feat: Account for model memory and retain worker launch ownership

.gitattributes

@@ -0,0 +1 @@
+*.py text eol=lf

.github/workflows/ci.yml

@@ -52,6 +52,9 @@ jobs:
           cache-dependency-path: pyproject.toml
       - name: Install dependencies
         run: pip install -e ".[all,dev]"
+      - name: Trace synthetic Windows file ownership
+        if: runner.os == 'Windows'
+        run: python tests/native/windows-owned-file-diagnostics-v2.py
       - name: Run tests (gate — deterministic, no network)
         # Issue #426: exclude `network`-marked tests (real HuggingFace model
         # downloads) from the gate so a transient HfHubHTTPError can't fail CI
@@ -115,17 +118,38 @@ jobs:
 
   # Hunter F26: catch packaging regressions (missing py.typed, metadata
   # errors, broken README rendering, sdist contents) in CI instead of
-  # at release time. Runs once on 3.12 only — duplicating across the
-  # matrix wastes CI minutes without additional signal.
+  # at release time. Exercise both MCP dependency boundaries on each OS.
   build-check:
-    runs-on: ubuntu-latest
-    timeout-minutes: 10
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 20
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        python-version: ["3.12"]
+        mcp-version: [minimum, latest]
+        pydantic-version: [latest]
+        include:
+          - os: ubuntu-latest
+            python-version: "3.10"
+            mcp-version: minimum
+            pydantic-version: latest
+          # Exercise the MCP/Pydantic floor pair and the release that removed
+          # eval_type_backport. Other dependencies still use normal resolution.
+          - os: ubuntu-latest
+            python-version: "3.10"
+            mcp-version: minimum
+            pydantic-version: "2.11.0"
+          - os: ubuntu-latest
+            python-version: "3.12"
+            mcp-version: minimum
+            pydantic-version: "2.14.0"
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
-      - name: Set up Python 3.12
+      - name: Set up Python ${{ matrix.python-version }}
         uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
         with:
-          python-version: "3.12"
+          python-version: ${{ matrix.python-version }}
           cache: 'pip'
           cache-dependency-path: pyproject.toml
       - name: Install build tooling
@@ -134,10 +158,28 @@ jobs:
         run: python -m build
       - name: Verify package metadata
         run: twine check --strict dist/*
-      # Issue #654 (M-83): actually install the built wheel and exercise the
-      # console entry point, so a broken/missing entry-point or import-time
-      # failure is caught here (not just at release).
-      - name: Install wheel and smoke-test console script
+      # Issue #757: an editable install or --help alone does not exercise
+      # the installed wheel's MCP handshake and serialized tool metadata.
+      - name: Install wheel and verify MCP compatibility
+        shell: python
+        env:
+          MCP_VERSION: ${{ matrix.mcp-version }}
+          PYDANTIC_VERSION: ${{ matrix.pydantic-version }}
         run: |
-          pip install dist/*.whl
-          truememory-mcp --help
+          import os
+          from pathlib import Path
+          import subprocess
+          import venv
+
+          wheel, = Path("dist").glob("*.whl")
+          env_dir = Path(os.environ["RUNNER_TEMP"]) / "mcp-wheel-env"
+          venv.EnvBuilder(with_pip=True).create(env_dir)
+          python = env_dir / ("Scripts/python.exe" if os.name == "nt" else "bin/python")
+          requirements = [str(wheel.resolve())]
+          if os.environ["MCP_VERSION"] == "minimum":
+              requirements.append("mcp[cli]==1.21.1")
+          if os.environ["PYDANTIC_VERSION"] != "latest":
+              requirements.append("pydantic==" + os.environ["PYDANTIC_VERSION"])
+          subprocess.run([str(python), "-m", "pip", "install", *requirements], check=True)
+          subprocess.run([str(python), "-m", "pip", "check"], check=True)
+          subprocess.run([str(python), "-I", "tests/mcp-wheel-smoke.py"], check=True)

.github/workflows/publish.yml

@@ -32,7 +32,17 @@ jobs:
 
   build-check:
     runs-on: ubuntu-latest
-    timeout-minutes: 10
+    timeout-minutes: 20
+    strategy:
+      fail-fast: false
+      matrix:
+        mcp-version: [minimum, latest]
+        pydantic-version: [latest]
+        include:
+          - mcp-version: minimum
+            pydantic-version: "2.11.0"
+          - mcp-version: minimum
+            pydantic-version: "2.14.0"
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
@@ -46,6 +56,29 @@ jobs:
         run: python -m build
       - name: Verify package metadata
         run: twine check --strict dist/*
+      - name: Install wheel and verify MCP compatibility
+        shell: python
+        env:
+          MCP_VERSION: ${{ matrix.mcp-version }}
+          PYDANTIC_VERSION: ${{ matrix.pydantic-version }}
+        run: |
+          import os
+          from pathlib import Path
+          import subprocess
+          import venv
+
+          wheel, = Path("dist").glob("*.whl")
+          env_dir = Path(os.environ["RUNNER_TEMP"]) / "mcp-wheel-env"
+          venv.EnvBuilder(with_pip=True).create(env_dir)
+          python = env_dir / ("Scripts/python.exe" if os.name == "nt" else "bin/python")
+          requirements = [str(wheel.resolve())]
+          if os.environ["MCP_VERSION"] == "minimum":
+              requirements.append("mcp[cli]==1.21.1")
+          if os.environ["PYDANTIC_VERSION"] != "latest":
+              requirements.append("pydantic==" + os.environ["PYDANTIC_VERSION"])
+          subprocess.run([str(python), "-m", "pip", "install", *requirements], check=True)
+          subprocess.run([str(python), "-m", "pip", "check"], check=True)
+          subprocess.run([str(python), "-I", "tests/mcp-wheel-smoke.py"], check=True)
 
   version-check:
     runs-on: ubuntu-latest

install.ps1

@@ -86,7 +86,7 @@ if ($LASTEXITCODE -ne 0) {
 }
 
 # ---------- step 3: install truememory as a uv tool ----------
-Say "installing $PKG_SPEC (~3-5 min on first run, downloads all tier models)..."
+Say "installing $PKG_SPEC (~3-5 min on first run)..."
 & uv tool uninstall truememory *> $null
 if ($LASTEXITCODE -gt 1) {
     Warn "uv tool uninstall returned $LASTEXITCODE — proceeding, but result may be partial"
@@ -118,6 +118,8 @@ if ($uvToolDir) {
 }
 
 # ---------- step 4: auto-configure Claude ----------
+$setupStatus = "skipped"
+$hookStatus = "skipped"
 if ($env:TRUEMEMORY_SKIP_SETUP -eq "1") {
     Say "skipping Claude setup (TRUEMEMORY_SKIP_SETUP=1)"
 } elseif (-not $toolPython) {
@@ -126,41 +128,52 @@ if ($env:TRUEMEMORY_SKIP_SETUP -eq "1") {
 } else {
     Say "configuring Claude Code / Claude Desktop..."
     & $toolPython -m truememory.mcp_server --setup
-    if ($LASTEXITCODE -ne 0) {
+    if ($LASTEXITCODE -eq 0) { $setupStatus = "command succeeded" }
+    else {
+        $setupStatus = "failed"
         Warn "auto-setup returned non-zero (re-run: python -m truememory.mcp_server --setup)"
     }
 
     Say "installing hooks and CLAUDE.md instructions..."
     & $toolPython -m truememory.ingest.cli install
-    if ($LASTEXITCODE -ne 0) {
+    if ($LASTEXITCODE -eq 0) { $hookStatus = "command succeeded" }
+    else {
+        $hookStatus = "failed"
         Warn "hook install returned non-zero (re-run: python -m truememory.ingest.cli install)"
     }
 }
 
 # ---------- step 5: pre-download models for all tiers ----------
-Say "pre-downloading models for all tiers (Edge + Base + Pro)..."
-Say "  this takes 2-5 min but means tier switching just works afterward."
-if (Test-Path $toolPython) {
-    Say "  [1/3] Edge reranker (MiniLM-L-6-v2, ~22MB)..."
-    & $toolPython -c "from sentence_transformers import CrossEncoder; CrossEncoder('cross-encoder/ms-marco-MiniLM-L-6-v2')"
-    if ($LASTEXITCODE -eq 0) { Ok "  [1/3] Edge reranker ready" }
-    else { Warn "  [1/3] Edge reranker download failed (search still works without it)" }
-
-    Say "  [2/3] Base/Pro embedder (Qwen3-Embedding-0.6B, ~1.2GB)..."
-    & $toolPython -c "from sentence_transformers import SentenceTransformer; SentenceTransformer('Qwen/Qwen3-Embedding-0.6B', truncate_dim=256)"
-    if ($LASTEXITCODE -eq 0) { Ok "  [2/3] Base/Pro embedder ready" }
-    else { Warn "  [2/3] Base/Pro embedder download failed (you can retry later or use Edge tier)" }
-
-    Say "  [3/3] Base/Pro reranker (gte-modernbert, ~600MB)..."
-    & $toolPython -c "from sentence_transformers import CrossEncoder; CrossEncoder('Alibaba-NLP/gte-reranker-modernbert-base')"
-    if ($LASTEXITCODE -eq 0) { Ok "  [3/3] Base/Pro reranker ready" }
-    else { Warn "  [3/3] Base/Pro reranker download failed (you can retry later or use Edge tier)" }
-
-    Ok "all models pre-downloaded — tier switching is instant."
+Say "pre-downloading the Edge reranker and Base/Pro models..."
+Say "  this can take 2-5 min; uncached models require network access."
+$modelChecks = @(
+    @{ Label = "Edge reranker"; Status = "not checked"; Code = "from sentence_transformers import CrossEncoder; CrossEncoder('cross-encoder/ms-marco-MiniLM-L-6-v2')" },
+    @{ Label = "Base/Pro embedder"; Status = "not checked"; Code = "from sentence_transformers import SentenceTransformer; SentenceTransformer('Qwen/Qwen3-Embedding-0.6B', truncate_dim=256)" },
+    @{ Label = "Base/Pro reranker"; Status = "not checked"; Code = "from sentence_transformers import CrossEncoder; CrossEncoder('Alibaba-NLP/gte-reranker-modernbert-base')" }
+)
+if ($toolPython -and (Test-Path -LiteralPath $toolPython)) {
+    foreach ($modelCheck in $modelChecks) {
+        Say "  Pre-downloading $($modelCheck.Label)..."
+        $modelCheck.Status = "failed"
+        try {
+            & $toolPython -c $modelCheck.Code
+            if ($LASTEXITCODE -eq 0) { $modelCheck.Status = "ready" }
+        } catch {
+            Warn "  $($modelCheck.Label) could not be loaded: $_"
+        }
+        if ($modelCheck.Status -eq "ready") {
+            Ok "  $($modelCheck.Label) ready"
+        } else {
+            Warn "  $($modelCheck.Label) pre-download failed"
+            Warn "Retry this model only: & (Join-Path (uv tool dir) 'truememory\Scripts\python.exe') -c `"$($modelCheck.Code)`""
+        }
+    }
 } else {
     Warn "could not locate tool Python at $toolPython — skipping model pre-download"
-    Warn "models will download on first use instead"
+    Warn "Model readiness is unverified. Locate the tool environment with: uv tool dir"
+    Warn "After locating its Python, run: python -m truememory.ingest.cli setup"
 }
+$modelReadyCount = @($modelChecks | Where-Object { $_.Status -eq "ready" }).Count
 
 # ---------- done ----------
 Write-Host ""
@@ -174,10 +187,31 @@ Write-Host @"
                                   a sauron company
 "@ -ForegroundColor Green
 
-$installedVer = & $toolPython -c "from importlib.metadata import version; print(version('truememory'))" 2>$null
+$installedVer = $null
+if ($toolPython) {
+    try {
+        $installedVer = & $toolPython -c "from importlib.metadata import version; print(version('truememory'))" 2>$null
+    } catch {
+        $installedVer = $null
+    }
+}
 if (-not $installedVer) { $installedVer = "unknown" }
 Write-Host ""
-Ok "TrueMemory v$installedVer installed successfully."
+Ok "TrueMemory v$installedVer package installed."
+Say "Claude registration: $setupStatus"
+Say "Hooks: $hookStatus"
+Say "Model pre-download checks: $modelReadyCount/3 succeeded."
+foreach ($modelCheck in $modelChecks) {
+    Say "  $($modelCheck.Label): $($modelCheck.Status)"
+}
+if ($modelReadyCount -eq 3) {
+    Ok "All three requested model pre-download checks passed."
+} else {
+    Warn "Model readiness incomplete; package installation is retained. Retry the failed or skipped checks before relying on those models."
+}
+Say "The Edge embedder was not checked by this pre-download step."
+Say "Registration and hook command success do not verify runtime operation."
+Say "Changing embedding models can still require re-embedding stored memories."
 Write-Host ""
 Write-Host "  First time? Start a new Claude session and type:" -ForegroundColor Green
 Write-Host ""