Skip to content

The AI engine leaves the content surface — the gate lands it as a module (#2276) - #2524

Merged
rbuergi merged 2 commits into
mainfrom
feat/2276-ai-leaves-the-content-surface
Aug 27, 2026
Merged

rbuergi merged 2 commits into
mainfrom
feat/2276-ai-leaves-the-content-surface

Conversation

@rbuergi

@rbuergi rbuergi commented Aug 27, 2026

Copy link
Copy Markdown
Contributor

Summary

The core half of clearing #2276's remaining gate (the 409): FrameworkBuildIdentity.ContentSurfaceAssemblies drops MeshWeaver.AI. The engine is a registry module — composed into content compiles per-mesh via CompileReferences.ComposeWithModules, never part of the framework identity. The plugins half (dropping the portal Gui / LocalMesh ProjectReferences + the hosts' cutover to registry delivery) lands against this commit's platform pin; until then plugins builds at its current pin are untouched.

The bake host keeps the engine — as a module

Plugin packages ship Agent/Skill nodes, and a mesh without the AI node types refuses those installs. So mw-plugin-test now runs the engine exactly the way a portal does:

  • The tester csproj replaces its ProjectReference with a tester-local MeshModuleClosure row, written AFTER the MeshModulesPublish import so the shared inventory is untouched — no portal image gains a closure copy (that would be the 409 back again). The real closure lane lays modules/MeshWeaver.AI/ (entry DLL + its private NuGet closure) into build and publish output.
  • PluginGateRunner activates it via an in-memory configuration (Modules:Assemblies + Modules:Required + one Features:StaticRepoSync partition, union-expanded to all AI content partitions by the module's served-as-a-unit rule) through InstallConfiguredModules → AiMeshModuleAttribute. No compiled call.
  • A missing module fails the gate RED before the mesh boots — a gate never tests its own inputs.
  • The import-settled signal registers pre-settled (the documented "nothing to import" state — gate content arrives via installs).

The closure equality holds

Four projects that arrived transitively through the engine and are still content surface (Mesh.Operations, Application.Styles, Markdown.Collaboration, ContentCollections.Indexing.Graph) become direct tester references, so CanonicalList_MatchesTheTesterClosure and the manifest-host closure check pass unchanged in shape (all 23 identity tests green).

Tests

  • FrameworkBuildIdentityTest: 23/23 green.
  • MeshWeaver.PluginTester.Test: 91/91 green — nine of those suites boot the gate in-process, so this is the in-proc proof that installs, compiles, renders and Tests areas all work with the engine module-activated.

Deploy note

Merging this changes the bake identity core CI publishes under. Deployed portals keep their own (pin-frozen) identity and their existing bakes until the plugins pin bump + image roll, which should follow promptly — the flip sequence is tracked on #2276.

No What's New entry: pure-internal platform step — the user-facing story (the AI Store entry) ships with MeshWeaver.Plugins#687.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Tz1i2fFcJhvR1Eje8yFH5F

…ULE (#2276)

FrameworkBuildIdentity.ContentSurfaceAssemblies drops MeshWeaver.AI: the
engine is a registry module now, composed into content compiles per-mesh via
ComposeWithModules, never part of the framework identity. An image shipping
AI.dll in its app closure would 409 the registry module — this is the core
half of clearing that gate; the plugins half drops the portal Gui and
LocalMesh ProjectReferences against this commit's pin.

The bake host (mw-plugin-test) keeps needing the AI node types — plugin
packages ship Agent/Skill nodes, and a mesh without the types refuses those
installs — so the gate now activates the engine the way a portal does:

- the tester csproj replaces the ProjectReference with a tester-local
  MeshModuleClosure row (written AFTER the MeshModulesPublish import, so the
  shared inventory is untouched and no portal image gains the module), laying
  modules/MeshWeaver.AI/ into build and publish output;
- PluginGateRunner builds an in-memory configuration (Modules:Assemblies +
  Modules:Required + one StaticRepoSync partition, union-expanded by the
  module's served-as-a-unit rule) and InstallConfiguredModules activates the
  engine through AiMeshModuleAttribute — no compiled call;
- a missing module fails the gate RED before the mesh boots (a gate never
  tests its own inputs);
- the import-settled signal is registered pre-settled — the documented
  "nothing to import" state — since gate content arrives via installs.

Four projects that arrived transitively through the engine and are still part
of the content surface (Mesh.Operations, Application.Styles,
Markdown.Collaboration, ContentCollections.Indexing.Graph) become direct
tester references, so the canonical list and the tester closure stay equal —
CanonicalList_MatchesTheTesterClosure passes unchanged in shape.

MeshWeaver.PluginTester.Test imports the same lane with the same single row
(nine of its suites boot the gate in-process); all 91 pass.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tz1i2fFcJhvR1Eje8yFH5F
Copilot AI lite review requested due to automatic review settings August 27, 2026 18:58

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR continues #2276 by removing MeshWeaver.AI from the platform “content surface” identity and treating the AI engine as a module that is activated via Modules:Assemblies, rather than a compiled reference. It also updates the plugin gate/tester to load the AI engine from the modules/MeshWeaver.AI/ layout (closure lane) and fail early when the required module bits are missing.

Changes:

  • Dropped MeshWeaver.AI from FrameworkBuildIdentity.ContentSurfaceAssemblies, reflecting that the AI engine no longer contributes to the framework identity.
  • Updated mw-plugin-test gate boot to activate the AI engine through module configuration (InstallConfiguredModules) and to hard-fail when the required module cannot be resolved.
  • Reworked the PluginTester and its test project to lay out the AI engine via MeshModuleClosure (module publish layout) instead of a ProjectReference, and added direct references for assemblies that remain on the content surface.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File Description
tools/MeshWeaver.PluginTester/PluginGateRunner.cs Replaces compiled .AddAI(...) with module activation + required-module preflight; registers StaticRepoImportSettled pre-settled for the gate.
tools/MeshWeaver.PluginTester/MeshWeaver.PluginTester.csproj Removes AI ProjectReference; adds direct references for remaining surface assemblies; imports module publish targets and declares a tester-local MeshModuleClosure for AI.
test/MeshWeaver.PluginTester.Test/MeshWeaver.PluginTester.Test.csproj Ensures the test output also gets the AI module layout by running the same closure lane (targets import + single-row MeshModuleClosure).
src/MeshWeaver.Compiler/FrameworkBuildIdentity.cs Removes MeshWeaver.AI from the framework content-surface assembly list and documents the module-based composition rationale.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 3)

1 090 tests  ±0   1 090 ✅ ±0   17m 38s ⏱️ + 3m 44s
    7 suites ±0       0 💤 ±0 
    7 files   ±0       0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 2)

1 087 tests  ±0   895 ✅ ±0   4m 59s ⏱️ -10s
    6 suites ±0   192 💤 ±0 
    6 files   ±0     0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 0)

1 308 tests  ±0   1 308 ✅ ±0   6m 6s ⏱️ -9s
    6 suites ±0       0 💤 ±0 
    6 files   ±0       0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 4)

    8 files  ±0      8 suites  ±0   11m 55s ⏱️ +42s
4 408 tests ±0  4 405 ✅ ±0  3 💤 ±0  0 ❌ ±0 
5 007 runs  ±0  5 004 ✅ ±0  3 💤 ±0  0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 5)

1 591 tests  ±0   1 590 ✅ ±0   8m 8s ⏱️ - 1m 32s
    9 suites ±0       1 💤 ±0 
    9 files   ±0       0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results (shard 1)

739 tests  ±0   735 ✅ ±0   7m 26s ⏱️ +9s
  8 suites ±0     4 💤 ±0 
  8 files   ±0     0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

@github-actions

github-actions Bot commented Aug 27, 2026 •

Copy link
Copy Markdown
Contributor

Test Results

    44 files  ±0      44 suites  ±0   56m 14s ⏱️ + 2m 43s
10 221 tests ±0  10 021 ✅ ±0  200 💤 ±0  0 ❌ ±0 
10 822 runs  ±0  10 622 ✅ ±0  200 💤 ±0  0 ❌ ±0 

Results for commit f6127fd. ± Comparison against base commit 96a022b.

♻️ This comment has been updated with latest results.

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.

2 participants