Point the docs index at the current site, and index the docs - #35
Merged
Conversation
The task battery failed on T3.8 because get_reviewed_page fetched https://v2.virtualflybrain.org/circuit-browser and got a 404. There is no such page. The model constructed it, because the index had nothing about the Circuit Browser and the only v2 URL it had seen was the Geppetto one — and any path on an allow-listed host is fetchable. The index held 13 entries and two of them were documentation pages. The site publishes 77 pages under /docs/. Asked how to use any part of the site, the model had nothing to read. Three corrections. The home page pointed at the bare host, which is not the canonical form — hugo.toml sets baseURL to www — so it answered a 302 to http. The 3D viewer pointed at the v2 single-page app, which serves no readable text; the site now documents the viewer properly. And /reports/ is a routed namespace rather than a document: bare /reports/ redirects into the Geppetto app, so it has no text to review. It belongs in the outbound allow-list, which is where linking is decided, not here. Then fifteen additions: the whole Website Features section, and a landing page for each of Concepts, Data, Tutorials, APIs, Anatomy Diagrams, Resources and Overview, so there is an entry point into every part of the documentation rather than a guess. Titles and summaries are taken from each page's own front matter in the VFB2 wrapper site rather than written here, so they stay true to the page. Every one of the 27 URLs was checked and returns 200 with no redirect.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The task battery failed on T3.8 because
get_reviewed_pagefetchedhttps://v2.virtualflybrain.org/circuit-browserand got a 404. There is no suchpage — the model constructed it, because the index had nothing about the Circuit
Browser and the only
v2.URL it had seen was the Geppetto viewer. Any path onan allow-listed host is fetchable, so a plausible guess gets tried.
The underlying problem is bigger than one task: the index held 13 entries, two of
which were documentation pages. The site publishes 77 pages under
/docs/. Askedhow to use any part of the site, the model had almost nothing to read.
Corrections
hugo.tomlsetsbaseURL = "https://www.virtualflybrain.org/", so the bare form is notcanonical and answered a 302 — to
http, at that.The site documents the viewer properly now, so it points there.
/reports/removed. It is a routed namespace rather than a document: bare/reports/redirects into the Geppetto app, so there is nothing to review. Itbelongs in the outbound allow-list, where linking is decided, not in an index
of pages to read.
Additions
The whole Website Features section — search and query, 3D viewer, slice
viewer, term info, term context, thumbnails, circuit browser — plus a landing
page for each of Concepts, Data, Tutorials, APIs, Anatomy Diagrams, Resources and
Overview, so there is an entry point into every part of the documentation rather
than a guess.
Titles and summaries are taken from each page's own front matter in the VFB2
wrapper site rather than written here, so they stay true to the page.
Not addressed here
This removes most of the reason for the model to invent a URL; it does not stop
it.
get_reviewed_pagewill still fetch any path on an allow-listed host.Constraining it to indexed URLs is the actual fix, and it is the same shape as
the existing grounding rule for identifiers — worth its own change rather than
folding a behaviour change into a data update.
Verification
redirect: 'manual'— every one returns 200 with noredirect.
node --test tests/unit/*.test.mjs— 1277 pass, 0 fail.