Storage Editor for Apify datasets, key-value stores and request queues

Storage Editor is a web-based editor for the three Apify storage types: datasets, key-value stores and request queues. It lets you browse, search, create, rename, duplicate and delete storages, edit their contents, and, most importantly, edit a dataset by dropping fields and rows and saving the result as a new dataset. Datasets have no edit API on the Apify platform, so this is the practical way to clean one up without writing a script.

The Actor runs only as a Standby Actor: it behaves like an always-on web app at a stable URL, starts in about a second when first opened, and costs nothing while idle. You never start it as a run; a run started by hand only prints these instructions and finishes. It is written in Go, ships as a 7.5 MB Docker image and runs in the smallest 128 MB memory slot.

What you can do with it

• Clean up a scraped dataset before handing it over: remove test rows, drop internal fields such as debug URLs or raw HTML, and save a tidy copy under a new name.
• Split or trim data: keep only rows matching a search term (a domain, a status, a category) and save them as a separate dataset.
• Fix a request queue: change the URL, method, userData or handledAt of individual requests, add missing ones, or delete stale ones.
• Manage key-value records: read, write and delete records of any content type, including images and other binary files, with a preview for images.
• Organise storages: find a storage by name or ID among thousands, rename it, duplicate it, or delete it. Duplicating copies everything, including binary records and request metadata.

Features

• Three tabs, one per storage type, each with paging, an "include unnamed" switch and a search box that scans your whole account (up to 20 000 storages) by name or ID, optionally case sensitive.
• Custom view for any storage: shows every top-level field as a checkbox to drop or keep, offers a row search with "remove matching", "keep only matching" and per-row remove, and saves the result as a new storage or replaces the original in place.
• Datasets of any size: the dataset view previews only the first 100 rows (or 1 000, or 10 000). You pick fields and search rules on the preview, and saving applies them to every row on the server, streamed in the background. A dataset with 50 000 or 5 million rows is edited the same way within the 128 MB memory slot. A dataset that fits in the preview is edited row by row exactly as you see it.
• Dataset editor: paged item viewer with a text filter and a push-items box.
• Key-value store editor: key list with prefix filter, record editor with content type, base64 handling and file upload for binary records, image preview.
• Request queue editor: request list with filter, a field-by-field form for URL, method, unique key, userData, headers, payload, retryCount, noRetry and handledAt, and a raw-JSON escape hatch for anything else. Only changed fields are written.
• Duplicate any storage under a new name; refuses to write into a storage that already exists.
• Closes itself when unused: the run shuts down ten seconds after the last tab is closed, or after three minutes without any activity in an open tab, with a 30-second countdown first. Any click or key press keeps it alive, and reopening the page starts a fresh run in a few seconds.
• Built-in diagnostics: the header shows the build, run and whether your token can see account storages, and 403 responses explain what to fix.

How to use the web UI

1. Open the Actor in Apify Console and go to the Endpoints tab. Copy the Standby URL.

2. Open it in your browser with your API token appended once:

   https://<your-username>--storage-editor.apify.actor/?token=<YOUR_APIFY_TOKEN>
   

   The page removes the token from the address bar, keeps it in the browser tab and sends it with every request. The platform authenticates each request and routes it to a Standby run owned by your account, and the Actor uses the same token to call the Apify API for you.

3. Pick a tab, find a storage, and use Open to edit contents, View to build a filtered copy, or the row actions to rename, duplicate or delete.

Editing a dataset, step by step:

1. Open the dataset and click Custom view. The first 100 rows load; a bigger preview can be picked in the Preview panel.
2. In Fields, tick the fields to drop (or switch to "keep only the checked fields"). Fields that do not appear in the preview rows can be added by name.
3. In Rows, type a search term and click Remove matching or Keep only matching. Each click adds a rule, listed below the search box, that applies to every row of the dataset. The search looks at the whole original row as JSON, so you can filter on a field and hide it afterwards, or match an exact value such as "status":"error". The cross next to a row removes just that row.
4. Enter a name and click Save as new dataset, or Save and replace original to keep the old name. The server copies the dataset through your rules in the background and the page shows the progress; the run stays alive until the copy finishes even if you close the tab. Replacing writes to a temporary name and deletes the original only after every row is written, then renames the copy.

Permissions

The Actor runs with Limited permissions. A Limited run token only reaches the run's own default storages, so the editor does not use it for your data: every storage call is made with the API token you opened the page with, for that request only. The token is kept in memory for the life of the run and is never logged or stored.

This means the editor can reach exactly what your token can reach. Use a scoped token limited to some storages and the editor sees only those. Because the Actor does not need Full permissions, accounts that may not run Full-permission Actors owned by other users, such as Apify admin accounts, can use it.

The header of the web UI shows "full access" when your token can list the account's storages, and "(run token)" if your token did not reach the Actor and the run's own token was used.

The editor needs an API token in the URL. If the Standby endpoint is set to Console login instead, the browser has no token to send, and the editor can reach only the run's own storages.

Limits

Limit	Value	Reason
Dataset size for a custom view	no limit	the preview holds up to 10 000 rows; saving streams every row on the server
Rows loaded into a key-value store or request queue view	10 000	browser memory; the page warns when a storage is larger
Items per dataset push	as many as fit in 5 MB	Apify API payload limit; bigger pastes are split automatically
Storages scanned per search	20 000	the API has no name filter, so pages are scanned
Record or item body per request	10 MB	server-side cap
Time per request	5 minutes	Standby platform limit; affects duplicating very large storages
Field filtering	top-level fields	nested keys are shown inside their parent value
Idle shutdown	3 minutes unused, or 10 s after the last tab closes	the run closes when nobody uses the editor; a countdown warns 30 s ahead

Troubleshooting and FAQ

The page shows "403" and "LIMITED access" in the header. The token you opened the page with cannot list the account's storages, usually because it is a scoped token. Open the Standby URL again with a token that has access to the storages you want to edit. If the header also says "(run token)", the token did not reach the Actor; report it as a bug.

I rebuilt the Actor but see the old UI. Standby runs outlive builds. Abort them or wait for the idle timeout. The header shows the build number of the run you are talking to.

Duplicate says the name already exists. The Actor refuses to merge into an existing storage. Choose a new name, or delete the old storage first.

The preview says "about N rows" but the saved dataset has a different count. The number is estimated from the preview. The rules run on every row when saving, and the final toast shows the exact count.

Why did the cross remove only one row? A cross removes that exact preview row. To remove all rows like it across the whole dataset, search for something they share and use Remove matching.

Pressing Start does nothing useful. That is by design: a normal run only prints instructions, sets its status message accordingly and finishes. Use the Endpoints tab instead.

The page says the editor closed. Nobody used it for a while, so the run stopped to save compute. Open the Standby URL again with your token; a new run starts in a few seconds and everything is where you left it, because all data lives in your storages, not in the run.

Can I use a scoped token? Yes for authentication, but a scoped token that restricts the Actor scope narrows the run's access and the editor will get 403s. Use a token without that restriction.

Found a bug or missing a feature? Open a report in the Issues tab of this Actor with the storage type, the action you took and the message from the toast or the header.

How it works, in short

The editor is a small web application served by the Actor itself. Everything it does is an ordinary Apify API call made with the run's own token, so it can only ever touch the storages of the account that opened it, and nothing is stored anywhere except in your storages. No data passes through third-party services.