MCP Registry listing, Intel-Mac indexing fix, README first screen (v0.3.1)

.github/workflows/publish.yml

@@ -44,7 +44,7 @@ jobs:
         run: |
           python -m venv /tmp/clean
           /tmp/clean/bin/pip install dist/*.whl
-          for cmd in sgm-index sgm-ask sgm-mcp sgm-turnstate-hook sgm-turnstate-show; do
+          for cmd in sgm-index sgm-ask sgm-mcp sgm-turnstate-hook sgm-turnstate-show sqlite-graph-memory; do
             test -x "/tmp/clean/bin/$cmd" || { echo "::error::$cmd missing from wheel"; exit 1; }
           done
       - uses: actions/upload-artifact@v4
@@ -66,3 +66,37 @@ jobs:
       - uses: pypa/gh-action-pypi-publish@release/v1
         with:
           skip-existing: true   # the package may already be uploaded before the GitHub release object exists
+
+  # List the server in the official MCP Registry once PyPI has the release.
+  # Auth is GitHub OIDC (no stored secret): the registry grants the
+  # io.github.tonydzi/* namespace to workflows running in tonydzi's repos.
+  # Ownership is proven by the `mcp-name:` line in README.md, which PyPI shows
+  # as the package description, so this job must run AFTER the PyPI upload.
+  mcp-registry:
+    needs: publish
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+      - name: server.json must describe the version being released
+        run: |
+          pkg=$(python3 -c "import tomllib;print(tomllib.load(open('pyproject.toml','rb'))['project']['version'])")
+          srv=$(python3 -c "import json;d=json.load(open('server.json'));p=d['packages'][0];print(d['version'], p['version'], p['runtimeArguments'][0]['value'])")
+          [ "$srv" = "$pkg $pkg sqlite-graph-memory[embeddings]==$pkg" ] || { echo "::error::server.json says '$srv', pyproject says $pkg"; exit 1; }
+          grep -q "mcp-name: $(python3 -c "import json;print(json.load(open('server.json'))['name'])")" README.md \
+            || { echo "::error::README.md lacks the mcp-name line the registry verifies"; exit 1; }
+      - name: Wait until PyPI serves this version
+        run: |
+          v=$(python3 -c "import json;print(json.load(open('server.json'))['version'])")
+          for i in $(seq 1 30); do
+            curl -sf "https://pypi.org/pypi/sqlite-graph-memory/$v/json" >/dev/null && exit 0
+            sleep 20
+          done
+          echo "::error::PyPI never served $v"; exit 1
+      - name: Install mcp-publisher
+        run: |
+          curl -sL "https://github.com/modelcontextprotocol/registry/releases/latest/download/mcp-publisher_linux_amd64.tar.gz" | tar xz mcp-publisher
+      - run: ./mcp-publisher login github-oidc
+      - run: ./mcp-publisher publish

.github/workflows/tests.yml

@@ -63,7 +63,7 @@ jobs:
           /tmp/clean/bin/pip install dist/*.whl
       - name: every advertised command must start
         run: |
-          for cmd in sgm-index sgm-ask sgm-turnstate-hook sgm-turnstate-show; do
+          for cmd in sgm-index sgm-ask sgm-mcp sgm-turnstate-hook sgm-turnstate-show sqlite-graph-memory; do
             test -x "/tmp/clean/bin/$cmd" || { echo "::error::$cmd was not installed"; exit 1; }
           done
           /tmp/clean/bin/sgm-turnstate-show --help > /dev/null

CHANGELOG.md

@@ -4,14 +4,31 @@ What shipped, in plain words. Entries below v0.1.3 were written on 2026-09-05 fr
 release notes that already existed — the file itself did not exist until then, so those three are
 recorded after the fact rather than backdated to look contemporaneous.
 
-## Unreleased
+## v0.3.1 — unreleased
 
 **Credit where it is due.** The MCP server that ships as `sgm-mcp` was written by Vardhman Gupta
 ([@Kaap10](https://github.com/Kaap10)) in PR #8: the server, its tests and its docs. His
 follow-up fixes are the shared answer-file race (commit 3ce7061), the stdout fallback (PR #10)
 and the swallowed read errors (PR #14).
 He is now listed in the `authors` field of `pyproject.toml`, which is the attribution he asked for.
 
+**Indexing works on Intel Macs again.** A fresh `pip install 'sqlite-graph-memory[embeddings]'`
+on an x86_64 Mac with Python 3.12 crashed `sgm-index` with `NameError: name 'nn' is not
+defined`. Torch stops at 2.2 there, transformers 5.x needs 2.6 and quietly disables torch,
+and sentence-transformers 6.x requires transformers 5.x. The embeddings extra now caps
+transformers below 4.50 and sentence-transformers below 5 on that one platform. Measured
+both ways on the same machine: 0.3.0 from PyPI crashes, this build indexes the 202-note
+bench vault cleanly (204 chunks).
+
+**Listed in the official MCP Registry.** `server.json` describes the server as
+`io.github.tonydzi/sqlite-graph-memory`, the README carries the `mcp-name` line the registry
+checks, and the publish workflow lists the release after PyPI has it. The package also
+installs the MCP server under its own name, `sqlite-graph-memory`, so registry clients can
+start it with `uvx`.
+
+**The README opens with three commands and their real output** on the bench vault, including
+the honest caveat that vector-only ranks ahead of graph mode on that vault overall.
+
 ## bench-v0.1 — 2026-10-09
 
 **A benchmark other people can run.** Until now every retrieval number in this repo came from

README.md

@@ -1,13 +1,60 @@
 # sqlite-graph-memory
 
+<!-- mcp-name: io.github.tonydzi/sqlite-graph-memory -->
+
 **Your agent forgets everything between sessions, and plain vector search gives it back
-flat, unrelated chunks. This gives it associative recall over your own markdown notes —
-one SQLite file, no graph database, no ETL pass, no server.**
+flat, unrelated chunks. This gives it associative recall over your own markdown notes:
+vector search finds the entry points, the `[[wikilinks]]` you already wrote are the graph.
+One SQLite file, no graph database, no ETL pass, no server.**
+
+## Three commands
 
 ```bash
-pip install sqlite-graph-memory
+pip install 'sqlite-graph-memory[embeddings]'
+sgm-index ./notes
+sgm-ask --ab "In which programming language is the tool behind Project Ouzel written?"
+```
+
+`./notes` is any folder of markdown notes. To get exactly the output below, use the
+202-note synthetic vault from this repo (it is not inside the pip package):
+
+```bash
+git clone --depth 1 https://github.com/tonydzi/sqlite-graph-memory && cp -R sqlite-graph-memory/bench/vault ./notes
+```
+
+Real output, top 5 of 12:
+
+```
+--- DIRECT memory (vector) ---
+ 1. [  1.66] project-ouzel
+ 2. [ -1.33] project-avocet
+ 3. [ -1.79] tool-cobbleset
+ 4. [ -2.44] meeting-2025-10-21-lighthouse-sync
+ 5. [ -2.70] person-severin-ibarrondo
+
+--- ASSOCIATIVE memory (vector+graph) ---
+ 1. [  1.66] project-ouzel
+ 2. [  1.14] tool-lanternfish  <== +graph
+ 3. [ -1.33] project-avocet
+ 4. [ -1.79] tool-cobbleset
+ 5. [ -2.44] meeting-2025-10-21-lighthouse-sync
 ```
 
+The answer lives in `tool-lanternfish` ("Lanternfish is written in Elixir"). The question
+never names Lanternfish, so the vector pass does not find it in the top 12 at all.
+`project-ouzel` links to it, so one graph hop puts it second.
+
+That is the idea in one query, and it is not a blanket win. On this same vault vector-only
+ranks ahead of graph mode overall; the numbers are in [`bench/`](bench/README.md). `--ab`
+exists so you can measure which one wins on your own notes.
+
+**From any MCP client** (Claude Desktop, Cursor, Zed, Windsurf): `sgm-mcp` after the install
+above, or `uvx --with 'sqlite-graph-memory[embeddings]' sqlite-graph-memory`. Set
+`BRAIN_INDEX_DIR` to the `index` folder `sgm-index` wrote. Config:
+[`examples/mcp-config.json`](examples/mcp-config.json).
+
+## What it is
+
 Graph RAG on SQLite for AI agents — a working pilot, not a framework. It is the extracted
 memory layer of a personal "second brain" agent setup: vector search finds the entry
 points, the `[[wikilinks]]` you already wrote by hand are the graph, and a cross-encoder

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "sqlite-graph-memory"
-version = "0.3.0"
+version = "0.3.1"
 description = "Associative memory for LLM agents: Graph RAG over a folder of markdown notes, in one SQLite file, with no server to run."
 readme = "README.md"
 requires-python = ">=3.9"
@@ -47,9 +47,16 @@ dependencies = ["numpy>=1.24"]
 # Intel Macs top out at torch 2.2.x, which is built against NumPy 1.x and dies on
 # NumPy 2 with "RuntimeError: Numpy is not available" (measured 2026-10-15 on an
 # x86_64 Mac, Python 3.9: sgm-index crashed; with numpy<2 it indexed cleanly).
+# The same ceiling bites one layer up: transformers 5.x needs torch >= 2.6, so on
+# torch 2.2 it disables torch and sgm-index dies with "NameError: name 'nn' is not
+# defined" inside transformers/integrations/accelerate.py (measured 2026-10-15 on
+# an x86_64 Mac, Python 3.12, fresh `pip install` from PyPI). sentence-transformers
+# 6.x in turn requires transformers 5.x, so both are capped there.
 embeddings = [
   "sentence-transformers>=2.6",
   "numpy<2; sys_platform == 'darwin' and platform_machine == 'x86_64'",
+  "transformers<4.50; sys_platform == 'darwin' and platform_machine == 'x86_64'",
+  "sentence-transformers<5; sys_platform == 'darwin' and platform_machine == 'x86_64'",
 ]
 
 [project.urls]
@@ -64,6 +71,9 @@ sgm-ask = "sqlite_graph_memory.brain_ask:main"
 sgm-turnstate-hook = "sqlite_graph_memory.turnstate_hook:main"
 sgm-turnstate-show = "sqlite_graph_memory.turnstate_show:main"
 sgm-mcp = "sqlite_graph_memory.mcp_server:main"
+# Same server under the package's own name, so `uvx sqlite-graph-memory` (what MCP
+# registry clients run for a PyPI package) starts it without a --from flag.
+sqlite-graph-memory = "sqlite_graph_memory.mcp_server:main"
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/sqlite_graph_memory"]

server.json

@@ -0,0 +1,45 @@
+{
+  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
+  "name": "io.github.tonydzi/sqlite-graph-memory",
+  "title": "sqlite-graph-memory",
+  "description": "Recall over your markdown notes: vector search plus your [[wikilinks]] as a graph, in SQLite.",
+  "repository": {
+    "url": "https://github.com/tonydzi/sqlite-graph-memory",
+    "source": "github"
+  },
+  "version": "0.3.1",
+  "packages": [
+    {
+      "registryType": "pypi",
+      "registryBaseUrl": "https://pypi.org",
+      "identifier": "sqlite-graph-memory",
+      "version": "0.3.1",
+      "runtimeHint": "uvx",
+      "runtimeArguments": [
+        {
+          "type": "named",
+          "name": "--with",
+          "value": "sqlite-graph-memory[embeddings]==0.3.1",
+          "description": "Recall needs the embedding and reranker models; the base package leaves them out to stay small."
+        }
+      ],
+      "transport": {
+        "type": "stdio"
+      },
+      "environmentVariables": [
+        {
+          "name": "BRAIN_INDEX_DIR",
+          "description": "Directory holding the index. Build it once with `sgm-index /path/to/notes`.",
+          "format": "filepath",
+          "isRequired": true
+        },
+        {
+          "name": "TURNSTATE_DB",
+          "description": "SQLite file for recall telemetry.",
+          "format": "filepath",
+          "default": "./turnstate.db"
+        }
+      ]
+    }
+  ]
+}

src/sqlite_graph_memory/__init__.py

@@ -21,6 +21,6 @@
 on an ``ImportError`` at startup.
 """
 
-__version__ = "0.3.0"
+__version__ = "0.3.1"
 
 __all__ = ["__version__"]

src/sqlite_graph_memory/mcp_server.py

@@ -22,6 +22,8 @@
 import sys, os, json, subprocess, tempfile
 from pathlib import Path
 
+from sqlite_graph_memory import __version__
+
 # Ensure UTF-8 for stdio communication on all platforms (including Windows console)
 try:
     sys.stdin.reconfigure(encoding='utf-8')
@@ -43,7 +45,7 @@ def _brain_ask_cmd():
 
 SERVER_INFO = {
     "name": "sqlite-graph-memory",
-    "version": "0.3.0"
+    "version": __version__
 }
 
 TOOLS = [

tests/test_mcp_server.py

@@ -12,6 +12,11 @@
 
 
 
+def test_server_info_version_tracks_package():
+    from sqlite_graph_memory import __version__
+    assert SERVER_INFO["version"] == __version__
+
+
 def test_mcp_initialize():
     req = {
         "jsonrpc": "2.0",