Skip to content

testing-instance.md

Latest commit

 

History

History
74 lines (60 loc) · 3.45 KB

testing-instance.md

File metadata and controls

74 lines (60 loc) · 3.45 KB
Raw

Local test instance (for agents verifying changes)

A ready-to-run, LLM-capable seeded database lives on this dev machine so you can launch the real app and verify a change instead of only running tests. Use it. Founders strongly prefer "I booted it and saw it work" over "typecheck passed".

What's seeded

Canonical seed data dir: ~/.local/share/hivekeep-testdata/ (self-contained: SQLite DB + persisted .encryption-key + workspaces/).

  • Admin user: admin@local.test / Password123! (onboarding complete)
  • Anthropic "Claude (subscription)" provider in cli mode → reads the host's ~/.claude/.credentials.json (Claude Max OAuth) at runtime. No secret is stored in the DB, so it's safe to copy around; it just works on this machine. Models are synced and the default LLM is set, so chat / tasks / tools work.
  • One regular Agent ("Tester") bound to the default model
  • 25 contacts, 12 vault secrets, 10 webhooks — enough to exercise the settings list screens (search / filter / pagination thresholds are 8 items)

Booting an isolated instance against it

Never boot against the canonical dir directly — tests mutate data. Copy it first. And never use the prod env (see warnings below).

# 1. Copy the seed to a throwaway dir
SRC=~/.local/share/hivekeep-testdata
DST=/tmp/hk-test-$$
cp -r "$SRC" "$DST"

# 2. Boot on a dedicated port with the env FULLY overridden (this shell inherits
#    PROD env vars — see below). Leave ENCRYPTION_KEY unset; the dir carries its own.
DB_PATH="$DST/hivekeep.db" HIVEKEEP_DATA_DIR="$DST" \
PORT=4178 PUBLIC_URL=http://localhost:4178 HIVEKEEP_PUBLIC_URL=http://localhost:4178 \
TRUSTED_ORIGINS=http://localhost:4178 HIVEKEEP_MODEL_REGISTRY=false \
NODE_OPTIONS=--max-old-space-size=4096 \
bun src/server/index.ts

Then log in at http://localhost:4178 with the admin creds above, or drive it with Playwright (chromium is installed). Stop it by its exact PID when done.

Regenerating / extending the seed

bun scripts/seed-test-db.ts          # top up what's missing (idempotent)
FRESH=1 bun scripts/seed-test-db.ts  # wipe + reseed from scratch
TESTDATA_DIR=/path bun scripts/seed-test-db.ts

scripts/seed-test-db.ts is the source of truth (survives schema migrations — just re-run it). Extend it there if you need more seeded data (channels, crons, mini-apps, etc.).

Critical gotchas (this machine)

  • This shell inherits PROD env vars from the user profile: DB_PATH=~/.local/share/kinbot/kinbot.db (the live 185 MB prod DB), HIVEKEEP_DATA_DIR=~/.local/share/kinbot, PORT=3000. A bare bun run start or bun run db:migrate targets PROD. Always override DB_PATH + HIVEKEEP_DATA_DIR + PORT inline. scripts/migrate.ts reads DB_PATH (not HIVEKEEP_DATA_DIR).
  • Prod runs live on port 3000. Never kill/pkill by port or pattern — kill your isolated server by its exact PID only.
  • Settings is a modal, not a route. Open it deterministically by navigating to /?email_connected=x (a URL hook), then click the section in the nav.
  • Playwright: use waitUntil: 'domcontentloaded', NOT 'networkidle' — the global SSE connection never lets the network go idle. Launch chromium with args: ['--no-sandbox'].
  • Typecheck OOMs without NODE_OPTIONS=--max-old-space-size=8192.
  • The worktree has no node_modules of its own; Vite/bun need a symlink: ln -s <repo-root>/node_modules node_modules (gitignored).