# Instance Access

How to reach the deployed inkreach-teable instance, acquire credentials, and
verify health. This file describes the **reference deployment** — adapt
endpoint/ports for your own deployment.

## Connection

| Service | URL / Port | Mode | Purpose |
|---------|-----------|------|---------|
| Backend API | `http://localhost:3000` | production (`NODE_ENV=production`) | REST API, automation engine, websocket |
| Frontend | `http://localhost:3002` | production | End-user web UI (Next.js) |
| PostgreSQL | `localhost:5432` | — | Database (db: `teable`, user: `teable`) |
| Redis | `localhost:6379` | — | BullMQ queue for automation runs, distributed lock |

- The frontend reads `SOCKET_PORT=3000` from `.env` to point its websocket at the backend. Don't let it fall back to `3001`.
- End users browse `:3002`. The backend `:3000` is the API only (`BACKEND_SKIP_NEXT_START=true` disables the embedded Next server).

## Health Checks

```bash
# Backend API up?
curl -s -o /dev/null -w '%{http_code}' http://localhost:3000/api/auth/signin -X POST \
  -H 'Content-Type: application/json' -d '{}'   # expect 401 (endpoint exists, bad creds)

# Frontend up?
curl -s -o /dev/null -w '%{http_code}' http://localhost:3002   # expect 307 or 200

# Ports listening?
ss -tlnp | grep -E ':3000|:3002|:5432|:6379'
```

If `:3000` is unreachable but systemd shows `active`, the worker likely OOM-crashed while the `pnpm` parent stayed alive:
```bash
systemctl --user stop teable-backend.service
pkill -9 -f 'dist/index.js'
systemctl --user start teable-backend.service
```

## Acquiring a PAT (Personal Access Token)

The CLI authenticates with a PAT (`teable_{id}_{encryptedSign}`). Two ways to get one:

### Option A — via the CLI (needs a session cookie or existing token)
```bash
inkreach-teable auth token create --name my-cli \
  --scopes automation|read,automation|create,automation|update,automation|delete,table|read,record|read,record|create,record|update \
  --base-ids bseXXX --expires 2099-12-31
# → prints full token ONCE
```
Note: `auth token list/create` may require a **session** (browser login) rather than a PAT — if you get 403, use Option B.

### Option B — via the web UI (recommended for the first token)
1. Log in to `http://localhost:3002` as an admin.
2. User menu → **Settings** → **Personal Access Tokens**.
3. Create a token with scopes: `automation|*`, `table|*`, `record|*`, `field|*`, `view|*` as needed. Limit to the base if desired.
4. Copy the `teable_xxx_yyy` string.

### Configure the CLI
```bash
inkreach-teable auth login --endpoint http://localhost:3000 --token <PAT> --base-id bseXXX
inkreach-teable auth status   # verify
```

## Reference Credentials (local dev)

> Never commit real secrets. The reference dev instance uses:
> - Admin: `hr_admin@xingchen.tech` / (ask maintainer) — for browser login only.
> - Base: `bseeZR7yiDNeMPCa4CN` (人事薪资管理) — the test base with ~600 automations.
> - Test table: `tblzPOXQvUruQosKNP5` (has button fields for automation testing).

For CLI use, **always prefer a PAT** over the admin password.

## Permissions

- PAT scope ∩ base role = effective permission. To manage automations you need `automation|read/create/update/delete` scopes **and** a role of editor+ on the base.
- Roles: owner (all), creator (all on the base), editor (read/create/update), commenter (read), viewer (read).
- Workflows auto-create an internal token on activation (name `workflow:{id}`, scope limited to table/record read+write) — used by action nodes, not for CLI auth.

## Troubleshooting Access

| Symptom | Fix |
|---------|-----|
| `401 Unauthorized` | Token wrong/expired → recreate PAT, `auth login` again |
| `403 Forbidden` | Token lacks scope, or base role too low → widen scopes / raise role |
| `404 Base not found` | Wrong base id, or endpoint points to a different instance → `auth status`, check `--base-id` |
| Connection refused | Backend down → health check, restart service (see [development.md](development.md)) |