\n       networks:\n         - db-network\n   networks:\n     db-network:\n       external: true\n   ```\n4. **Configure the connection** — point the app at the shared service. If the shared service has a DNS alias matching the old bundled service name (e.g., `local-valkey` has alias `valkey`), the app may work with zero config changes.\n5. **Handle authentication** — shared services may have passwords the bundled ones didn't. Add credentials via `settings.yml`, `.env`, or environment variables.\n\n### DNS Alias Trick\n\n`local-valkey` on `db-network` has aliases `[\"local-valkey\", \"valkey\"]`. Applications that reference `valkey:6379` resolve to the shared instance automatically — no config rename needed.\n\n### Verifying Connection\n\n```bash\n# Check container is on the right network\ndocker ps --filter \"name=APP\" --format \"{{.Networks}}\"\n\n# Check logs for connection errors\ndocker logs APP 2>&1 | grep -i \"redis\\|valkey\\|connect\\|refused\"\n\n# HTTP health check\ncurl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" http:\u002F\u002F127.0.0.1:PORT\u002F\n```\n\n### Pitfalls\n\n- **Remove `depends_on` healthcheck conditions when deleting bundled DBs** — compose files often declare `depends_on: {bundled-db: {condition: service_healthy}}`. Once the bundled service is removed, these blocks break `docker compose up` with \"dependency failed to start\". Delete them along with the service (external services can't satisfy `depends_on` anyway).\n- **Multi-service apps may reference the DB from several services** — e.g., Penpot's backend AND exporter both declare `PENPOT_REDIS_URI`. Grep the whole compose for DB connection env vars and update every occurrence, not just the main app service.\n- **File permissions (UID mismatch):** Container config files are often owned by the container's internal UID (e.g., SearXNG uses UID 977). Regular shell redirection (`>`) fails with \"Permission denied\". Use `sudo tee` instead:\n  ```bash\n  sudo tee \u002Fpath\u002Fto\u002Fcontainer-owned\u002Fconfig.yml > \u002Fdev\u002Fnull \u003C\u003C 'EOF'\n  content here\n  EOF\n  ```\n- **Env var overrides config file:** Some apps (like SearXNG) let environment variables override config files. If you set the connection in `settings.yml` AND as an env var, the env var wins — and may use the wrong key name (causing deprecation warnings). Pick one approach, not both.\n- **Network name collision:** If the bundled service had the same container name as a shared service alias, Docker DNS conflicts. Rename the bundled service before removing it, or just remove it entirely.\n- **SQLite-based apps cannot integrate:** Some apps (e.g., ByteStash uses `better-sqlite3`) are file-based with no network database support. Check `package.json` or container dependencies before attempting migration.\n- **Port conflicts:** Always check occupied ports (`ss -tlnp | grep 127.0.0.1:`) before deploying. Common conflicts: port 3000 (Grafana), port 80 (system services).\n\n### Documentation Workflow\n\nWhen consolidating services, create a running guide in the user's **Obsidian vault** (`F:\\obsidian_note\\oralita_md\\home-lab\\miscellaneous\\` — verify the path first; the vault gets reorganized). The user reviews and edits docs locally in Obsidian — server-side docs get ignored.\n\nAlso update the port reference in `F:\\obsidian_note\\oralita_md\\Quick Note\\Home Lab App.md` when ports change.\n\n### Sequential Review Preference\n\nProcess one service at a time. After each migration:\n1. Start the container\n2. Verify it's healthy (HTTP check, logs, DB connectivity)\n3. Present the results to the user\n4. **Wait for user approval** before proceeding to the next service\n\nDo NOT batch all migrations in one shot — the user reviews each change before moving on.\n\n### Port Conflicts\n\nPort 3000 is commonly taken by Grafana. Always check before deploying:\n\n```bash\nss -tlnp | grep -E \"127\\.0\\.0\\.1:\" | awk '{print $4}' | sort -t: -k2 -n\n```\n\nKnown port allocations on the Panomete homelab:\n\n| Port | Service |\n|------|---------|\n| 3000 | Grafana |\n| 3001 | Uptime Kuma |\n| 5432 | PostgreSQL |\n| 5984 | CouchDB |\n| 6379 | Valkey |\n| 7000 | AdGuard |\n| 7004 | SearXNG |\n| 7005 | Infisical |\n| 7006 | OTS |\n| 7007 | StirlingPDF |\n| 7008 | ByteStash |\n| 7009 | Penpot |\n| 7010 | GitHub Stats Extended |\n| 8000 | Spring Cloud Gateway |\n| 8001 | Keycloak |\n| 8999 | Eureka |\n| 9000 | Portainer |\n| 9001 | Penpot (alt, direct access) |\n| 9090 | Prometheus |\n| 27017 | MongoDB |\n\n## CI\u002FCD Migration: build → image (Critical Pattern)\n\nOnce CI\u002FCD is set up, the server must ONLY pull pre-built images from GHCR. Never build on the server.\n\n**Wrong (building on server):**\n```yaml\nflowero-discover:\n  build:\n    context: .\u002Fflowerodiscovery\n    dockerfile: Dockerfile\n```\n\n**Correct (pulling from GHCR):**\n```yaml\nflowero-discover:\n  image: ghcr.io\u002Foat431\u002Fflowero-discovery:latest\n```\n\n**Migration steps:**\n1. Update compose: replace `build:` with `image: ghcr.io\u002F...`\n2. Remove local repo copies from server (`rm -rf ~\u002Fplatform\u002Fflowerodiscovery`, etc.)\n3. Server layout becomes: `.env`, `docker-compose.platform.yml`, `docker-compose.observability.yml`, `observability\u002F` config dir\n4. Verify: `docker ps` should show GHCR image names, not local build names\n\n**Why:** The user explicitly corrected this — \"why you modified the local repo in the first place, i thought we just setup ci\u002Fcd\". Source code lives in GitHub. CI builds + pushes to GHCR. Server pulls. No source code on server.\n\n## CI\u002FCD for Tailscale-Behind Servers\n\nWhen the homelab resolves to a Tailscale IP, GitHub Actions can't SSH directly.\nUse `tailscale\u002Fgithub-action` to bridge the runner onto the Tailnet temporarily,\nthen deploy via SSH. Deploy should be manual (`workflow_dispatch`) for single-dev\nsafety. See `references\u002Ftailscale-cicd-bridge.md` for the full workflow pattern.\n\n### CI Pitfalls (Gradle projects)\n\n- **`gradlew: Permission denied`** — Windows doesn't preserve Unix execute bits.\n  Fix: `git update-index --chmod=+x gradlew` AND add `chmod +x gradlew` step in CI.\n- **`Task 'checkstyleMain' not found`** — Only run if Checkstyle plugin is in `build.gradle`.\n  If the project doesn't have it, remove the lint step entirely.\n- **`.dockerignore` excluding needed files** — The Guard CI failed because\n  `panomete-realm.json` was listed in `.dockerignore`. Always verify `.dockerignore`\n  doesn't exclude files referenced by `COPY` in the Dockerfile.\n\n### GHCR Package Visibility Pitfall\n\nGHCR package visibility is **separate from repo visibility**. Making a repo public\ndoes NOT make its container image public. You must also set the package to public:\n\n1. Go to `https:\u002F\u002Fgithub.com\u002F\u003Cowner>\u002F\u003Crepo>\u002Fpackages`\n2. Click the container package → **Package settings**\n3. Under **Danger Zone** → **Change visibility** → **Public**\n\nOr link the package to the repo (inherits visibility). Without this,\n`docker pull` from a public repo still returns `unauthorized`.\n\n### Docker Volume Naming\n\nWhen using named volumes in Docker Compose, the actual volume name on disk is\n`{project}_{service}_{volume}`, NOT just the volume key. Always check with\n`docker volume ls` before writing backup scripts. Example:\n- Compose key: `postgres_data` in project `platform`\n- Actual name: `platform_postgres_data` or `postgres_postgres_data` (depends on compose setup)\n\n## Backup Automation Pattern\n\nSee `references\u002Fbackup-automation.md` for the full pg_dumpall + rclone + cron pattern.\n\n## Related\n\n- References: `references\u002Fkeycloak-pitfalls.md`, `references\u002Fdocker-compose-patterns.md`, `references\u002Ftailscale-cicd-bridge.md`, `references\u002Fsearxng.md`, `references\u002Finfisical.md`, `references\u002Fots.md`, `references\u002Fbytestash.md`, `references\u002Fpenpot.md`, `references\u002Fcouchdb.md`, `references\u002Fadguard-dns-filtering.md`, `references\u002Fobservability-stack.md`, `references\u002Fbackup-automation.md`; scripts `scripts\u002Fadguard-qlog-stats.py` (AdGuard query-log stats), `scripts\u002Fverify-router-dns.sh` (verify router-side IPv6 DNS hand-out)\n- See also: `homelab-infra-audit` for server auditing before deployment\n",{"name":5,"triggers":24,"description":7},[25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41],"deploying a new service to the homelab","configuring Keycloak \u002F OAuth2 \u002F OIDC","setting up Spring Cloud Gateway or Eureka","Docker Compose for multi-service platforms","Nginx reverse proxy behind Cloudflare","troubleshooting container healthchecks or networking","consolidating bundled databases into shared infrastructure","connecting containers to db-network","Infisical deployment or configuration","OTS \u002F one-time-secret deployment","migrating bundled Redis\u002FValkey\u002FPostgreSQL to shared infrastructure","SearXNG configuration or rate limiting","MCP server pointing at homelab services","Tailscale routing for local development tools","AdGuard Home DNS filtering \u002F blocklist auditing","DNS filtering coverage — devices bypassing the filter","router-side IPv6 DNS (RDNSS \u002F DHCPv6) — make the router hand out the homelab DNS, then verify it","main","2988869c5a7d7be2225c9ea7a54a4a56a59b0d45","workflow\u002Fdevops\u002Fhomelab-microservice-deployment",0,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup","https:\u002F\u002Fgithub.com\u002Foat431","https:\u002F\u002Fpanomete.com\u002F","Hope and Luck is not a part of the strategy, \r\n\r\nSoftskill produces plan Hard skill produce product.","2026-11-03 00:43:08",{"source":8,"skillName":5,"commands":52,"prompt":62,"zipAvailable":63},[53,56,59],{"tool":54,"command":55},"npx","npx skills add oat431\u002Fhermes_config_backup --skill homelab-microservice-deployment",{"tool":57,"command":58},"bunx","bunx skills add oat431\u002Fhermes_config_backup --skill homelab-microservice-deployment",{"tool":60,"command":61},"pnpm","pnpm dlx skills add oat431\u002Fhermes_config_backup --skill homelab-microservice-deployment","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002FSKILL.md",true,[65,71,75,79,83,87,91,95,99,103,107,111,115,119,125],{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"references\u002Fadguard-dns-filtering.md","REFERENCE","参考文档",8359,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fadguard-dns-filtering.md",{"path":72,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":73,"sourceHtmlUrl":74},"references\u002Fbackup-automation.md",5356,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fbackup-automation.md",{"path":76,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":77,"sourceHtmlUrl":78},"references\u002Fbytestash.md",1928,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fbytestash.md",{"path":80,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":81,"sourceHtmlUrl":82},"references\u002Fcouchdb.md",10983,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fcouchdb.md",{"path":84,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":85,"sourceHtmlUrl":86},"references\u002Fdocker-compose-patterns.md",3233,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fdocker-compose-patterns.md",{"path":88,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":89,"sourceHtmlUrl":90},"references\u002Fgithub-stats-extended.md",6635,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fgithub-stats-extended.md",{"path":92,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":93,"sourceHtmlUrl":94},"references\u002Finfisical.md",2913,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Finfisical.md",{"path":96,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":97,"sourceHtmlUrl":98},"references\u002Fkeycloak-pitfalls.md",4866,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fkeycloak-pitfalls.md",{"path":100,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":101,"sourceHtmlUrl":102},"references\u002Fobservability-stack.md",9817,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fobservability-stack.md",{"path":104,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":105,"sourceHtmlUrl":106},"references\u002Fots.md",1655,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fots.md",{"path":108,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":109,"sourceHtmlUrl":110},"references\u002Fpenpot.md",5446,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fpenpot.md",{"path":112,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":113,"sourceHtmlUrl":114},"references\u002Fsearxng.md",13235,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Fsearxng.md",{"path":116,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":117,"sourceHtmlUrl":118},"references\u002Ftailscale-cicd-bridge.md",4195,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Freferences\u002Ftailscale-cicd-bridge.md",{"path":120,"resourceType":121,"resourceTypeDesc":122,"sizeBytes":123,"sourceHtmlUrl":124},"scripts\u002Fadguard-qlog-stats.py","SCRIPT","脚本",1996,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Fscripts\u002Fadguard-qlog-stats.py",{"path":126,"resourceType":121,"resourceTypeDesc":122,"sizeBytes":85,"sourceHtmlUrl":127},"scripts\u002Fverify-router-dns.sh","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fdevops\u002Fhomelab-microservice-deployment\u002Fscripts\u002Fverify-router-dns.sh",[129,137,144,151,158,165],{"skillNo":130,"name":131,"nameZh":6,"description":132,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":133,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":136,"publishedAt":19},"01M7AYZ2JY3NFKAPQSC39N5F3F","project-spec-docs","Fill\u002Fupdate project_spec docs (templates, MM handoffs).","MIT","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fproject-spec-docs\u002FSKILL.md",[],[],{"skillNo":138,"name":139,"nameZh":6,"description":140,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":133,"licenseFileUrl":6,"sourceHtmlUrl":141,"categories":142,"keywords":143,"publishedAt":19},"01M7AYYRZPE7HQX7778558X52S","penpot-mcp","Penpot MCP: connect, verify, and fix streaming via nginx.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fpenpot-mcp\u002FSKILL.md",[],[],{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":150,"publishedAt":19},"01M7AYYNKP61CN2281VMJDXMB5","nginx-streaming-proxy","Configure nginx for streaming endpoints (SSE, MCP, WS).","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fnginx-streaming-proxy\u002FSKILL.md",[],[],{"skillNo":152,"name":153,"nameZh":6,"description":154,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":133,"licenseFileUrl":6,"sourceHtmlUrl":155,"categories":156,"keywords":157,"publishedAt":19},"01M7AYYJP0MDV1D3WVEEZB3SGA","branding-ci","Use for brand\u002FCI identity, style guides, and brand docs.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fbranding-ci\u002FSKILL.md",[],[],{"skillNo":159,"name":160,"nameZh":6,"description":161,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":162,"categories":163,"keywords":164,"publishedAt":19},"01M7AYYGCXVT97DDTCM0ZJT6F6","yuanbao","Yuanbao (元宝) groups: @mention users, query info\u002Fmembers.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fshared-skills\u002Fyuanbao\u002FSKILL.md",[],[],{"skillNo":166,"name":167,"nameZh":6,"description":168,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":169,"categories":170,"keywords":171,"publishedAt":19},"01M7AYYC2Y65NMGMD4GQXAZMMS","slidev","Create and present web-based slidedecks for developers using Slidev with Markdown, Vue components, code highlighting, animations, and interactive features. Use when building technical presentations, conference talks, code walkthroughs, teaching materials, or developer decks.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fshared-skills\u002Fslidev\u002FSKILL.md",[],[]]