> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vendo.run/llms.txt
> Use this file to discover all available pages before exploring further.

# E-STORE-001

> Doctor warns when the store's data directory sits on disk a platform wipes on every redeploy.

```text store/persistence theme={null}
warning: the store's data directory "<dataDir>" is on ephemeral disk —
<platform> wipes it on every redeploy and your users' apps and data go with
it.
```

`<dataDir>` is the local path the store resolved (`.vendo/data` unless you
passed a custom one), and `<platform>` names what doctor detected: Railway,
Render, Fly.io, Heroku, or a generic `/tmp` mount.

`check: store/persistence` · `error_code: E-STORE-001` · `doctor exits: 0`

## What you're seeing

This is a warning, not a failure; doctor still exits 0. The app runs fine
today. The risk is the next redeploy, which wipes every app your users built
and all of their data with it.

## Why

Without a `VENDO_API_KEY`, the store defaults to PGlite, which writes to a
directory on the local container filesystem. That's the right default on a
laptop or a machine with a real disk. On a platform that recreates its
container filesystem on every deploy, it's a data-loss bug waiting for the
next release.

## The fix

Set `VENDO_API_KEY` to move onto Vendo Cloud's hosted store, which survives
redeploys by default.

```bash .env.local theme={null}
VENDO_API_KEY=vnd_...
```

Run `vendo login` if you don't have a key yet; it mints one and writes it to
`.env.local` for you. This check only runs while no key is set, so once
Cloud is wired, it stops firing.

## Related errors

* [E-CFG-002](/production/troubleshooting/e-cfg-002): the gitignore that keeps this same data directory out of git
* [E-CFG-001](/production/troubleshooting/e-cfg-001): a missing `.vendo/` config file