Skip to content

Private workspace in the kernel (FA-91); sidebar gains Search, More and Prism ID - #235

Merged
humuhumu33 merged 5 commits into
mainfrom
user-workspace
Nov 16, 2026
Merged

humuhumu33 merged 5 commits into
mainfrom
user-workspace

Conversation

@humuhumu33

Copy link
Copy Markdown
Contributor

Why

Each visitor needs a workspace only their Prism ID can open, and the sidebar needs one way to search and one way to reach Prism ID.

What

Kernel: app/src/Workspace.lex.tex (FA-91) decides whether a workspace opens and whether a record may be read, stored, moved in from before a Prism ID, or its plain copy removed. ForgeKernel::dispatch routes workspace requests to it. Eight breaks theorems and nineteen workspaceWords_ theorems, replayed as acceptance vectors. No Studio code calls it yet: sealing records is the next change.

Sidebar (hologram.screens/AppSidebar.tsx): Search under Home with Ctrl/Cmd K, More with Edit sidebar (kept in holo.sidebar.v1), and Prism ID at the foot, loading the Prism frame's own panel only when its bytes match their SHA-256. The product is named Prism Foundry.

Non-Prism lines shrink. The palette lives inside the sidebar file because the ratchet refuses a new non-Prism file, so upstream's AppSearchDialog.tsx is restored byte for byte and the home input's app picker stays upstream's.

Checks

Lean verified all 42 modules (attestation dcc25ea1…). Both gates fail on planted defects and pass once restored (verification/private-workspace.md). Full test suite, format, model and audit gates passed locally on the first commit; the ratchet follow-up commit is left to CI.

Note

#231 also registers FA-91. Whichever merges second must renumber.

The Prism ID panel's redesign lands separately in UOR-Foundation/template#94; this sidebar picks it up once that is published.

🤖 Generated with Claude Code

humuhumu33 and others added 5 commits November 16, 2026 16:34
…nd Prism ID

The kernel answers whether a workspace opens, and whether a record may be read,
stored, moved in from before a Prism ID, or its plain copy removed
(app/src/Workspace.lex.tex, FA-91). ForgeKernel::dispatch sends "workspace "
requests to it ahead of the protocols. Eight breaks theorems say a workspace
opens only for its owner, with the current boundary key and a live session,
never on a removed device; a record is read only when authentic and in place
and stored only sealed. Nineteen workspaceWords_ theorems are proved by
evaluation and replayed by the guest as acceptance vectors. Both gates were
shown to fail on a planted defect (verification/private-workspace.md). No
Studio code asks the kernel yet: sealing records is the next change.

The sidebar (hologram.screens/AppSidebar.tsx):
- Search under Home, and Ctrl or Cmd K: one centred palette over the apps and
  chats already in memory, with no debounce or round trip. It replaces
  upstream's AppSearchDialog at build, so the home input's app picker uses it too.
- More: the items the visitor tucks away, and Edit sidebar to choose them
  (Templates and Plugins by default; kept in localStorage holo.sidebar.v1).
- Prism ID at the foot: the Prism frame's own panel, loaded on first use and
  run only if its bytes match its pointer's SHA-256, as an app page runs it.
  This makes true what the Studio shell surface already claimed.

The product is renamed Prism Foundry everywhere it is named, and the facets are
projected again from app/product.json.

Non-Prism lines shrink by 342. Upstream's app-sidebar.tsx and AppSearchDialog.tsx,
which ours replace whole at build, are reduced to typed pointers (VENDORED.md),
and dead code goes from host.js, prism.js, lexlean.mjs, obligations.mjs and
build.mjs. Two tests move with it: the kernel's vector count is 355, and the
patchSpec slice ends at stepSpec.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…earch dialog returns

The non-Prism ratchet refuses any new non-Prism file, so the Search palette moves
into hologram.screens/AppSidebar.tsx and upstream's AppSearchDialog.tsx is restored
byte for byte. The home input's app picker is upstream's again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n's imports

ForgeKernel imports Workspace, so its linear_arithmetic proofs staged
LexLeanRuntime.Laws into the application build, where Lean's C emitter loops
(LexLean #465). The SDK's Lean stage timed out on CI twice (PP5007).
WorkspaceProofs is an entrypoint nothing in the application imports, as
ForgeKernelProofs is. lexlean fmt, lock and check pass on 43 modules in the
pinned SDK image.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
product.json's three workspaceWords rows point at WorkspaceProofs; the facets and
lexlean.toml's entrypoints are projected again by foundry_model.mjs. lexlean fmt,
lock and check pass on 43 modules in the pinned SDK image.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@humuhumu33
humuhumu33 merged commit cb12b5f into main Nov 16, 2026
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant