v0.9.11 — architecture map and route sidebar

paperclip-codegraph v0.9.11

Install

npm install -g @colbymchenry/codegraph@1.6.0   # the CodeGraph CLI, if you do not have it
paperclipai plugin install paperclip-codegraph@0.9.11

Pin the version: Paperclip stores a caret range against what was installed, and for a
0.x version a caret does not cross a minor release, so a bare plugin install is a
no-op once the next minor is out.

The plugin installs disabled. Nothing changes for any agent until an organization
turns it on, in Settings → Plugins → CodeGraph.

The Map view is now the architecture map CodeGraph actually has, and the host's
Back button is gone.

The Map was wrong, and I could look

The previous release offered a "Map" tab that drew the selected symbol's
callers and callees. That is not what CodeGraph's map is. Loading
http://127.0.0.1:4747/#/map and reading its DOM showed a different thing
entirely — an architecture map:

each module sits one layer above the modules it depends on, so reading top to
bottom follows the dependency direction. Line weight is how many calls, imports
and type references cross the link.

That is worth having, so the tab now does that:

• modules with their file and symbol counts, and a bar showing how much leans
  on them, scaled against the most depended-on box;
• links weighted by reference count, thicker for more;
• cycles reported in words, because a layered drawing cannot express one. The
  edge that closes a cycle is drawn dashed and excluded from the layering rather
  than hidden, so the picture does not claim a hierarchy it does not have.
• weak links held back until a module they touch is selected, with the count
  stated. Hidden half a cycle would be a lie, so links touching a cycle are never
  held back.
• controls that matter: Showing scopes to a subtree with real file counts,
  Grouping picks the depth (automatic, or 1–4 folders), and Include test modules un-hides the test-only boxes.

Against the real POS index it produces 48 modules, 137 links, 17 layers, 2
cycles from 5,415 cross-file edges — comparable to the viewer's 54/108/6 on the
same repository, which is the check that matters: not identical, but the same kind
of picture.

Three things the original does that this does not pretend to: hover-a-link to see
the symbol pairs behind the weight, copy-image, and download-SVG. The exclusions
note is there without the confidence threshold, because this plugin has no
confidence scores to threshold.

The Back button is the host's, and a slot removes it

PluginPage.tsx:166 renders {!routeSidebarActive && <Back to dashboard>}, and
routeSidebarActive is true only when resolveRouteSidebarSlot finds a
routeSidebar slot whose routePath matches the page slot's, in the same plugin.
No amount of page-side styling changes that: the answer is to declare one.

So the manifest now declares a routeSidebar beside the page, and it is used for
something real rather than to satisfy the check — the view rail. That is also
where the viewer keeps its views, so the page loses its tab bar and gains a proper
rail, with the view held in the URL (#symbol, #map) so the rail can link to it,
a reload keeps it, and the two cannot disagree.

The manifest test gained the capability mapping for the new slot type and an
assertion that the route sidebar's routePath still matches the page's — because
losing that pairing silently brings the Back button back, and nothing else would
notice.

Testing

486 tests, 469 passing and 17 skipped without an index. New: 39 for the
architecture layout and map derivation, plus 7 that run against the real POS
index — including that no two modules overlap in a layer and that every drawn link
references a real module.