tesseract-agent 0.9.15 → 0.9.20

package/dist-bundle/cli.js

@@ -22277,6 +22277,54 @@ ${r?.url ?? ""}`);
     tool3("browser_type", `Type text into an input/textarea by CSS selector (fires input + change events). ${REAL}`, { selector: external_exports.string().describe("CSS selector of the field."), text: external_exports.string().describe("Text to set as the field value.") }, async (a) => {
       const r = await drive("type", { selector: a.selector, text: a.text });
       return text2(r?.typed ? `Typed into ${a.selector}` : `No element matched ${a.selector}`, !r?.typed);
+    }),
+    // ── FULL CONTROL / runtime-debugger tools (need the Companion extension; the built-in engine
+    //    supports the structured verbs above). Powerful — gated by the user's opt-in + indicator. ──
+    tool3("browser_eval", `Run ARBITRARY JavaScript in the page and return its result. Executes in the real page context via the DevTools protocol (bypasses page CSP for the top-level expression) \u2014 use for anything the structured tools don't cover: read variables & site functions, walk the DOM, compute values, mutate state, await async work. The last expression value (or an awaited promise) is returned, serialized by value. ${REAL}`, { expression: external_exports.string().describe("JS expression or statements to evaluate in the page."), awaitPromise: external_exports.boolean().optional().describe("Await a returned promise (default true).") }, async (a) => {
+      const r = await drive("eval", { expression: a.expression, awaitPromise: a.awaitPromise });
+      if (r?.exceptionDetails)
+        return text2(`Exception: ${JSON.stringify(r.exceptionDetails)}`, true);
+      return text2(typeof r?.value === "string" ? r.value : JSON.stringify(r?.value ?? null, null, 2));
+    }),
+    tool3("browser_cdp", `Send a RAW Chrome DevTools Protocol command to the driven tab and return the result \u2014 full runtime-debugger access to ANY CDP domain: DOM, CSS, Runtime, Debugger (breakpoints / pause / step / evaluateOnCallFrame), Network (enable, then getResponseBody), Page, Input, Emulation, Storage, Performance, Accessibility, \u2026 Enable a domain (e.g. Network.enable) then read its async notifications with browser_events. ${REAL}`, { method: external_exports.string().describe('CDP method, e.g. "DOM.getDocument", "Debugger.setBreakpointByUrl", "Network.enable", "Emulation.setDeviceMetricsOverride".'), params: external_exports.record(external_exports.string(), external_exports.unknown()).optional().describe("Params object for the CDP method.") }, async (a) => {
+      const r = await drive("cdp", { method: a.method, params: a.params ?? {} });
+      return text2(JSON.stringify(r ?? null, null, 2));
+    }),
+    tool3("browser_events", `Drain buffered CDP notifications from the driven tab \u2014 console messages (Runtime.consoleAPICalled), uncaught errors, Network.* traffic, Debugger.paused, Log.entryAdded, etc. Enable the relevant domain first with browser_cdp (Runtime.enable / Network.enable / Log.enable / Debugger.enable), then poll this. ${REAL}`, { filter: external_exports.string().optional().describe('Only return events whose CDP method starts with this prefix, e.g. "Network." or "Runtime.consoleAPICalled".'), keep: external_exports.boolean().optional().describe("Keep events in the buffer instead of draining (default false = drain).") }, async (a) => {
+      const r = await drive("pullEvents", { method: a.filter, text: a.keep ? "keep" : "" });
+      return text2(JSON.stringify(r?.events ?? [], null, 2));
+    }),
+    tool3("browser_wait_for", `Wait until a CSS selector appears OR a JS expression becomes truthy (polls up to timeout ms). ${REAL}`, { selector: external_exports.string().optional().describe("CSS selector to wait for."), expression: external_exports.string().optional().describe("JS expression to poll until truthy (used when no selector is given)."), timeout: external_exports.number().optional().describe("Max wait in ms (default 10000).") }, async (a) => {
+      const r = await drive("waitFor", { selector: a.selector, expression: a.expression, timeout: a.timeout });
+      return text2(r?.ok ? `Ready after ${r.waitedMs}ms` : "Timed out waiting", !r?.ok);
+    }),
+    tool3("browser_get_html", `Return the full outer HTML of the current page (document.documentElement.outerHTML). ${REAL}`, {}, async () => {
+      const r = await drive("html");
+      return text2(r?.html ?? "");
+    }),
+    tool3("browser_real_click", `Click an element with a REAL, trusted mouse event dispatched at its center via the DevTools protocol (isTrusted=true) \u2014 use when a site ignores the synthetic browser_click (OAuth buttons, some SPAs, drag handles). ${REAL}`, { selector: external_exports.string().describe("CSS selector of the element to click.") }, async (a) => {
+      const r = await drive("realClick", { selector: a.selector });
+      return text2(r?.clicked ? `Trusted-clicked ${a.selector}` : `No element matched ${a.selector}`, !r?.clicked);
+    }),
+    tool3("browser_real_type", `Type text as REAL, trusted keyboard input via the DevTools protocol (optionally focusing a selector first) \u2014 use when a site ignores the synthetic browser_type. ${REAL}`, { selector: external_exports.string().optional().describe("CSS selector to focus first (optional; types into the already-focused element otherwise)."), text: external_exports.string().describe("Text to type.") }, async (a) => {
+      await drive("realType", { selector: a.selector, text: a.text });
+      return text2(`Typed${a.selector ? ` into ${a.selector}` : ""}`);
+    }),
+    tool3("browser_key", `Press a single trusted key in the driven tab \u2014 "Enter" / "Tab" / "Escape" / "Backspace" or a character. ${REAL}`, { key: external_exports.string().describe('Key to press, e.g. "Enter".') }, async (a) => {
+      await drive("key", { text: a.key });
+      return text2(`Pressed ${a.key}`);
+    }),
+    tool3("browser_cookies", `Read cookies for the driven tab (includes httpOnly) via the DevTools protocol. ${REAL}`, { url: external_exports.string().optional().describe("Restrict to cookies for this URL (optional).") }, async (a) => {
+      const r = await drive("cookies", { url: a.url });
+      return text2(JSON.stringify(r ?? {}, null, 2));
+    }),
+    tool3("browser_screenshot_full", `Capture a full-page PNG (beyond the visible viewport) via the DevTools protocol. ${REAL}`, {}, async () => {
+      const r = await drive("screenshotFull");
+      const dataUrl = r?.dataUrl ?? "";
+      const b642 = dataUrl.startsWith("data:") ? dataUrl.slice(dataUrl.indexOf(",") + 1) : dataUrl;
+      if (!b642)
+        return text2("No screenshot data returned.", true);
+      return { content: [{ type: "image", data: b642, mimeType: "image/png" }] };
     })
   ];
 }
@@ -73864,11 +73912,35 @@ async function replayTurn(turn) {
 var extensionBrowserConn = null;
 function setExtensionBrowserConn(send3) {
   extensionBrowserConn = send3;
+  if (send3) {
+    browserPushActive = false;
+    browserEventBuf.length = 0;
+  }
 }
 function clearExtensionBrowserConnIf(send3) {
   if (extensionBrowserConn === send3)
     extensionBrowserConn = null;
 }
+var browserEventBuf = [];
+var browserPushActive = false;
+var MAX_BROWSER_EVENTS = 1e3;
+function pushBrowserEvent(e) {
+  browserPushActive = true;
+  browserEventBuf.push(e ?? {});
+  if (browserEventBuf.length > MAX_BROWSER_EVENTS)
+    browserEventBuf.splice(0, browserEventBuf.length - MAX_BROWSER_EVENTS);
+}
+function drainBrowserEvents(params) {
+  const prefix = typeof params.method === "string" ? params.method : "";
+  const match = (e) => !prefix || (e.method ?? "").startsWith(prefix);
+  const events = browserEventBuf.filter(match);
+  if (params.text !== "keep") {
+    for (let i = browserEventBuf.length - 1; i >= 0; i--)
+      if (match(browserEventBuf[i]))
+        browserEventBuf.splice(i, 1);
+  }
+  return events;
+}
 setBrowserExtensionProbe(() => extensionBrowserConn != null);
 function registerAgentSink(sink) {
   sinks.add(sink);
@@ -73957,6 +74029,10 @@ function buildSharedBackend(engineId) {
     },
     (requestId, action, params, browserId) => {
       if (extensionBrowserConn) {
+        if (action === "pullEvents" && browserPushActive && backend) {
+          backend.sendBrowserResponse?.(requestId, { events: drainBrowserEvents(params ?? {}) });
+          return;
+        }
         extensionBrowserConn("browser", { id: requestId, op: action, ...params ?? {} });
         return;
       }
@@ -83651,6 +83727,8 @@ async function startLocalServer(opts) {
           setExtensionBrowserConn(send3);
         } else if (p?.id)
           peekSharedBackend()?.sendBrowserResponse?.(p.id, p.result, p.error);
+      } else if (env.channel === "browser:event") {
+        pushBrowserEvent(env.payload);
       } else if (env.channel === "rdesk") {
         const rd = opts.getRdesk?.();
         if (rd && env.sessionId) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tesseract-agent",
-  "version": "0.9.15",
+  "version": "0.9.20",
   "license": "UNLICENSED",
   "publishConfig": {
     "access": "public",