feat(algolia): 生成 easyops 格式 search-index.json 并覆盖上传 OSS

.github/workflows/algolia-index.yml

@@ -46,3 +46,28 @@ jobs:
             exit 1
           fi
           npm run algolia:index
+
+      - name: Setup ossutil v2
+        run: |
+          if [ ! -x "$HOME/ossutil" ]; then
+            curl -o ossutil.zip https://gosspublic.alicdn.com/ossutil/v2/2.1.2/ossutil-2.1.2-linux-amd64.zip
+            unzip ossutil.zip
+            cd ossutil-2.1.2-linux-amd64
+            chmod 755 ossutil
+            mv ossutil "$HOME/ossutil"
+          fi
+
+      # 覆盖式单文件上传：覆盖 OSS 上残留的旧 search-index.json（8 月过期产物）。
+      # 不能带 -r / --update（那只会增不删），必须覆盖根下同名文件。
+      - name: Upload easyops search-index.json to OSS
+        env:
+          OSS_ACCESS_KEY_ID: ${{ secrets.OSS_ACCESS_KEY_ID }}
+          OSS_ACCESS_KEY_SECRET: ${{ secrets.OSS_ACCESS_KEY_SECRET }}
+        run: |
+          ~/ossutil cp scripts/_algolia_tmp/search-index.json \
+            oss://cn-rdk-front-prod/case_doc/search-index.json \
+            -e https://d-robotics.cc \
+            -i $OSS_ACCESS_KEY_ID \
+            -k $OSS_ACCESS_KEY_SECRET \
+            --region cn-beijing \
+            --retry-times=3

scripts/algolia-index.mjs

@@ -16,8 +16,14 @@
  */
 import "dotenv/config";
 import { createHash } from "node:crypto";
+import { existsSync, mkdirSync, writeFileSync } from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
 import { load } from "cheerio";
 
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const rootDir = path.resolve(__dirname, "..");
+
 const APP_ID = process.env.ALGOLIA_APP_ID || "1VU781LYTV";
 const INDEX_NAME = process.env.ALGOLIA_INDEX_NAME || "case_doc";
 const SITE_URL = ensureTrailingSlash(
@@ -638,6 +644,124 @@ async function smokeQuery() {
   }
 }
 
+function relativePathFromUrl(url) {
+  try {
+    return new URL(url).pathname;
+  } catch {
+    return url;
+  }
+}
+
+function deepestHeadingText(hierarchy) {
+  for (const lvl of ["lvl6", "lvl5", "lvl4", "lvl3", "lvl2", "lvl1"]) {
+    const text = (hierarchy?.[lvl] || "").trim();
+    if (text) {
+      return text;
+    }
+  }
+  return "";
+}
+
+/**
+ * 把 Algolia DocSearch records 映射成 easyops `search-index.json` 同构格式（5 shard），
+ * 供 rdk-docs-mcp 读取（仅 zh-Hans）。
+ */
+function buildEasyopsIndex(records) {
+  const zh = records.filter((r) => r.language === "zh-Hans");
+  const byPage = new Map();
+  for (const r of zh) {
+    const pagePath = relativePathFromUrl(r.url_without_anchor);
+    if (!byPage.has(pagePath)) {
+      byPage.set(pagePath, []);
+    }
+    byPage.get(pagePath).push(r);
+  }
+
+  const pages = [];
+  const headings = [];
+  const pageSnippets = [];
+  const contentSnippets = [];
+  let id = 0;
+  let pageSeq = 0;
+
+  for (const [u, recs] of byPage) {
+    pageSeq += 1;
+    let pageTitle = "";
+    for (const r of recs) {
+      if (r.type === "lvl1" && r.hierarchy?.lvl1) {
+        pageTitle = r.hierarchy.lvl1.trim();
+        break;
+      }
+    }
+    const breadcrumbs = (recs[0]?.hierarchy?.lvl0 || "")
+      .split(" / ")
+      .map((s) => s.trim())
+      .filter(Boolean);
+    if (!pageTitle) {
+      pageTitle = breadcrumbs[breadcrumbs.length - 1] || "";
+    }
+    pages.push({ i: ++id, t: pageTitle, u, b: breadcrumbs });
+
+    let firstHeading = "";
+    for (const r of recs) {
+      if (
+        r.type &&
+        r.type.startsWith("lvl") &&
+        r.type !== "lvl1" &&
+        r.hierarchy?.[r.type]
+      ) {
+        firstHeading = r.hierarchy[r.type].trim();
+        break;
+      }
+    }
+    if (firstHeading) {
+      pageSnippets.push({ i: ++id, t: firstHeading, s: pageTitle, u, p: pageSeq });
+    }
+
+    const seenHeadings = new Set();
+    const seenContent = new Set();
+    for (const r of recs) {
+      const type = r.type;
+      if (type && type.startsWith("lvl") && type !== "lvl1") {
+        if (r.anchor && !seenHeadings.has(r.anchor)) {
+          seenHeadings.add(r.anchor);
+          headings.push({
+            i: ++id,
+            t: (r.hierarchy?.[type] || "").trim(),
+            u,
+            h: `#${r.anchor}`,
+            p: pageSeq,
+          });
+        }
+      } else if (type === "content" && r.content) {
+        const text = r.content.trim();
+        const key = `${r.anchor || ""}\u0000${text}`;
+        if (seenContent.has(key)) {
+          continue;
+        }
+        seenContent.add(key);
+        const doc = { i: ++id, t: text, u, p: pageSeq };
+        const source = deepestHeadingText(r.hierarchy);
+        if (source) {
+          doc.s = source;
+        }
+        if (r.anchor) {
+          doc.h = `#${r.anchor}`;
+        }
+        contentSnippets.push(doc);
+      }
+    }
+  }
+
+  return [
+    { documents: pages },
+    { documents: headings },
+    { documents: pageSnippets },
+    { documents: [] },
+    { documents: contentSnippets },
+  ];
+}
+
 async function main() {
   if (!ADMIN_API_KEY) {
     console.error(
@@ -662,6 +786,17 @@ async function main() {
     process.exit(1);
   }
 
+  const easyopsIndex = buildEasyopsIndex(records);
+  const easyopsPath = path.join(rootDir, "scripts/_algolia_tmp/search-index.json");
+  mkdirSync(path.dirname(easyopsPath), { recursive: true });
+  writeFileSync(easyopsPath, JSON.stringify(easyopsIndex));
+  console.log(
+    `Wrote easyops search-index.json ` +
+      `(${easyopsIndex[0].documents.length} pages, ` +
+      `${easyopsIndex[1].documents.length} headings, ` +
+      `${easyopsIndex[4].documents.length} content snippets) to ${easyopsPath}`,
+  );
+
   if (DRY_RUN) {
     const sample = records[0];
     console.log("Dry-run sample record:", JSON.stringify(sample, null, 2));