continuum-backend/scripts/dev-env.sh

#!/usr/bin/env bash
# Brings the local dependency stack (Valkey + Postgres/pgvector) up and down,
# and applies the Drizzle schema to it.
#
# Two backends behind one interface, chosen automatically:
#
#   docker  docker-compose.yml - Valkey 8 + pgvector/pgvector:pg17. What CI
#           and a normal dev machine should use; closest to production.
#   native  apt packages started directly - postgresql-16 + pgvector + Valkey
#           7.2 from Ubuntu's own repositories.
#
# The native path exists because Docker provably cannot run in some sandboxes,
# including the one this was written in: dockerd needs CAP_NET_ADMIN to build
# its bridge, and even with `--bridge=none --iptables=false
# --storage-driver=vfs` it cannot unpack an image, because registering a layer
# needs user namespaces. Both are host-level grants, so there is nothing a
# script can do about it. Rather than leave the database untestable there, this
# falls back to running the same two services from apt. Set CONTINUUM_DEV_BACKEND
# to force one.
#
# Usage:
#   scripts/dev-env.sh up        start the services, wait until they answer
#   scripts/dev-env.sh down      stop them (keeps data)
#   scripts/dev-env.sh reset     stop them AND delete all data
#   scripts/dev-env.sh status    what is running, and which backend
#   scripts/dev-env.sh psql      open a shell on the database
#   scripts/dev-env.sh schema    apply the committed migrations (drizzle-kit migrate)
#   scripts/dev-env.sh push      reshape the db from src/db/schema, no migration
#   scripts/dev-env.sh env       print the DATABASE_URL/VALKEY_URL to use
set -euo pipefail

REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
cd "$REPO_ROOT"

PG_VERSION=16
PG_PORT=5432
VALKEY_PORT=6379
# Straight at Postgres, not pgCat on 6432: .env.example points at the pooler for
# production, and neither backend here runs one.
DEV_DATABASE_URL="postgres://continuum:continuum@localhost:${PG_PORT}/continuum"
DEV_VALKEY_URL="redis://localhost:${VALKEY_PORT}"

NATIVE_STATE="${XDG_RUNTIME_DIR:-/tmp}/continuum-dev-env"

# --- backend selection -----------------------------------------------------

# Docker is used when its daemon is already answering. Deliberately does not
# try to start dockerd: on a machine where it works there is a service manager
# doing that, and on a machine where it does not, starting it just produces a
# daemon that cannot unpack an image (see the header). Force either backend
# with CONTINUUM_DEV_BACKEND=docker|native.
docker_available() {
  command -v docker >/dev/null 2>&1 || return 1
  sudo -n docker info >/dev/null 2>&1
}

backend() {
  if [ -n "${CONTINUUM_DEV_BACKEND:-}" ]; then
    echo "$CONTINUUM_DEV_BACKEND"
  elif docker_available; then
    echo docker
  else
    echo native
  fi
}

# --- waiting ---------------------------------------------------------------

# Poll for readiness rather than sleeping a fixed amount: Postgres accepts
# connections before it will serve them, and a failure should report why.
wait_for_postgres() {
  for _ in $(seq 1 60); do
    if pg_isready -h localhost -p "$PG_PORT" -U continuum -d continuum >/dev/null 2>&1; then
      return 0
    fi
    sleep 0.5
  done
  echo "postgres did not become ready on port $PG_PORT" >&2
  return 1
}

wait_for_valkey() {
  for _ in $(seq 1 40); do
    if valkey-cli -p "$VALKEY_PORT" ping 2>/dev/null | grep -q PONG; then
      return 0
    fi
    sleep 0.5
  done
  echo "valkey did not answer on port $VALKEY_PORT" >&2
  return 1
}

# --- native backend --------------------------------------------------------

native_up() {
  mkdir -p "$NATIVE_STATE"

  if ! pg_isready -h localhost -p "$PG_PORT" >/dev/null 2>&1; then
    echo "==> starting postgresql-$PG_VERSION"
    # pg_ctlcluster rather than systemctl: there is no service manager here.
    sudo pg_ctlcluster "$PG_VERSION" main start
  fi
  wait_for_postgres_socket

  # Idempotent: both of these are no-ops on a second run.
  if ! sudo -u postgres psql -tAc "SELECT 1 FROM pg_roles WHERE rolname='continuum'" | grep -q 1; then
    echo "==> creating role 'continuum'"
    sudo -u postgres psql -qc "CREATE ROLE continuum LOGIN PASSWORD 'continuum' SUPERUSER" >/dev/null
  fi
  if ! sudo -u postgres psql -tAc "SELECT 1 FROM pg_database WHERE datname='continuum'" | grep -q 1; then
    echo "==> creating database 'continuum'"
    sudo -u postgres createdb -O continuum continuum
  fi
  # frame_embeddings needs it, and it is why the compose file uses the
  # pgvector image rather than plain postgres.
  sudo -u postgres psql -d continuum -qc "CREATE EXTENSION IF NOT EXISTS vector" >/dev/null

  if ! valkey-cli -p "$VALKEY_PORT" ping >/dev/null 2>&1; then
    echo "==> starting valkey"
    valkey-server --port "$VALKEY_PORT" --daemonize yes \
      --dir "$NATIVE_STATE" --appendonly yes \
      --logfile "$NATIVE_STATE/valkey.log"
  fi
  wait_for_valkey
}

# Postgres has to be up before the role exists, so this waits on the server
# rather than on a successful `continuum` login.
wait_for_postgres_socket() {
  for _ in $(seq 1 60); do
    if sudo -u postgres pg_isready -q 2>/dev/null; then return 0; fi
    sleep 0.5
  done
  echo "postgres did not start" >&2
  return 1
}

native_down() {
  if valkey-cli -p "$VALKEY_PORT" ping >/dev/null 2>&1; then
    echo "==> stopping valkey"
    valkey-cli -p "$VALKEY_PORT" shutdown nosave 2>/dev/null || true
  fi
  if sudo -u postgres pg_isready -q 2>/dev/null; then
    echo "==> stopping postgresql-$PG_VERSION"
    sudo pg_ctlcluster "$PG_VERSION" main stop || true
  fi
}

native_reset() {
  native_down
  echo "==> dropping the continuum database and valkey's data"
  sudo pg_ctlcluster "$PG_VERSION" main start
  wait_for_postgres_socket
  sudo -u postgres dropdb --if-exists continuum
  sudo pg_ctlcluster "$PG_VERSION" main stop || true
  rm -rf "$NATIVE_STATE"
}

native_status() {
  echo "backend: native (apt packages, no container runtime)"
  if sudo -u postgres pg_isready -q 2>/dev/null; then
    echo "  postgres: up on :$PG_PORT ($(sudo -u postgres psql -tAc 'SELECT version()' | cut -d, -f1))"
  else
    echo "  postgres: down"
  fi
  if valkey-cli -p "$VALKEY_PORT" ping >/dev/null 2>&1; then
    echo "  valkey:   up on :$VALKEY_PORT"
  else
    echo "  valkey:   down"
  fi
}

# --- docker backend --------------------------------------------------------

docker_up() {
  sudo docker compose up -d
  echo "==> waiting for containers to report healthy"
  for _ in $(seq 1 60); do
    local unhealthy
    unhealthy="$(sudo docker compose ps --format '{{.Service}} {{.Health}}' \
      | awk '$2 != "healthy" { print $1 }' || true)"
    [ -z "$unhealthy" ] && return 0
    sleep 1
  done
  echo "timed out: $(sudo docker compose ps --format '{{.Service}} {{.Health}}')" >&2
  return 1
}

# --- commands --------------------------------------------------------------

cmd_up() {
  case "$(backend)" in
    docker) echo "==> backend: docker compose"; docker_up ;;
    native) echo "==> backend: native (docker is unavailable here - see this script's header)"; native_up ;;
  esac
  echo
  cmd_env
  echo
  echo "Next: scripts/dev-env.sh schema"
}

cmd_down()   { case "$(backend)" in docker) sudo docker compose down ;; native) native_down ;; esac; }
cmd_reset()  { case "$(backend)" in docker) sudo docker compose down -v ;; native) native_reset ;; esac; }
cmd_status() { case "$(backend)" in docker) echo "backend: docker"; sudo docker compose ps ;; native) native_status ;; esac; }
cmd_psql()   { PGPASSWORD=continuum psql -h localhost -p "$PG_PORT" -U continuum -d continuum "$@"; }
cmd_env()    { echo "DATABASE_URL=$DEV_DATABASE_URL"; echo "VALKEY_URL=$DEV_VALKEY_URL"; }

# Applies the committed migrations in drizzle/, the same way a deploy would.
# This is the default because it is the path that has to work in production -
# testing `push` locally and `migrate` in production tests two different things.
cmd_schema() {
  echo "==> applying committed migrations to $DEV_DATABASE_URL"
  DATABASE_URL="$DEV_DATABASE_URL" bunx drizzle-kit migrate
}

# Diffs the live database against src/db/schema and reshapes it in place, with
# no migration file. Faster while iterating on the schema, and destructive: it
# will drop a column to make the database match. Local only - and whatever it
# does is not recorded anywhere, so run `bun run db:generate` before committing
# a schema change made this way.
cmd_push() {
  echo "==> pushing src/db/schema straight onto $DEV_DATABASE_URL"
  DATABASE_URL="$DEV_DATABASE_URL" bunx drizzle-kit push --force
}

case "${1:-}" in
  up)     cmd_up ;;
  down)   cmd_down ;;
  reset)  cmd_reset ;;
  status) cmd_status ;;
  psql)   shift; cmd_psql "$@" ;;
  schema) cmd_schema ;;
  push)   cmd_push ;;
  env)    cmd_env ;;
  *)      awk 'NR>1 && /^#/ { sub(/^# ?/, ""); print; next } NR>1 { exit }' "${BASH_SOURCE[0]}"; exit 1 ;;
esac