You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
feat: add Helm chart and Dockerfile for deploying Latitude to Kubernetes
Adds a production-ready Helm chart under charts/latitude/ that deploys all
four services: web (TanStack Start SSR), api (Hono HTTP), ingest (telemetry
ingestion), and workers (Redpanda background jobs).
Includes a multi-stage Dockerfile with per-service build targets (api, web,
ingest, workers, migrations) and a health endpoint for the web app.
Key design decisions:
- All four app services deployed with health probes, optional HPA/PDB/Ingress
- Secret management supports both inline values (dev/test) and an external
pre-existing Kubernetes Secret (production) via existingSecret
- ConfigMap and Secret cover all LAT_* env vars including Kafka/Redpanda,
object storage (S3), Weaviate, ClickHouse migration vars, and admin DB URL
- Pre-install/pre-upgrade migration Job runs Postgres (drizzle-kit),
ClickHouse (goose), and Weaviate migrations before app pods roll out
- Workers deployment uses extended terminationGracePeriodSeconds for
in-flight job completion
- Pod annotations include config/secret checksums to trigger rolling
restarts on configuration changes
- VITE_LAT_* vars documented as build-time (client bundle) with SSR
runtime fallback
Made-with: Cursor
0 commit comments