Harden web search and docs defaults

docker/web-search/.dockerignore

@@ -1,2 +1,5 @@
 *
 !Dockerfile
+!patch-mcp-web-search.mjs
+!overrides/
+!overrides/bing.js

docker/web-search/Dockerfile

@@ -1,7 +1,14 @@
-FROM node:22-bookworm-slim
+FROM node:22-bookworm-slim@sha256:813a7480f28fdadac1f7f5c824bcdad435b5bc1322a5968bbbdef8d058f9dff4
 
 ARG MCP_WEB_SEARCH_VERSION=1.3.0
+ARG MCP_WEB_SEARCH_MAX_BYTES=52428800
 
+COPY patch-mcp-web-search.mjs /tmp/patch-mcp-web-search.mjs
+COPY overrides/bing.js /tmp/context-kit-bing-provider.js
+
+# Chromium intentionally tracks Debian security updates inside the pinned base
+# image family; Bing's browser path is more likely to break with stale Chromium
+# than with patched OS packages.
 RUN apt-get update \
   && apt-get install -y --no-install-recommends \
     ca-certificates \
@@ -10,11 +17,15 @@ RUN apt-get update \
   && rm -rf /var/lib/apt/lists/*
 
 RUN npm install -g "@zhafron/mcp-web-search@${MCP_WEB_SEARCH_VERSION}" \
+  && cp /tmp/context-kit-bing-provider.js /usr/local/lib/node_modules/@zhafron/mcp-web-search/dist/src/providers/bing.js \
+  && node /tmp/patch-mcp-web-search.mjs \
+  && rm /tmp/patch-mcp-web-search.mjs /tmp/context-kit-bing-provider.js \
   && npm cache clean --force
 
 ENV CHROME_PATH=/usr/bin/chromium \
   DEFAULT_SEARCH_PROVIDER=searxng \
   HTTP_TIMEOUT=15000 \
+  MAX_BYTES=${MCP_WEB_SEARCH_MAX_BYTES} \
   MAX_RESULTS=10 \
   SEARXNG_URL=http://searxng:8080
 

docker/web-search/overrides/bing.js

@@ -0,0 +1,114 @@
+import { PUPPETEER_TIMEOUT } from "../constants.js";
+import { browserPool } from "../utils/browser-pool.js";
+import { getAcceptLanguageHeader, getMarketFromLang } from "../utils/user-agent.js";
+import { searchCache, createCacheKey } from "../utils/cache.js";
+
+// Context Kit override for @zhafron/mcp-web-search 1.3.0.
+// The upstream provider can read Bing before result cards render and return an
+// empty fallback. Keep this as a direct provider replacement until upstream
+// waits for cards and decodes current /ck/a redirects reliably.
+const DEFAULT_BROWSER_SEARCH_USER_AGENT = process.env.BROWSER_SEARCH_USER_AGENT ||
+  "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36";
+
+function decodeBase64Url(value) {
+  const normalized = value.replace(/-/g, "+").replace(/_/g, "/");
+  const padded = normalized.padEnd(normalized.length + ((4 - normalized.length % 4) % 4), "=");
+  return Buffer.from(padded, "base64").toString("utf-8");
+}
+
+export class BingProvider {
+  name = "bing";
+
+  decodeBingRedirect(href) {
+    try {
+      const url = new URL(href, "https://www.bing.com/");
+      if (url.hostname === "www.bing.com" && url.pathname === "/ck/a") {
+        const encoded = url.searchParams.get("u");
+        if (encoded) {
+          const candidates = [encoded];
+          if (/^[a-z][0-9]/i.test(encoded)) candidates.push(encoded.slice(2));
+          for (const candidate of candidates) {
+            try {
+              const decoded = decodeBase64Url(candidate);
+              if (/^https?:\/\//i.test(decoded)) return decoded;
+            }
+            catch { }
+          }
+        }
+      }
+      return url.toString();
+    }
+    catch {
+      return href;
+    }
+  }
+
+  async search(q, limit, lang) {
+    const cacheKey = createCacheKey("bing", q, limit, lang);
+    const cached = searchCache.get(cacheKey);
+    if (cached)
+      return cached;
+    const market = getMarketFromLang(lang);
+    const results = await browserPool.withBrowser(async (browser) => {
+      const page = await browser.newPage();
+      try {
+        await page.setViewport({ width: 1365, height: 768 });
+        await page.setUserAgent(DEFAULT_BROWSER_SEARCH_USER_AGENT);
+        await page.setExtraHTTPHeaders(getAcceptLanguageHeader(lang));
+        const url = new URL("https://www.bing.com/search");
+        url.searchParams.set("q", q);
+        url.searchParams.set("mkt", market);
+        const response = await page.goto(url.toString(), {
+          waitUntil: "domcontentloaded",
+          timeout: PUPPETEER_TIMEOUT
+        });
+        if (response && response.status() >= 400) {
+          throw new Error(`Bing HTTP ${response.status()}`);
+        }
+        await page.waitForSelector("li.b_algo h2 a[href], li.b_algo a[href]", { timeout: 10000 }).catch(() => undefined);
+        const items = await page.evaluate(maxResults => {
+          const parsed = [];
+          for (const card of Array.from(document.querySelectorAll("li.b_algo"))) {
+            const anchor = card.querySelector("h2 a[href]") || card.querySelector("a[href]");
+            const title = anchor?.textContent?.trim() || "";
+            const href = anchor?.getAttribute("href") || "";
+            if (!title || !href)
+              continue;
+            const snippetElement = card.querySelector("div.b_caption p, div.b_snippet, p");
+            const snippet = snippetElement?.textContent?.trim() || undefined;
+            parsed.push({ title, url: href, snippet });
+            if (parsed.length >= maxResults)
+              break;
+          }
+          return parsed;
+        }, limit);
+        return items.flatMap(result => {
+          try {
+            const absolute = new URL(result.url, "https://www.bing.com/").toString();
+            const decoded = this.decodeBingRedirect(absolute);
+            new URL(decoded);
+            return [{ ...result, url: decoded, source: "bing" }];
+          }
+          catch {
+            return [];
+          }
+        });
+      }
+      finally {
+        await page.close();
+      }
+    });
+    searchCache.set(cacheKey, results);
+    return results;
+  }
+
+  async isAvailable() {
+    try {
+      await browserPool.getBrowser();
+      return true;
+    }
+    catch {
+      return false;
+    }
+  }
+}

docker/web-search/patch-mcp-web-search.mjs

@@ -0,0 +1,28 @@
+import fs from "node:fs";
+
+// Context Kit patch for @zhafron/mcp-web-search 1.3.0.
+// Upstream hard-codes the fetch_url schema limit to 25 MiB even though the
+// runtime extractor already uses MAX_BYTES. Keep this narrow and fail the build
+// if upstream changes the compiled source shape.
+const serverPath = "/usr/local/lib/node_modules/@zhafron/mcp-web-search/dist/src/server.js";
+let source = fs.readFileSync(serverPath, "utf8");
+
+const replacements = [
+  [
+    'import { MAX_RESULTS } from "./constants.js";',
+    'import { MAX_BYTES, MAX_RESULTS } from "./constants.js";'
+  ],
+  [
+    "max_download_bytes: z.number().int().min(1).max(26214400).optional()",
+    "max_download_bytes: z.number().int().min(1).max(MAX_BYTES).optional()"
+  ]
+];
+
+for (const [before, after] of replacements) {
+  if (!source.includes(before)) {
+    throw new Error(`mcp-web-search patch target not found: ${before}`);
+  }
+  source = source.replace(before, after);
+}
+
+fs.writeFileSync(serverPath, source);

docker/web-search/searxng/settings.yml

@@ -15,8 +15,8 @@ search:
     - json
 
 server:
-  # Local placeholder. The Docker service also sets SEARXNG_SECRET from .env;
-  # keep SearXNG bound to 127.0.0.1 unless you review this config separately.
+  # Local placeholder. Keep SearXNG bound to 127.0.0.1 unless you review this
+  # config and replace the secret_key for a deliberate non-local deployment.
   secret_key: "local-only-change-if-exposed"
   limiter: false
   image_proxy: true