Make the interface answer the way the tools answer

[Elnazkarami]

Follows #18, which is merged. Two things found by actually using the interface rather than assuming it worked.

The query page was giving a wrong answer

The page decided for itself which sessions matched, by comparing strings in the browser. A session whose species had never been entered compared unequal to mouse and was dropped — so it answered a question about missing evidence as though it were a question about contradicting evidence, and showed a cohort cleaner than the data supports.

That is the one mistake this project exists to prevent, and the page was making it.

It happened because the standard was implemented twice: once in Python, once in TypeScript. So the TypeScript is gone. The query runs where the rules live, and the page renders what comes back — matched, excluded, and cannot be ruled out — with the evidence that placed each session in its group.

On a real project:

considered 11 · matched 0 · excluded 0 · unresolved 11
"11 could not be ruled out because species was never entered."

The old page showed 0 of 11, which reads as none of these are mice. None of them are anything yet. Nobody has said.

Four endpoints, all read-only

open	Read a JSON document by name, instead of making someone drag a file in — a workaround for a page that could not reach a disk
link	Build a project's linked records in memory. table check --emit writes them to a file; nothing here needs that file to exist, so querying leaves the project untouched
check	Check the metadata tables against each other — whether a procedure names an animal that exists. The entry form judges one value at a time and cannot see across three files
query	The above

browse now offers files as well as folders, so a manifest is chosen the way a person refers to it: by name.

The other pages follow. Manifest opens from disk. Validate checks the tables a project already has, beside the form for entering what only a person knows. The home page no longer calls itself browser-only or promises that nothing leaves the browser — it reads drives now, and what it should promise is that nothing leaves the machine.

The progress the interface was meant to show

A scan reported a file count climbing towards nothing. progress asks for output on a terminal, which a page does not want — but it was also, silently, what decided whether the tree was walked up front to count the work. Passing progress=False and a callback, exactly what a page does, disabled the count and the estimate together.

before:  9,039/None files   39 MB/s   eta=None
after:   3,598/28,345 files 24 MB/s   eta=7.1s

One flag answering two questions, again. Three regression tests cover it; they fail with the fix reverted.

Also in here: picking a file rather than a folder returned a 500 showing the word ValueError to someone who had clicked the wrong thing — now a 400 in words. And the tab had no favicon, because the icon ships in the bundle but nothing referenced it.

Verified

Every page rendered in a real browser, headless: all four mount, every picker lists actual directories, and each endpoint answers over HTTP. A query nobody can parse comes back as words, not a stack trace.

353 → 364 tests. All ten CI jobs green.