State
space/ holds a working Gradio application — the Credibility Inspector, deployed as a HuggingFace Space and embedded at uofa.net/demo — plus its own 20-file test suite under tests/space/.
It is absent from every entry point a contributor would use to find it:
docs/repo-layout.md, whose stated job is "top-level orientation", does not list space/ in its top-level directories table at all.README.md, AGENTS.md and CHANGELOG.md contain zero occurrences of "inspector".
There is a further naming split that compounds it: the code, the HF Space metadata and space/README.md call it Gap-Finder, while the UI title, docs/credibility-inspector.md and site/src/components/GapFinderEmbed.astro call it the Credibility Inspector. Someone grepping for one name does not find the other.
Why it matters
The Inspector is not a side project. AGENTS.md:77-84 carries a rule specifically about it ("Don't hardcode pack-derived data … in src/ or the Space"), and tests/space/test_emittability.py enforces a hard invariant that the Space must never emit anything the CLI would not. A contributor can therefore violate a documented rule about a component they had no way to discover from the repo's own navigation, and only find out when an anti-fork test fails.
Suggested next step
Add a space/ row to the docs/repo-layout.md top-level table, in the same shape as the existing rows, naming both spellings so either search term lands. Its "Where to find things" table is a natural second spot — a "How do I run the Inspector locally?" row would carry most of the value on its own.
Whether to reconcile Gap-Finder and Credibility Inspector into one name is a larger call and probably a separate decision; naming both in one place is cheap and fixes the discoverability half regardless.
Discovered during
PR #85, while adding a render-only panel to space/ and looking for where that component was documented.
State
space/holds a working Gradio application — the Credibility Inspector, deployed as a HuggingFace Space and embedded atuofa.net/demo— plus its own 20-file test suite undertests/space/.It is absent from every entry point a contributor would use to find it:
docs/repo-layout.md, whose stated job is "top-level orientation", does not listspace/in its top-level directories table at all.README.md,AGENTS.mdandCHANGELOG.mdcontain zero occurrences of "inspector".There is a further naming split that compounds it: the code, the HF Space metadata and
space/README.mdcall it Gap-Finder, while the UI title,docs/credibility-inspector.mdandsite/src/components/GapFinderEmbed.astrocall it the Credibility Inspector. Someone grepping for one name does not find the other.Why it matters
The Inspector is not a side project.
AGENTS.md:77-84carries a rule specifically about it ("Don't hardcode pack-derived data … insrc/or the Space"), andtests/space/test_emittability.pyenforces a hard invariant that the Space must never emit anything the CLI would not. A contributor can therefore violate a documented rule about a component they had no way to discover from the repo's own navigation, and only find out when an anti-fork test fails.Suggested next step
Add a
space/row to thedocs/repo-layout.mdtop-level table, in the same shape as the existing rows, naming both spellings so either search term lands. Its "Where to find things" table is a natural second spot — a "How do I run the Inspector locally?" row would carry most of the value on its own.Whether to reconcile Gap-Finder and Credibility Inspector into one name is a larger call and probably a separate decision; naming both in one place is cheap and fixes the discoverability half regardless.
Discovered during
PR #85, while adding a render-only panel to
space/and looking for where that component was documented.