mempalace-code 1.6.2__tar.gz → 1.8.0__tar.gz

mempalace_code-1.8.0/.github/workflows/ci.yml

@@ -0,0 +1,92 @@
+name: Tests
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+  workflow_dispatch:
+    inputs:
+      model_tests:
+        description: "Run needs_network model-backed tests with HuggingFace cache"
+        required: false
+        default: "false"
+        type: choice
+        options:
+          - "false"
+          - "true"
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.13"
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+      - run: pip install -e ".[dev,treesitter]"
+      - run: python3 -m pytest tests/ -v -m "not needs_network"
+
+  chroma-compat:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.13"
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+      - run: pip install -e ".[dev,chroma]"
+      - name: ChromaStore compatibility tests
+        run: python -m pytest tests/test_chroma_compat.py -v
+      - run: python -m pytest tests/ -v -m "not needs_network"
+
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.13"
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+      - run: pip install ruff
+      - run: ruff check mempalace_code/ tests/
+      - run: ruff format --check mempalace_code/ tests/
+
+  model-tests:
+    runs-on: ubuntu-latest
+    if: github.event_name == 'workflow_dispatch' && github.event.inputs.model_tests == 'true'
+    env:
+      HF_HOME: ${{ github.workspace }}/.cache/huggingface
+      MEMPALACE_TEST_HF_HOME: ${{ github.workspace }}/.cache/huggingface
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.13"
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+      - name: Restore HuggingFace model cache
+        id: hf-cache
+        uses: actions/cache@v5
+        with:
+          path: ${{ github.workspace }}/.cache/huggingface/hub/models--sentence-transformers--all-MiniLM-L6-v2
+          key: hf-${{ runner.os }}-py3.13-minilm-all-MiniLM-L6-v2-${{ hashFiles('pyproject.toml', 'uv.lock') }}
+          restore-keys: |
+            hf-${{ runner.os }}-py3.13-minilm-all-MiniLM-L6-v2-
+      - run: pip install -e ".[dev,treesitter]"
+      - name: Fetch embedding model (no-op on cache hit)
+        run: mempalace-code fetch-model
+      - name: Run model-backed tests
+        run: python -m pytest tests/ -v -m needs_network

mempalace_code-1.8.0/.github/workflows/dotnet-bench.yml

@@ -0,0 +1,73 @@
+name: .NET Benchmark
+
+on:
+  pull_request:
+    branches: [main]
+  push:
+    branches: [main]
+  workflow_dispatch:
+
+jobs:
+  dotnet-bench:
+    runs-on: ubuntu-latest
+    if: ${{ !(github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'skip-bench')) }}
+    env:
+      HF_HOME: ${{ github.workspace }}/.cache/huggingface
+      CLEAN_ARCHITECTURE_COMMIT: 5a600ab8749c110384bc3bd436b9c67f3067b489
+      R5_THRESHOLD: "0.900"
+    steps:
+      - uses: actions/checkout@v5
+
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.13"
+          cache: pip
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+
+      - name: Restore HuggingFace model cache
+        id: hf-cache
+        uses: actions/cache@v5
+        with:
+          path: ${{ github.workspace }}/.cache/huggingface/hub/models--sentence-transformers--all-MiniLM-L6-v2
+          key: hf-${{ runner.os }}-py3.13-minilm-all-MiniLM-L6-v2-${{ hashFiles('pyproject.toml', 'uv.lock') }}
+          restore-keys: |
+            hf-${{ runner.os }}-py3.13-minilm-all-MiniLM-L6-v2-
+
+      - run: pip install -e ".[dev,treesitter]"
+
+      - name: Fetch embedding model (no-op on cache hit)
+        run: mempalace-code fetch-model
+
+      - name: Fetch CleanArchitecture at pinned commit
+        run: |
+          git init /tmp/CleanArchitecture
+          git -C /tmp/CleanArchitecture remote add origin https://github.com/jasontaylordev/CleanArchitecture.git
+          git -C /tmp/CleanArchitecture fetch --depth=1 origin "$CLEAN_ARCHITECTURE_COMMIT"
+          git -C /tmp/CleanArchitecture checkout --detach FETCH_HEAD
+          ACTUAL=$(git -C /tmp/CleanArchitecture rev-parse HEAD)
+          if [ "$ACTUAL" != "$CLEAN_ARCHITECTURE_COMMIT" ]; then
+            echo "Commit mismatch: expected $CLEAN_ARCHITECTURE_COMMIT got $ACTUAL"
+            exit 1
+          fi
+
+      - name: Validate benchmark query set
+        run: |
+          python benchmarks/dotnet_bench.py \
+            --repo-dir /tmp/CleanArchitecture \
+            --validate-queries
+
+      - name: Run .NET retrieval benchmark with R@5 gate
+        run: |
+          python benchmarks/dotnet_bench.py \
+            --repo-dir /tmp/CleanArchitecture \
+            --out benchmarks/results_dotnet_bench_ci.json \
+            --fail-under-r5 ${{ env.R5_THRESHOLD }}
+
+      - name: Upload benchmark report
+        if: always()
+        uses: actions/upload-artifact@v6
+        with:
+          name: dotnet-bench-results
+          path: benchmarks/results_dotnet_bench_ci.json

{mempalace_code-1.6.2 → mempalace_code-1.8.0}/.github/workflows/publish.yml

@@ -17,7 +17,7 @@ jobs:
           python-version: "3.11"
       - run: pip install build
       - run: python -m build
-      - uses: actions/upload-artifact@v5
+      - uses: actions/upload-artifact@v6
         with:
           name: dist
           path: dist/
@@ -29,7 +29,7 @@ jobs:
     permissions:
       id-token: write
     steps:
-      - uses: actions/download-artifact@v5
+      - uses: actions/download-artifact@v7
         with:
           name: dist
           path: dist/

{mempalace_code-1.6.2 → mempalace_code-1.8.0}/CHANGELOG.md

@@ -1,5 +1,140 @@
 # Changelog
 
+Current command/package names: the CLI is `mempalace-code`, the import package is
+`mempalace_code`, and the MCP module is `python -m mempalace_code.mcp_server`.
+Older historical entries may mention legacy `mempalace` names that were valid
+when those changes landed.
+
+## v1.8.0 — 2026-05-03
+
+### Added
+
+- LanceDB storage cleanup: `mempalace-code cleanup` reclaims stale Lance
+  versions/fragments after optimization, with dry default retention and an
+  explicit `--unsafe-now` emergency mode.
+- Disk-budget guards for backup and watcher loops. Backups now fail before
+  opening archive files when projected post-backup free space would fall below
+  the configured floor; watchers skip write cycles while the palace is under
+  budget.
+- `mempalace-code watch <dir> status` reports free space, palace/backups size,
+  configured threshold, and macOS launchd state.
+- `code_search(..., rerank="hybrid")` and the MCP `mempalace_code_search`
+  `rerank` argument expose BM25-style token-overlap reranking for code search.
+- Conversation normalization now covers Gemini CLI JSONL and compacts Claude
+  Code tool-use/tool-result blocks.
+
+### Changed
+
+- .NET retrieval benchmark baseline on the pinned CleanArchitecture corpus is
+  now R@5 0.900 / R@10 1.000 in vector mode; local hybrid rerank comparison
+  measured R@5 1.000 / R@10 1.000 on 2026-05-03.
+- Generated `entities.json` files from init/entity detection are skipped during
+  project mining by default, unless explicitly force-included.
+- `MEMPALACE_BACKUP_MIN_FREE_BYTES` and `backup_min_free_bytes` remain supported
+  as compatibility aliases for the newer `backup_disk_min_free_bytes` floor.
+- The collected test suite is now 1769 tests.
+
+### Fixed
+
+- MCP request handling tolerates `params: null`, `arguments: null`,
+  notification messages, and client noise keys such as `wait_for_previous`.
+- Search/read paths tolerate rows with missing or `None` metadata without
+  crashing.
+
+## v1.7.0 — 2026-05-02
+
+### Added
+
+- Architecture extraction mode: `mempalace-code mine` now emits higher-level KG
+  facts for .NET (C#, F#, VB.NET) and Python projects, including pattern, layer,
+  namespace, and project membership facts.
+- Architecture fact refresh is now wing-scoped, so re-mining one project no
+  longer invalidates architecture KG facts from other projects in the same
+  palace.
+- Multi-repo palace sync: `mempalace-code mine-all` now mines multiple initialized
+  projects into one palace with one wing per project, explicit `wing:` overrides,
+  git-remote/folder wing auto-naming, duplicate-wing rejection before mining, and
+  incremental per-repo re-mining by default.
+- Release-grade coverage for backup/restore CLI dispatch, storage migration CLI
+  passthrough/error handling, Python multi-import dependency extraction,
+  devops/config file scanning, benchmark CI gates, watcher rule reloads, and
+  tree-sitter detached-comment behavior.
+
+### Changed
+
+- The shipped Python import namespace is `mempalace_code`; docs and generated
+  scheduler snippets prefer `mempalace-code` / `python -m mempalace_code` while
+  preserving the source-checkout `mempalace.mcp_server` compatibility shim.
+- Package metadata now matches the Python 3.11+ support floor.
+
+### Fixed
+
+- Watchers reload `scan_skip_dirs`, `scan_skip_files`, and `scan_skip_globs`
+  without restarting `mempalace-code watch`.
+- Generated backup/watch scheduler fallbacks no longer reference the legacy
+  `python -m mempalace` module path.
+
+## 2026-05-02 · ARCH-EXTRACTION-MODE
+
+Architecture extraction mode: `mempalace-code mine` now runs a post-mining pass
+that emits higher-level KG facts for .NET (C#, F#, VB.NET) and Python projects.
+
+### Added
+
+- `mempalace_code/architecture.py` — new module with pattern detection
+  (`is_pattern`: Service, Repository, Controller, ViewModel, Factory), layer
+  classification (`is_layer`: UI, Business, Data, Infrastructure), namespace
+  tagging (`in_namespace`), and project membership (`in_project`).
+- `KnowledgeGraph.invalidate_by_source_file` gains an optional `predicates`
+  parameter so architecture facts can be refreshed without expiring type
+  dependency facts (implements, inherits, depends_on, etc.).
+- Architecture config block in `mempalace.yaml` under `architecture:` — supports
+  custom `patterns` (name, suffixes, explicit `type_names`) and `layers`
+  (name, namespace_globs, type_suffixes, priority). Invalid rule entries are
+  silently ignored; the pass continues with built-in defaults.
+- KG queries: `entity="Service", direction="incoming"` shows all services;
+  `entity="Data", direction="incoming"` shows all data-layer types.
+
+## 2026-05-02 · FUT-MULTI-REPO
+
+Multi-repo palace sync: `mine-all` command mines multiple project directories into one palace with per-repo wing isolation, wing auto-naming from git remote/folder, and incremental per-repo re-mining.
+
+## 2026-05-01 · Recent completed task review
+
+### Added
+
+- `MINE-SCAN-RULES-LIVE-RELOAD`: watcher loops now reload `scan_skip_dirs`,
+  `scan_skip_files`, and `scan_skip_globs` between scan cycles, so app-level
+  exclude changes take effect without restarting `mempalace-code watch`.
+- `QUAL-E2E-REMAINING-MODULES`: end-to-end coverage now includes `convo_miner`,
+  `layers`, and `palace_graph` scenarios, including idempotent conversation
+  re-mining, tiered context loading, tunnel detection, traversal, and missing-room
+  boundary behavior.
+
+### Changed
+
+- `CLEAN-ONBOARDING`: `mempalace-code init` is config-file-first by default;
+  interactive guided setup is routed through explicit onboarding paths instead
+  of blocking the normal init flow.
+- The shipped Python import namespace is now `mempalace_code`. Packaged
+  `mempalace-code` installs no longer claim the top-level `mempalace` module,
+  allowing same-environment coexistence with upstream/vanilla MemPalace.
+- New MCP setup examples use `python -m mempalace_code.mcp_server`. Source
+  checkouts keep a minimal `mempalace.mcp_server` shim so older repo-local
+  Codex/Autopilot configs with checkout `PYTHONPATH` continue to start.
+
+### Fixed
+
+- `MINE-SCAN-GLOB-DIR-PRUNE`: glob rules that cover an entire generated
+  directory now prune that subtree during the walk instead of filtering only
+  after file discovery.
+- Legacy hook fallbacks and active docs now call `mempalace-code` or
+  `python -m mempalace_code`, matching the renamed package.
+
+## 2026-05-01 · MINE-APP-SCAN-EXCLUDES-PR4
+
+App-level scan excludes (`scan_skip_dirs`, `scan_skip_files`, `scan_skip_globs`) implemented in miner and watcher with hardened, tested outputs.
+
 ## v1.6.2 — 2026-05-01
 
 ### Added

{mempalace_code-1.6.2 → mempalace_code-1.8.0}/CLAUDE.md

@@ -45,16 +45,16 @@ python -m pytest tests/test_convo_miner.py -v
 
 ```bash
 # Check
-ruff check mempalace/ tests/
+ruff check mempalace_code/ tests/
 
 # Format check
-ruff format --check mempalace/ tests/
+ruff format --check mempalace_code/ tests/
 
 # Auto-fix lint
-ruff check --fix mempalace/ tests/
+ruff check --fix mempalace_code/ tests/
 
 # Auto-fix format
-ruff format mempalace/ tests/
+ruff format mempalace_code/ tests/
 ```
 
 Line length: 100. Target: py39. Quote style: double.
@@ -64,7 +64,7 @@ Line length: 100. Target: py39. Quote style: double.
 | Module | Purpose |
 |--------|---------|
 | `storage.py` | LanceDB vector storage — add, search, delete, health_check, recover |
-| `backup.py` | Tarball backup/restore — `mempalace backup`, scheduled backups |
+| `backup.py` | Tarball backup/restore — `mempalace-code backup`, scheduled backups |
 | `miner.py` | Code project miner — walks source files, extracts drawers |
 | `convo_miner.py` | Conversation miner — ingests Claude/ChatGPT/Slack exports |
 | `searcher.py` | Semantic search — query palace with optional wing/room filters |
@@ -73,7 +73,7 @@ Line length: 100. Target: py39. Quote style: double.
 | `palace_graph.py` | Graph traversal and tunnel detection across wings/rooms |
 | `mcp_server.py` | MCP server — exposes palace tools to Claude Code and other MCP clients |
 | `watcher.py` | File watcher — `watch_and_mine`, `watch_all`, launchd/cron schedule rendering |
-| `cli.py` | `mempalace` CLI entry point — init, mine, mine-all, watch, search, health, repair, backup |
+| `cli.py` | `mempalace-code` CLI entry point — init, mine, mine-all, watch, search, health, repair, backup |
 
 ## Architecture Principles
 
@@ -130,3 +130,16 @@ Per-category R@5:
   - `chore:` — maintenance, deps, tooling
 - **No force-push to `main`**.
 - PR merges go through the `feat/*` → `main` flow; squash if the branch is noisy.
+
+## Operational Lessons
+
+- **Keep this file public-safe.** `CLAUDE.md` is part of the publishable repository. Do not write private remotes, hostnames, credentials, local machine paths, customer/project details, incident specifics, or non-public operational history here. Put private or machine-local lessons in a local-only note outside the published tree.
+- **Record reusable lessons only when they are public knowledge.** When a session exposes a project gotcha, publish step, verification boundary, or agent-behavior correction, add a concise durable note here only if it is safe for public readers and useful to future contributors.
+- **Verify the environment that will actually run the change.** GitHub Actions runtime changes are not proven by Python tests alone. Use local YAML/static checks such as `actionlint`, then verify the real hosted workflow run when action runtime behavior matters.
+- **Name the verification boundary.** If a workflow is tag-only or release-only, say that it was syntax-checked and version-checked but not execution-tested unless a real trigger was run. Do not imply full local coverage for hosted-only behavior.
+- **Check the intended public release target.** Before publishing, verify the repository, branch, tag, and workflow that public users will see. Do not assume local remote names or private mirrors represent public release truth.
+- **Keep benchmark gates tied to measured baselines.** If a release benchmark fails, reproduce it locally against the pinned fixture, update the CI threshold only to the observed stable baseline, and backlog any desired quality increase separately.
+- **Do not call tests "local feature testing."** When asked to test new features locally, run the public CLI/MCP/API behavior itself, not only pytest. For each new feature, exercise at least one success path and one important failure/guard path when safe, record the exact command or request, and name any behavior that was covered only by tests.
+- **Exercise real integration surfaces before release claims.** Direct handler calls are useful for MCP compatibility, but they are not the same as a separate stdio MCP client. CLI help is not the same as executing the command. A release-readiness summary must distinguish unit tests, focused integration tests, direct API smoke, real CLI execution, and hosted/daemon behavior that was not run.
+- **Clean up smoke-test artifacts immediately.** Real backup, cleanup, and benchmark smokes can create archives, temp palaces, and result JSON. Put them under disposable temp paths, verify success/failure, then remove artifacts or explicitly report what remains.
+- **Treat palace disk growth as storage forensics first.** Compare backup size, live storage stats, row counts, and cleanup output before deleting anything. Preserve non-regenerable manual drawers and KG data, stop active writers when needed, use supported cleanup APIs, and verify health/status afterward.