Find all downloadable files on a webpage

01

This is not a bulk downloader

Almost every extension that scans a page for files exists to put all of them on your disk in one press. That is a genuinely useful thing and it is not this. If you already know you want the files, a mass downloader is the better tool and you should use one.

Foretaste answers the question one step earlier: which of these do I actually want? It lists them, says what each one is, and opens any single one in a reader. Your Downloads folder stays as it was unless you decide otherwise, and a page of forty attachments costs nothing to look through.

The two are not rivals so much as different halves of the same afternoon. Scan first, read the two files that matter, then save those two.

02

What counts as a file, in four rules

A page is full of links and most of them are navigation. Foretaste decides with the same four rules the button beside a link uses, so the two can never disagree with each other. They are tried cheapest and most standard first:

A link back to the page it is on is never a file. This is the rule that stops a "share this page" link becoming an entry.
What the page itself declared, in the download and type attributes. These mean the same thing on every website there has ever been, and a page that marks its own link as a download knows better than any guess about it. It is why /attachment?id=8812, an address with no filename and no extension anywhere in it, is listed correctly.
The rules for that particular site. A handful of sites put the file at a different address from the one the link points at. A GitHub /blob/ address is the page around a file, so it resolves to the raw file instead. These are plain data, and you can add your own for a site nobody here has heard of.
The filename, last, because it is the weakest of the four. A path ending in a slash names a folder whatever it says before the slash, which is why archive.org's "(View Contents)" link beside an archive no longer offers to open a directory listing as an archive.

Getting the order right is the difference between a list of files and a list of links. github.com/vercel/next.js stops being treated as a JavaScript file, and the attachment with no extension starts being listed.

03

What the list gives you

Grouped by kind, with a count on each group: documents, spreadsheets, archives, pictures, media, text and code, and the rest. On a release page that separates the six binaries from the four checksum files immediately.
Filter as you type, and chips to show only one kind. On a directory index of a hundred entries, typing three letters beats reading it.
Sort by name or by kind.
Ask how big they are, before downloading any of them. A page never says how big its files are — the size is in a response header — so this is a button rather than something that happens to you. Press it and Foretaste sends a HEAD request per file, which asks for the headers and no body, then sorts the answers into four bands: under 1 MB, 1 to 10 MB, 10 to 100 MB, and over 100 MB. You can then show one band at a time. Some servers refuse a HEAD, and those files are recorded as "it would not say" rather than guessed at or fetched to find out. Why a page cannot tell you this itself.
Open any one of them in the reader, in a tab or in a panel over the page you are already on. Every format Foretaste opens is on the format list, with what you actually see for each.
Save any one of them without opening it first. Saving never needs website access and never fails for a permission reason.
Step from one file to the next from the header of the file you are reading, so working through eight attachments does not mean going back to the list eight times. Choosing which ones first, and playing through them without a press between each, is Pro. Stepping is not.

The list stops at 200 entries. That is a real ceiling rather than a soft one: a file index with three thousand rows is a page to filter rather than a page to read, and a list that long is slower to use than the browser's own find.

04

Scanning fetches nothing

Opening the list makes no network request at all. It reads the addresses already in the page you are looking at, which your browser has already loaded, and works entirely from those. A page of two hundred links produces two hundred rows and zero requests.

That is a decision rather than a limitation. A scan that fetched each file to report its size would fire two hundred requests at somebody else's server the moment you opened a list, for files you were probably never going to open. Two things here do go to the network and both sit behind a press that says what it is about to do: thumbnails, which need the picture and tell you how many files they are about to fetch before they start, and opening a file, which fetches that file and nothing else.

Where those requests go: to the website the file is already on, and nowhere else. There is no Foretaste server for a file to reach.

05

Where it actually helps

These are the pages this was built against, not invented examples.

A course or lecture page with twenty attachments named week3-handout-v2-final.pdf. Scan, read the three you need, save nothing else.
A release page with a binary per platform plus signatures and checksums. The grouping separates them, and the checksum file is a text file you can read in place while checking the download against it.
An archive.org item page, which lists the same file several times at several addresses. One button per file rather than per link, so the list is as long as the item is.
A documentation or dataset folder listing, where the names are timestamps and the only way to tell them apart is to look inside. A ZIP is listed from its index, so looking inside a 40 MB archive costs about 67 KB.

Finding them is free. Acting on all of them is Pro.

The scan, the grouping, opening any of them and stepping from one to the next are free. These two are about doing something to the whole list at once.

Where it stops

It reads the page as loaded. Files behind a search box, a sign-in, or a button that fetches more results are not in the page yet, so they are not in the list. Load them first, then scan.

It needs access to that site. Reading the addresses in a page is something the browser asks your permission for, and it asks once, when you install Foretaste. There is no second prompt at the moment you scan.

A listed file can still refuse to open. Being linked and being readable are different things, and a site is entitled to say no. Why a site refuses, and what each answer means covers all of it. Saving to disk still works in every one of those cases.