AI向けに llms.txt と llms-full.txt を生成する仕組みを追加

AI（LLM）がサイトを参照・学習しやすくするため、以下を追加:

- public/llms.txt: 全ページをカテゴリ別に列挙した AI 向けサイトマップ
- scripts/generate-llms-full.mjs: 全 MDX ファイルの本文を収集・結合する生成スクリプト
- package.json の build スクリプトに生成処理を追加

ビルド時に dist/llms.txt と dist/llms-full.txt が自動生成される。

3 files changed, 196 insertions, 1 deletions

diff --git a/package.json b/package.json
index 5fa0846..ad8b7c0 100644
--- a/package.json
+++ b/package.json
@@ -5,7 +5,7 @@
   "scripts": {
     "dev": "astro dev",
     "start": "astro dev",
-    "build": "astro build && py -c \"import shutil; shutil.copy('public/.htaccess', 'dist/.htaccess')\" && py removeNullCharacters.py",
+    "build": "astro build && py -c \"import shutil; shutil.copy('public/.htaccess', 'dist/.htaccess')\" && py removeNullCharacters.py && node scripts/generate-llms-full.mjs",
     "deploy": "sh deploy.sh",
     "preview": "astro preview",
     "postinstall": "patch-package",
diff --git a/public/llms.txt b/public/llms.txt
new file mode 100644
index 0000000..bfbe905
--- /dev/null
+++ b/public/llms.txt
@@ -0,0 +1,51 @@
+# 安竹洋平 公式サイト（小平市議会議員）
+> 小平市議会議員 安竹洋平の公式サイトです。一般質問や議会での活動、いじめ、発達関連の情報をまとめています。
+
+## 基本情報
+- サイトURL: https://yasutakeyohei.com/
+- 運営者: 安竹洋平（小平市議会議員、完全無所属）
+- お問い合わせ: yohei@yasutakeyohei.com
+
+## メインページ
+- ホーム: https://yasutakeyohei.com/
+- 実績概要: https://yasutakeyohei.com/jisseki/
+- 私の方針: https://yasutakeyohei.com/policy/
+- ご支援: https://yasutakeyohei.com/support/
+- コンタクト: https://yasutakeyohei.com/contact/
+
+## 実績
+- 公文書管理の不正追及の軌跡: https://yasutakeyohei.com/koubunsyo-kanri/
+- いじめ重大事態への対応の軌跡: https://yasutakeyohei.com/ijime-judai-jitai/
+- 障害者福祉施設における虐待通報対応の軌跡: https://yasutakeyohei.com/fukushi-shisetsu-gyakutai/
+- 合気公園の軌跡: https://yasutakeyohei.com/aiki-kouen/
+- 情緒固定級の軌跡: https://yasutakeyohei.com/joutyo-koteikyu/
+- 過剰な制限緩和の軌跡: https://yasutakeyohei.com/kajo-seigen-kanwa/
+- 東京サレジオ学園北側開発問題の軌跡: https://yasutakeyohei.com/saresio-kaihatu/
+- ワクチン副反応救済制度の適正化の軌跡: https://yasutakeyohei.com/vaccine-kyuusai-tekiseika/
+- ディスレクシア（読み書き障害）対応の軌跡: https://yasutakeyohei.com/dislexia-taiou/
+
+## 一般質問
+- 一覧: https://yasutakeyohei.com/ippan-situmon/
+- 令和7年度: https://yasutakeyohei.com/ippan-situmon/r7d/
+- 令和6年度: https://yasutakeyohei.com/ippan-situmon/r6d/
+- 令和5年度: https://yasutakeyohei.com/ippan-situmon/r5d/
+- 令和4年度: https://yasutakeyohei.com/ippan-situmon/r4d/
+- 令和3年度: https://yasutakeyohei.com/ippan-situmon/r3d/
+- 令和2年度: https://yasutakeyohei.com/ippan-situmon/r2d/
+- 令和元年度: https://yasutakeyohei.com/ippan-situmon/r1d/
+
+## ディスレクシアについて
+- 概要: https://yasutakeyohei.com/about-dyslexia/
+- ディスレクシアとは: https://yasutakeyohei.com/about-dyslexia/what-is-dyslexia/
+- ディスレクシアの著名人: https://yasutakeyohei.com/about-dyslexia/celebrities/
+- 本人や保護者の声: https://yasutakeyohei.com/about-dyslexia/voices/
+- 潜在的人数と文科省調査: https://yasutakeyohei.com/about-dyslexia/potential-number/
+- 見過ごさないためのアセスメント: https://yasutakeyohei.com/about-dyslexia/assessment/
+- デイジー教科書と音声教材: https://yasutakeyohei.com/about-dyslexia/daisy-and-onsei/
+- 合理的配慮と関連法律: https://yasutakeyohei.com/about-dyslexia/reasonable-accommodation/
+- 小平市の状況: https://yasutakeyohei.com/about-dyslexia/kodaira/
+- 就学相談・通級指導: https://yasutakeyohei.com/about-dyslexia/school-consultation/
+- 情緒固定級とは: https://yasutakeyohei.com/about-dyslexia/emotional-support-class/
+- チャレンジスクール等: https://yasutakeyohei.com/about-dyslexia/high-school-options/
+- 長期欠席支援シート: https://yasutakeyohei.com/about-dyslexia/support-sheet/
+- 支援リソース・参考情報: https://yasutakeyohei.com/about-dyslexia/support-resources/
diff --git a/scripts/generate-llms-full.mjs b/scripts/generate-llms-full.mjs
new file mode 100644
index 0000000..cc3fdd4
--- /dev/null
+++ b/scripts/generate-llms-full.mjs
@@ -0,0 +1,144 @@
+import { readFileSync, writeFileSync } from "node:fs";
+import { readdir, readFile } from "node:fs/promises";
+import { join, extname, relative } from "node:path";
+
+const DOCS_DIR = join(import.meta.dirname, "..", "src", "content", "docs");
+const OUTPUT_FILE = join(import.meta.dirname, "..", "dist", "llms-full.txt");
+const SITE_URL = "https://yasutakeyohei.com";
+
+/**
+ * Recursively collect all .mdx files, excluding files starting with "_" (partials).
+ */
+async function collectMdxFiles(dir) {
+  const entries = await readdir(dir, { withFileTypes: true });
+  const files = [];
+
+  for (const entry of entries) {
+    if (entry.name.startsWith("_")) continue;
+    const fullPath = join(dir, entry.name);
+    if (entry.isDirectory()) {
+      const subFiles = await collectMdxFiles(fullPath);
+      files.push(...subFiles);
+    } else if (extname(entry.name) === ".mdx") {
+      files.push(fullPath);
+    }
+  }
+
+  return files;
+}
+
+/**
+ * Derive the page URL from the file path.
+ * Converts Starlight docs convention to URL path.
+ */
+function pathToUrl(filePath) {
+  const relPath = relative(DOCS_DIR, filePath).replace(/\\/g, "/");
+  let urlPath = relPath.replace(/\.mdx$/, "");
+
+  // index.mdx → directory URL
+  if (urlPath.endsWith("/index")) {
+    urlPath = urlPath.replace(/\/index$/, "/");
+  }
+
+  // Root index → /
+  if (urlPath === "index") {
+    urlPath = "";
+  }
+
+  return `${SITE_URL}/${urlPath}`;
+}
+
+/**
+ * Strip frontmatter (content between --- delimiters).
+ * Returns the body content after the second "---" line.
+ */
+function stripFrontmatter(content) {
+  // Match frontmatter delimited by --- at the start of the file
+  const match = content.match(/^---\r?\n[\s\S]*?\r?\n---/);
+  if (match) {
+    return content.slice(match[0].length).trimStart();
+  }
+  return content;
+}
+
+/**
+ * Strip JSX components (self-closing and paired tags with uppercase names).
+ * This is a simple heuristic that removes Astro/MDX components.
+ */
+function stripJsxComponents(content) {
+  // Remove import/export statements
+  content = content.replace(
+    /^import\s+[\s\S]*?(?:from\s+['"][^'"]+['"]|['"][^'"]+['"])\s*;?\s*$/gm,
+    "",
+  );
+  content = content.replace(
+    /^export\s+(?:const|let|var|function|default|async)\s+[\s\S]*?$/gm,
+    "",
+  );
+
+  return content;
+}
+
+/**
+ * Extract the title from the frontmatter of an MDX file.
+ */
+function extractTitle(content) {
+  const match = content.match(/^---\r?\n([\s\S]*?)\r?\n---/);
+  if (!match) return null;
+  const frontmatter = match[1];
+  const titleMatch = frontmatter.match(/^title:\s*(.+)$/m);
+  return titleMatch ? titleMatch[1].trim().replace(/['"]/g, "") : null;
+}
+
+async function main() {
+  console.log("Collecting MDX files...");
+  const files = await collectMdxFiles(DOCS_DIR);
+
+  // Sort: top-level files first, then by path
+  files.sort((a, b) => {
+    const aRel = relative(DOCS_DIR, a).replace(/\\/g, "/");
+    const bRel = relative(DOCS_DIR, b).replace(/\\/g, "/");
+    const aDepth = (aRel.match(/\//g) || []).length;
+    const bDepth = (bRel.match(/\//g) || []).length;
+    if (aDepth !== bDepth) return aDepth - bDepth;
+    return aRel.localeCompare(bRel);
+  });
+
+  console.log(`Found ${files.length} MDX files.`);
+
+  const sections = [];
+
+  for (const file of files) {
+    const rawContent = readFileSync(file, "utf-8");
+    const title = extractTitle(rawContent) || pathToUrl(file);
+    const url = pathToUrl(file);
+    let body = stripFrontmatter(rawContent);
+    body = stripJsxComponents(body);
+
+    // Skip empty or near-empty pages
+    if (body.trim().length < 10) {
+      console.log(`  Skipping (too short): ${relative(DOCS_DIR, file)}`);
+      continue;
+    }
+
+    sections.push(`# ${title}\n> ${url}\n\n${body.trim()}\n\n---\n`);
+  }
+
+  const fullContent = [
+    "# 安竹洋平 公式サイト - 全コンテンツ\n",
+    `> このファイルは AI（LLM）による学習・参照用に自動生成されています。\n`,
+    `> 生成元: ${SITE_URL}\n`,
+    `> 更新日: ${new Date().toISOString().split("T")[0]}\n\n`,
+    ...sections,
+  ].join("");
+
+  writeFileSync(OUTPUT_FILE, fullContent, "utf-8");
+  console.log(
+    `Generated: ${OUTPUT_FILE} (${fullContent.length.toLocaleString()} chars)`,
+  );
+}
+
+main().catch((err) => {
+  console.error("Error:", err);
+  process.exit(1);
+});