ax-grep 0.1.1 → 0.1.3

package/README.md

@@ -16,24 +16,19 @@
 
 <img src="./docs/assets/ax-grep-search.png" alt="ax-grep auto search comparison image" width="920">
 
-Compact semantic trees and agent-ready page checks from HTML, URLs, WebViews, and live browser pages.
-It approximates the accessibility tree browsers expose, but can do it from
-static HTML before launching browser automation.
-
-The CLI defaults to `impit` for browser-like HTTP/2/TLS static fetches, detects
-hCaptcha, reCAPTCHA, and Cloudflare challenge markers, then tells agents when a
-browser is actually needed. In the local benchmark, the decision handoff used
-15.4x less peak RAM on average and 3.0x fewer decision tokens on the content
-fixture than `agent-browser snapshot`.
-
-`--search` auto mode tries DuckDuckGo, Bing, and StartPage, records every
-engine attempt, and hands ranked source choices back to the agent.
+Compact semantic trees and agent-ready page checks from HTML, URLs, WebViews,
+and live browser pages. It approximates browser accessibility trees before
+automation, defaults to `impit` for browser-like HTTP/2/TLS fetches, detects
+major CAPTCHA and bot-challenge markers, saves 15.4x less peak RAM and 3.0x
+fewer decision tokens in the local benchmark, and `--search` auto mode tries
+DuckDuckGo, Bing, StartPage, and Google while recording every engine attempt.
 
 </div>
 
 ## 1. Try With A Prompt
 
-Paste this into a Codex session or subagent prompt before opening a browser:
+Paste this into a Codex session or subagent prompt before opening a browser
+(also works as guidance for Claude, Gemini, OpenRouter, and other agents):
 
 ```md
 Before opening a browser, inspect pages with ax-grep when possible.
@@ -52,7 +47,9 @@ not enough.
 curl -fsSL https://raw.githubusercontent.com/hmmhmmhm/ax-grep/main/skills.sh | sh
 ```
 
-This installs the Codex skill prompt only. Restart Codex if the new skill is not listed immediately.
+This installs the Codex skill prompt only. The same prompt pattern can be pasted
+into Claude, Gemini, OpenRouter, or other agent runners. Restart Codex if the
+new skill is not listed immediately.
 
 ## 3. Try The CLI
 
@@ -107,3 +104,5 @@ const text = await page.evaluate(script);
 - [Agent handoff loop](./docs/agent-handoff.md)
 - [Benchmarks](./docs/benchmarks.md)
 - [Library API](./docs/library-api.md)
+- [Server agent integration](./docs/server-agent.md)
+- [WebView and in-page usage](./docs/webview.md)

package/dist/browser.d.ts

@@ -1,4 +1,4 @@
-import { a as SemanticTreeOptions, S as SemanticNode, aE as SemanticTreeChange, aF as SemanticTreeObserverOptions } from './types-gwHWhYmw.js';
+import { a as SemanticTreeOptions, S as SemanticNode, aE as SemanticTreeChange, aF as SemanticTreeObserverOptions } from './types-K1hqb7Pq.js';
 
 declare function extractSemanticTree(options?: SemanticTreeOptions): SemanticNode;
 

package/dist/cli.js

@@ -135,7 +135,7 @@ var hiddenAgentPageCheckPaths = [
 var defaultTimeoutMs = 15e3;
 var defaultUserAgent = "ax-grep/0.1 (+https://github.com/hmmhmmhm/ax-grep)";
 var defaultFetcher = "impit";
-var autoSearchEngines = ["duckduckgo", "bing", "startpage"];
+var autoSearchEngines = ["duckduckgo", "bing", "startpage", "google"];
 function envelopeAgentCanReadCurrentPayload(envelope) {
   const agent = envelope.agent;
   return agent?.needsBrowserHtml === false && agent.primaryAction?.action === "read-content" && actionExecution(agent.primaryAction) === "read-current";
@@ -944,8 +944,8 @@ function parseMode(value) {
   throw new UsageError(`--mode must be compact, interactive, or full`);
 }
 function parseSearchEngine(value) {
-  if (value === "auto" || value === "bing" || value === "duckduckgo" || value === "startpage") return value;
-  throw new UsageError(`--engine must be auto, bing, duckduckgo, or startpage`);
+  if (value === "auto" || value === "bing" || value === "duckduckgo" || value === "startpage" || value === "google") return value;
+  throw new UsageError(`--engine must be auto, bing, duckduckgo, startpage, or google`);
 }
 function parseSearchLang(value) {
   const normalized = value.trim().toLowerCase();
@@ -973,6 +973,13 @@ function searchUrl(query, engine, lang, region) {
     if (region) url2.searchParams.set("region", region);
     return url2.toString();
   }
+  if (engine === "google") {
+    const url2 = new URL("https://www.google.com/search");
+    url2.searchParams.set("q", query);
+    if (lang) url2.searchParams.set("hl", lang);
+    if (region) url2.searchParams.set("gl", region);
+    return url2.toString();
+  }
   const url = new URL("https://duckduckgo.com/html/");
   url.searchParams.set("q", query);
   if (lang && region) url.searchParams.set("kl", `${region.toLowerCase()}-${lang}`);
@@ -1010,7 +1017,7 @@ Fetch a page and print a compact semantic accessibility-like tree.
 
 Options:
   --search <query>           Search the web and analyze the result page.
-  --engine <name>            Search engine for --search: auto, duckduckgo, bing, or startpage. Default: auto.
+  --engine <name>            Search engine for --search: auto, duckduckgo, bing, startpage, or google. Default: auto.
   --lang <code>              Search language hint, e.g. en, ko, ja, zh-cn.
   --region <code>            Search region hint, e.g. US, KR, JP, CN.
   --open-result <n|best>     With --search, fetch and analyze the selected result.
@@ -4511,6 +4518,7 @@ function detectSearchEngine(url) {
     if (hostname.endsWith("bing.com")) return "bing";
     if (hostname.endsWith("duckduckgo.com")) return "duckduckgo";
     if (hostname.endsWith("startpage.com")) return "startpage";
+    if (hostname.endsWith("google.com")) return "google";
     if (hostname.endsWith("search.yahoo.co.jp")) return "yahoo-japan";
     if (looksLikeGenericSearchUrl(url)) return "generic";
     return null;
@@ -4539,6 +4547,9 @@ function isResultCard(element, engine) {
   if (engine === "duckduckgo") {
     return hasClass(element, "result") || hasClass(element, "web-result") || hasClass(element, "result__body");
   }
+  if (engine === "google") {
+    return hasClass(element, "g") || hasClass(element, "MjjYud") || element.name === "div" && findElement(element.children, (child) => /^h[1-6]$/.test(child.name)) !== void 0;
+  }
   if (engine === "yahoo-japan") return hasClass(element, "sw-Card") || hasClass(element, "algo") || hasClass(element, "SearchResult");
   if (engine === "generic") {
     return hasClass(element, "result") || hasClass(element, "search-result") || element.name === "li" && findElement(element.children, (child) => child.name === "a") !== void 0 && findElement(element.children, (child) => child.name === "p") !== void 0;
@@ -4566,7 +4577,7 @@ function resultFromCard(card, baseUrl, engine, rank) {
   return withResultDateHint(result);
 }
 function resultTitleLink(card, engine) {
-  if (engine === "baidu" || engine === "bing" || engine === "yahoo-japan") {
+  if (engine === "baidu" || engine === "bing" || engine === "google" || engine === "yahoo-japan") {
     const heading2 = findElement(card.children, (element) => /^h[1-6]$/.test(element.name));
     const headingLink = heading2 ? firstUsefulAnchor(heading2, "https://example.invalid") : void 0;
     if (headingLink) return headingLink;
@@ -13661,6 +13672,10 @@ function summarizeBrowserHtmlReasonCode(needsBrowserHtml, analysis, primaryActio
   if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "HCAPTCHA_CHALLENGE")) return "hcaptcha";
   if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "RECAPTCHA_CHALLENGE")) return "recaptcha";
   if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "CLOUDFLARE_CHALLENGE")) return "cloudflare-challenge";
+  if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "AKAMAI_CHALLENGE")) return "akamai-challenge";
+  if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "DATADOME_CHALLENGE")) return "datadome-challenge";
+  if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "PERIMETERX_CHALLENGE")) return "perimeterx-challenge";
+  if (analysis.diagnostics.some((diagnostic) => diagnostic.code === "KASADA_CHALLENGE")) return "kasada-challenge";
   if (error?.code === "NO_INSPECTABLE_CONTENT") return "no-inspectable-content";
   if (error?.code === "HTTP_ERROR") return "http-error";
   if (error?.code === "FETCH_FAILED" || error?.code === "TIMEOUT") return "fetch-error";
@@ -16330,6 +16345,34 @@ function detectBarrierDiagnostics(fetched, tree, content) {
       message: "The page appears to contain a Cloudflare challenge."
     });
   }
+  if (/(akamai|akamai bot manager|bm-verify|_abck|ak_bmsc|sensor_data|abck)/i.test(challengeHaystack)) {
+    diagnostics.push({
+      severity: "warning",
+      code: "AKAMAI_CHALLENGE",
+      message: "The page appears to contain an Akamai bot challenge."
+    });
+  }
+  if (/(datadome|captcha-delivery\.com|geo\.captcha-delivery\.com|ddcid|dd_r)/i.test(challengeHaystack)) {
+    diagnostics.push({
+      severity: "warning",
+      code: "DATADOME_CHALLENGE",
+      message: "The page appears to contain a DataDome challenge."
+    });
+  }
+  if (/(perimeterx|px-captcha|_px|pxvid|px3)/i.test(challengeHaystack)) {
+    diagnostics.push({
+      severity: "warning",
+      code: "PERIMETERX_CHALLENGE",
+      message: "The page appears to contain a PerimeterX challenge."
+    });
+  }
+  if (/(kasada|x-kpsdk|kpsdk|ips\.js)/i.test(challengeHaystack)) {
+    diagnostics.push({
+      severity: "warning",
+      code: "KASADA_CHALLENGE",
+      message: "The page appears to contain a Kasada challenge."
+    });
+  }
   if (/(captcha|verify you are human|unusual traffic|checking your browser|just a moment|cf-browser-verification|cloudflare|access denied|request blocked|please wait for verification|please wait|enable javascript|enable java script|자바스크립트|봇이 아닙니다|자동입력|보안문자)/i.test(challengeHaystack)) {
     diagnostics.push({
       severity: "warning",
@@ -16369,7 +16412,7 @@ function dedupeDiagnostics(diagnostics) {
   });
 }
 function isChallengeDiagnosticCode(code) {
-  return code === "CHALLENGE_LIKELY" || code === "HCAPTCHA_CHALLENGE" || code === "RECAPTCHA_CHALLENGE" || code === "CLOUDFLARE_CHALLENGE";
+  return code === "CHALLENGE_LIKELY" || code === "HCAPTCHA_CHALLENGE" || code === "RECAPTCHA_CHALLENGE" || code === "CLOUDFLARE_CHALLENGE" || code === "AKAMAI_CHALLENGE" || code === "DATADOME_CHALLENGE" || code === "PERIMETERX_CHALLENGE" || code === "KASADA_CHALLENGE";
 }
 function looksLikeKnownSearchUrl(url) {
   try {
@@ -21972,7 +22015,7 @@ function inferSearchResultCommandContext(url) {
     }
     if (hostname.endsWith("google.com")) {
       const query = parsed.searchParams.get("q");
-      return query ? { query } : void 0;
+      return query ? { query, engine: "google" } : void 0;
     }
     if (hostname.endsWith("baidu.com")) {
       const query = parsed.searchParams.get("wd") ?? parsed.searchParams.get("word");
@@ -22296,7 +22339,7 @@ function parseArgMetadata(argv) {
     }
     if (arg === "--engine") {
       const value = argv[index + 1];
-      if (value === "auto" || value === "bing" || value === "duckduckgo" || value === "startpage") metadata.searchEngine = value;
+      if (value === "auto" || value === "bing" || value === "duckduckgo" || value === "startpage" || value === "google") metadata.searchEngine = value;
       index += 1;
       continue;
     }