omnius 1.0.206 → 1.0.208

package/dist/index.js

@@ -10326,6 +10326,108 @@ function pngDimensions(buffer2) {
   }
   return null;
 }
+async function describeFocusedEditableInContext(context2, frameMeta) {
+  const active = await context2.evaluate(`(() => {
+    const el = document.activeElement;
+    if (!el) return null;
+    const rect = el.getBoundingClientRect();
+    const role = (el.getAttribute("role") || "").toLowerCase();
+    const contentEditableAttr = el.getAttribute("contenteditable");
+    const contentEditable = contentEditableAttr !== null
+      && (contentEditableAttr === "" || String(contentEditableAttr).toLowerCase() === "true");
+    const disabled = !!el.disabled;
+    const readOnly = !!el.readOnly;
+    const isEditable = (el.matches("input, textarea") && !disabled && !readOnly)
+      || contentEditable || el.isContentEditable === true
+      || ["textbox", "searchbox", "combobox"].includes(role);
+    return {
+      tag: String(el.tagName || "").toLowerCase(),
+      id: el.id || "",
+      name: el.getAttribute("name") || "",
+      role,
+      ariaLabel: el.getAttribute("aria-label") || "",
+      type: el.getAttribute("type") || "",
+      placeholder: el.getAttribute("placeholder") || "",
+      text: String(el.textContent || "").trim().slice(0, 120),
+      isEditable,
+      disabled,
+      readOnly,
+      rect: { x: rect.x, y: rect.y, width: rect.width, height: rect.height },
+    };
+  })()`);
+  if (!active || typeof active !== "object")
+    return null;
+  return frameMeta ? { ...active, frame: frameMeta } : active;
+}
+async function describeFocusedEditable(pageHandle) {
+  const main2 = await describeFocusedEditableInContext(pageHandle, { kind: "main", url: pageHandle.url?.() ?? "" }).catch(() => null);
+  if (main2?.["isEditable"])
+    return main2;
+  const frames = typeof pageHandle.frames === "function" ? pageHandle.frames() : [];
+  const mainFrame = typeof pageHandle.mainFrame === "function" ? pageHandle.mainFrame() : null;
+  for (let i2 = 0; i2 < frames.length; i2++) {
+    const frame = frames[i2];
+    if (!frame || frame === mainFrame)
+      continue;
+    const active = await describeFocusedEditableInContext(frame, {
+      kind: "frame",
+      index: i2,
+      url: typeof frame.url === "function" ? frame.url() : "",
+      name: typeof frame.name === "function" ? frame.name() : ""
+    }).catch(() => null);
+    if (active?.["isEditable"])
+      return active;
+  }
+  return main2;
+}
+async function clickAndFillBrowserTarget(pageHandle, target, text, typingDelay) {
+  const viewport = pageHandle.viewportSize?.() ?? { width: 1280, height: 720 };
+  let candidate = await findBrowserVisualCandidate(pageHandle, target, viewport.width / 2, viewport.height / 2, true);
+  let source = "dom-candidate";
+  if (!candidate) {
+    candidate = await findBrowserVisualCandidate(pageHandle, target, viewport.width / 2, viewport.height / 2, true, true, true);
+    if (candidate?.["scrolledIntoView"] === true)
+      source += "+scroll";
+    if (candidate)
+      await pageHandle.waitForTimeout(150);
+  }
+  const center = candidate?.["center"];
+  const x = Number(center?.x);
+  const y = Number(center?.y);
+  if (!Number.isFinite(x) || !Number.isFinite(y)) {
+    throw new Error(`No visible editable candidate matched target "${target}". Run observe_bundle or dom_summary to inspect available labels/selectors.`);
+  }
+  await pageHandle.mouse.click(x, y);
+  await pageHandle.waitForTimeout(80);
+  const active = await describeFocusedEditable(pageHandle);
+  if (!active?.["isEditable"]) {
+    throw new Error(`Target "${target}" was clicked, but no editable element became focused. Matched element: ${candidate ? JSON.stringify({
+      tag: candidate["tag"],
+      text: candidate["text"],
+      ariaLabel: candidate["ariaLabel"],
+      placeholder: candidate["placeholder"],
+      name: candidate["name"]
+    }) : "(none)"}.`);
+  }
+  const selectAll = process.platform === "darwin" ? "Meta+A" : "Control+A";
+  await pageHandle.keyboard.press(selectAll);
+  await pageHandle.keyboard.type(text, { delay: typingDelay });
+  return { candidate, active, source };
+}
+function evaluateFailureMessage(err, code8) {
+  const raw = err instanceof Error ? err.message : String(err);
+  const hints = [];
+  if (/map is not a function/i.test(raw) && /querySelectorAll/i.test(code8)) {
+    hints.push("document.querySelectorAll() returns a NodeList; use Array.from(document.querySelectorAll(selector)).map(...) or [...document.querySelectorAll(selector)].map(...).");
+  }
+  if (/(?:\.value\s*=|setAttribute\(['"]value['"])/.test(code8) && /\b(input|textarea|querySelector)/i.test(code8)) {
+    hints.push("Do not fill modern React/Vue/Svelte forms by assigning .value in evaluate; use playwright_browser fill, or visual_click the field then type, so input/change events fire.");
+  }
+  if (/querySelectorAll|querySelector/.test(code8)) {
+    hints.push("For page inspection, prefer query_all, dom_summary, or observe_bundle before raw evaluate.");
+  }
+  return [raw.slice(0, 500), ...hints.map((hint) => `Hint: ${hint}`)].join("\n");
+}
 function buildImageMarker(buffer2) {
   let mimeType = "image/png";
   let out = buffer2;
@@ -10391,8 +10493,8 @@ ${input.text.slice(0, 2e4)}`.toLowerCase();
   }
   return { kind: "none", confidence: 0, evidence: [] };
 }
-async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY, forceCandidate, includeOffscreen = false, scrollIntoView = false) {
-  const candidate = await pageHandle.evaluate(`(() => {
+async function findBrowserVisualCandidateInContext(context2, target, visualX, visualY, forceCandidate, includeOffscreen = false, scrollIntoView = false) {
+  const candidate = await context2.evaluate(`(() => {
     const target = ${JSON.stringify(target)};
     const visualX = ${JSON.stringify(visualX)};
     const visualY = ${JSON.stringify(visualY)};
@@ -10423,12 +10525,55 @@ async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY,
       if (window.CSS && typeof window.CSS.escape === "function") return window.CSS.escape(id);
       return String(id).replace(/["\\\\]/g, "\\\\$&");
     };
+    const hasVisibleStyle = (el) => {
+      for (let cur = el; cur && cur.nodeType === 1; cur = cur.parentElement) {
+        const style = getComputedStyle(cur);
+        if (style.display === "none" || style.visibility === "hidden" || style.visibility === "collapse") return false;
+        const opacity = Number(style.opacity);
+        if (Number.isFinite(opacity) && opacity <= 0.02) return false;
+      }
+      return true;
+    };
+    const hasUsableBox = (el) => {
+      const rect = el.getBoundingClientRect();
+      return rect.width > 1 && rect.height > 1;
+    };
+    const inViewport = (rect) => !(rect.bottom < 0 || rect.right < 0 || rect.top > window.innerHeight || rect.left > window.innerWidth);
+    const isRendered = (el) => !!el && hasUsableBox(el) && hasVisibleStyle(el);
     const associatedControl = (el) => {
       if (/^label$/i.test(el.tagName || "") && el.getAttribute("for")) {
         return document.getElementById(el.getAttribute("for"));
       }
+      if (/^label$/i.test(el.tagName || "")) {
+        if (el.control) return el.control;
+        const nested = el.querySelector("input, textarea, select, [contenteditable='true'], [role='textbox']");
+        if (nested) return nested;
+        const labelRect = el.getBoundingClientRect();
+        let scope = el.parentElement;
+        for (let depth = 0; depth < 5 && scope; depth++, scope = scope.parentElement) {
+          const controls = Array.from(scope.querySelectorAll("input, textarea, select, [contenteditable='true'], [role='textbox']"))
+            .filter(control => {
+              const rect = control.getBoundingClientRect();
+              return rect.width > 1 && rect.height > 1 && rect.top >= labelRect.top - 12 && Math.abs(rect.left - labelRect.left) < 260;
+            })
+            .sort((a, b) => {
+              const ar = a.getBoundingClientRect();
+              const br = b.getBoundingClientRect();
+              return (Math.abs(ar.top - labelRect.bottom) - Math.abs(br.top - labelRect.bottom))
+                || (Math.abs(ar.left - labelRect.left) - Math.abs(br.left - labelRect.left));
+            });
+          if (controls[0]) return controls[0];
+        }
+      }
       return el;
     };
+    const clickTargetFor = (el) => {
+      const control = associatedControl(el) || el;
+      if (control === el) return el;
+      if (isRendered(control)) return control;
+      if (/^label$/i.test(el.tagName || "") && isRendered(el)) return el;
+      return control;
+    };
     const associatedLabelText = (el) => {
       const control = associatedControl(el);
       const id = control && control.id ? control.id : el.id;
@@ -10471,20 +10616,25 @@ async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY,
       el.id || "",
       el.className || "",
       el.getAttribute("type") || "",
+      el.getAttribute("autocomplete") || "",
     ].join(" ").toLowerCase();
     const infoFor = (el, score) => {
-      const rect = el.getBoundingClientRect();
-      const visible = !(rect.bottom < 0 || rect.right < 0 || rect.top > window.innerHeight || rect.left > window.innerWidth);
+      const control = associatedControl(el) || el;
+      const clickTarget = clickTargetFor(el);
+      const rect = clickTarget.getBoundingClientRect();
+      const visible = inViewport(rect) && hasVisibleStyle(clickTarget);
       return {
-        tag: String(el.tagName || "").toLowerCase(),
-        id: el.id || "",
-        className: String(el.className || "").slice(0, 160),
-        role: el.getAttribute("role") || "",
-        ariaLabel: el.getAttribute("aria-label") || "",
-        name: el.getAttribute("name") || "",
-        type: el.getAttribute("type") || "",
-        placeholder: el.getAttribute("placeholder") || "",
-        text: String(el.innerText || el.textContent || el.getAttribute("value") || "").trim().slice(0, 240),
+        tag: String(control.tagName || el.tagName || "").toLowerCase(),
+        id: control.id || el.id || "",
+        className: String(control.className || el.className || "").slice(0, 160),
+        role: control.getAttribute("role") || el.getAttribute("role") || "",
+        ariaLabel: control.getAttribute("aria-label") || el.getAttribute("aria-label") || "",
+        name: control.getAttribute("name") || el.getAttribute("name") || "",
+        type: control.getAttribute("type") || el.getAttribute("type") || "",
+        autocomplete: control.getAttribute("autocomplete") || el.getAttribute("autocomplete") || "",
+        placeholder: control.getAttribute("placeholder") || el.getAttribute("placeholder") || "",
+        text: String(el.innerText || el.textContent || control.innerText || control.textContent || control.getAttribute("value") || "").trim().slice(0, 240),
+        clickTag: String(clickTarget.tagName || "").toLowerCase(),
         rect: { x: rect.x, y: rect.y, width: rect.width, height: rect.height },
         center: { x: rect.x + rect.width / 2, y: rect.y + rect.height / 2 },
         visible,
@@ -10506,9 +10656,11 @@ async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY,
     let best = null;
     let bestElement = null;
     for (const el of Array.from(document.querySelectorAll(selectors))) {
-      const rect = el.getBoundingClientRect();
-      if (rect.width <= 1 || rect.height <= 1) continue;
-      const visible = !(rect.bottom < 0 || rect.right < 0 || rect.top > window.innerHeight || rect.left > window.innerWidth);
+      if (!isRendered(el)) continue;
+      const clickTarget = clickTargetFor(el);
+      if (!isRendered(clickTarget)) continue;
+      const rect = clickTarget.getBoundingClientRect();
+      const visible = inViewport(rect);
       if (!includeOffscreen && !visible) continue;
       const hay = textFor(el);
       const tokenHits = tokens.filter(t => hay.includes(t)).length;
@@ -10519,7 +10671,7 @@ async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY,
       if (score <= (forceCandidate ? 8 : 0)) continue;
       if (!best || score > best.score) {
         best = infoFor(el, score);
-        bestElement = el;
+        bestElement = clickTarget;
       }
     }
     if (bestElement && scrollIntoView && best && !best.visible) {
@@ -10531,6 +10683,82 @@ async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY,
   })()`);
   return candidate && typeof candidate === "object" ? candidate : null;
 }
+function offsetBrowserCandidate(candidate, offset, viewport, frameMeta) {
+  const rect = candidate["rect"];
+  const center = candidate["center"];
+  const x = Number(rect?.x) + offset.x;
+  const y = Number(rect?.y) + offset.y;
+  const width = Number(rect?.width);
+  const height = Number(rect?.height);
+  const cx = Number(center?.x) + offset.x;
+  const cy = Number(center?.y) + offset.y;
+  const globalRect = {
+    x,
+    y,
+    width,
+    height
+  };
+  const visible = Number.isFinite(x) && Number.isFinite(y) && Number.isFinite(width) && Number.isFinite(height) && !(y + height < 0 || x + width < 0 || y > viewport.height || x > viewport.width);
+  return {
+    ...candidate,
+    rect: globalRect,
+    center: { x: cx, y: cy },
+    visible: candidate["visible"] === true && visible,
+    ...frameMeta ? {
+      frame: frameMeta,
+      frameLocalRect: rect,
+      frameLocalCenter: center
+    } : {}
+  };
+}
+async function findBrowserVisualCandidate(pageHandle, target, visualX, visualY, forceCandidate, includeOffscreen = false, scrollIntoView = false) {
+  const viewport = pageHandle.viewportSize?.() ?? { width: 1280, height: 720 };
+  const candidates = [];
+  const top = await findBrowserVisualCandidateInContext(pageHandle, target, visualX, visualY, forceCandidate, includeOffscreen, scrollIntoView).catch(() => null);
+  if (top)
+    candidates.push(offsetBrowserCandidate(top, { x: 0, y: 0 }, viewport));
+  const frames = typeof pageHandle.frames === "function" ? pageHandle.frames() : [];
+  const mainFrame = typeof pageHandle.mainFrame === "function" ? pageHandle.mainFrame() : null;
+  for (let i2 = 0; i2 < frames.length; i2++) {
+    const frame = frames[i2];
+    if (!frame || frame === mainFrame)
+      continue;
+    const elementHandle = typeof frame.frameElement === "function" ? await frame.frameElement().catch(() => null) : null;
+    if (!elementHandle)
+      continue;
+    let box = await elementHandle.boundingBox().catch(() => null);
+    if (!box || box.width <= 1 || box.height <= 1)
+      continue;
+    const frameVisible = !(box.y + box.height < 0 || box.x + box.width < 0 || box.y > viewport.height || box.x > viewport.width);
+    if (!includeOffscreen && !frameVisible)
+      continue;
+    const localX = Math.max(0, Math.min(box.width, visualX - box.x));
+    const localY = Math.max(0, Math.min(box.height, visualY - box.y));
+    let candidate = await findBrowserVisualCandidateInContext(frame, target, localX, localY, forceCandidate, includeOffscreen, scrollIntoView).catch(() => null);
+    if (!candidate)
+      continue;
+    if (scrollIntoView && (!candidate["visible"] || !frameVisible)) {
+      if (typeof elementHandle.scrollIntoViewIfNeeded === "function") {
+        await elementHandle.scrollIntoViewIfNeeded().catch(() => void 0);
+      }
+      box = await elementHandle.boundingBox().catch(() => box);
+      candidate = {
+        ...candidate,
+        scrolledIntoView: true
+      };
+    }
+    if (!box)
+      continue;
+    candidates.push(offsetBrowserCandidate(candidate, { x: box.x, y: box.y }, viewport, {
+      kind: "frame",
+      index: i2,
+      url: typeof frame.url === "function" ? frame.url() : "",
+      name: typeof frame.name === "function" ? frame.name() : "",
+      rect: { x: box.x, y: box.y, width: box.width, height: box.height }
+    }));
+  }
+  return candidates.filter((candidate) => includeOffscreen || candidate["visible"] === true).sort((a2, b) => Number(b["score"] ?? 0) - Number(a2["score"] ?? 0))[0] ?? null;
+}
 function ok(output, start2) {
   return { success: true, output, durationMs: Date.now() - start2 };
 }
@@ -10559,7 +10787,7 @@ var init_playwright_browser = __esm({
     PLAYWRIGHT_BROWSERS_DIR = join13(PLAYWRIGHT_RUNTIME_DIR, "browsers");
     PlaywrightBrowserTool = class {
       name = "playwright_browser";
-      description = "Full-scope Playwright browser automation + diagnostic capture. Launches a persistent headless Chromium session by default, with optional visible/headed mode when a GUI display is available. Beyond navigation/interaction, this tool buffers everything the running app emits (console messages, network requests, JS exceptions, accessibility tree) so the agent can verify what is ACTUALLY happening — not just what the build/test reports. Auto-installs Playwright + Chromium on first use without sudo or OS package manager escalation. Diagnostic actions: observe_bundle, dom_summary, dom, console_logs, network_log, page_errors, a11y_snapshot, bounding_box, query_all, performance, cookies, storage, viewport, clear_diagnostics. Interaction actions: navigate, click, visual_click, fill, type, press, select, check, hover. Capture actions: screenshot, pdf, content, innerText, innerHTML, getAttribute, evaluate. Loopback URLs (localhost, 127.0.0.1, ::1) are allowed for local development servers; private LAN and metadata URLs remain blocked. Workflow for user-facing work: start/serve the system with the stack-native tool, navigate to the real URL, then inspect page_errors, console_logs, network_log, DOM/accessibility, and screenshot evidence before completion. Build/typecheck/test output is only one layer; runtime browser evidence is required when the delivered artifact is a page, app, dashboard, game, form, visualization, or other UI. Repeat navigate/act/observe until the actual user flow is clean.";
+      description = "Full-scope Playwright browser automation + diagnostic capture. Launches a persistent headless Chromium session by default, with optional visible/headed mode when a GUI display is available. Beyond navigation/interaction, this tool buffers everything the running app emits (console messages, network requests, JS exceptions, accessibility tree) so the agent can verify what is ACTUALLY happening — not just what the build/test reports. Auto-installs Playwright + Chromium on first use without sudo or OS package manager escalation. Diagnostic actions: observe_bundle, dom_summary, dom, console_logs, network_log, page_errors, a11y_snapshot, bounding_box, query_all, performance, cookies, storage, viewport, clear_diagnostics. Interaction actions: navigate, click, visual_click, fill, type, press, select, check, hover. Use fill with a selector or natural-language target for form fields; avoid raw evaluate for form filling because direct .value assignment does not fire app input/change events. This is a separate browser/runtime from browser_action; once you start a workflow here, continue here unless you intentionally navigate browser_action to the same URL. Capture actions: screenshot, pdf, content, innerText, innerHTML, getAttribute, evaluate. Loopback URLs (localhost, 127.0.0.1, ::1) are allowed for local development servers; private LAN and metadata URLs remain blocked. Workflow for user-facing work: start/serve the system with the stack-native tool, navigate to the real URL, then inspect page_errors, console_logs, network_log, DOM/accessibility, and screenshot evidence before completion. Build/typecheck/test output is only one layer; runtime browser evidence is required when the delivered artifact is a page, app, dashboard, game, form, visualization, or other UI. Repeat navigate/act/observe until the actual user flow is clean.";
       parameters = {
         type: "object",
         properties: {
@@ -10609,7 +10837,7 @@ var init_playwright_browser = __esm({
               "clear_diagnostics",
               "close"
             ],
-            description: "Action to perform:\n- navigate: go to a URL\n- click: click element by selector\n- fill: clear input and type text (for form fields)\n- type: type text character by character into a selector, or into the currently focused element after visual_click\n- press: press a key (Enter, Tab, Escape, etc.)\n- screenshot: capture the headless browser page, not the desktop; use value to choose the output file path\n- observe_bundle: capture URL/title/viewport, DOM summary, a11y, diagnostics, screenshot, and gate assessment\n- visual_click: browser screenshot -> Moondream point -> elementFromPoint -> human-like Playwright mouse click -> post-action screenshot\n- evaluate: run JavaScript in page context\n- content: get page text content (readable, stripped)\n- dom: get raw page HTML (truncated)\n- dom_summary: compact interactive DOM summary with selectors\n- innerText: get innerText of a specific element\n- select: select dropdown option by value\n- check/uncheck: toggle checkbox\n- hover: hover over element\n- wait: wait for a selector to appear\n- waitForNavigation: wait for page navigation to complete\n- waitForSelector: wait for element matching selector\n- title: get page title\n- url: get current URL\n- getAttribute: get element attribute value\n- innerHTML: get element's innerHTML\n- textContent: get element's textContent\n- goBack/goForward/reload: browser navigation\n- pdf: save page as PDF\n- close: close browser session"
+            description: "Action to perform:\n- navigate: go to a URL\n- click: click element by selector\n- fill: clear input and type text by selector, or by natural-language target when selector is absent\n- type: type text character by character into a selector, or into the currently focused element after visual_click\n- press: press a key (Enter, Tab, Escape, etc.)\n- screenshot: capture the headless browser page, not the desktop; use value to choose the output file path\n- observe_bundle: capture URL/title/viewport, DOM summary, a11y, diagnostics, screenshot, and gate assessment\n- visual_click: browser screenshot -> Moondream point -> elementFromPoint -> human-like Playwright mouse click -> post-action screenshot\n- evaluate: run JavaScript in page context\n- content: get page text content (readable, stripped)\n- dom: get raw page HTML (truncated)\n- dom_summary: compact interactive DOM summary with selectors\n- innerText: get innerText of a specific element\n- select: select dropdown option by value\n- check/uncheck: toggle checkbox\n- hover: hover over element\n- wait: wait for a selector to appear, or sleep for timeout ms when no selector is provided\n- waitForNavigation: wait for page navigation to complete\n- waitForSelector: wait for element matching selector\n- title: get page title\n- url: get current URL\n- getAttribute: get element attribute value\n- innerHTML: get element's innerHTML\n- textContent: get element's textContent\n- goBack/goForward/reload: browser navigation\n- pdf: save page as PDF\n- close: close browser session"
           },
           url: {
             type: "string",
@@ -10625,7 +10853,7 @@ var init_playwright_browser = __esm({
           },
           target: {
             type: "string",
-            description: "Natural-language browser visual target for visual_click, for example 'the green Continue button' or 'the search field'."
+            description: "Natural-language browser visual target for visual_click or selector-less fill, for example 'the green Continue button', 'username field', or 'password field'."
           },
           value: {
             type: "string",
@@ -10745,12 +10973,22 @@ var init_playwright_browser = __esm({
               return ok(`Clicked: ${resolvedSelector}${resolvedSelector !== selector ? ` (from ${selector})` : ""}`, start2);
             }
             case "fill": {
-              if (!selector)
-                return fail("selector is required", start2);
               if (text === void 0)
                 return fail("text is required", start2);
-              await page.fill(selector, text, { timeout: timeout2 });
-              return ok(`Filled ${selector} with "${text}"`, start2);
+              const typingDelay = typeof args.typing_delay_ms === "number" ? Math.max(0, Math.min(500, Math.round(args.typing_delay_ms))) : 20;
+              if (selector) {
+                const resolvedSelector = resolveDomSummarySelector(selector);
+                if (!resolvedSelector)
+                  return fail(`No selector known for DOM summary reference ${selector}; run dom_summary and use the emitted selector.`, start2);
+                await page.fill(resolvedSelector, text, { timeout: timeout2 });
+                return ok(`Filled ${resolvedSelector}${resolvedSelector !== selector ? ` (from ${selector})` : ""} with "${text}"`, start2);
+              }
+              const target = typeof args.target === "string" && args.target.trim() ? args.target.trim() : "";
+              if (!target)
+                return fail("selector or target is required for fill. Prefer target for visual/natural-language form fields, e.g. target='username field'.", start2);
+              const result = await clickAndFillBrowserTarget(page, target, text, typingDelay);
+              const active = result.active ?? {};
+              return ok(`Filled target "${target}" via ${result.source} into <${active["tag"] || "element"}>${active["name"] ? ` name=${JSON.stringify(active["name"])}` : ""}${active["placeholder"] ? ` placeholder=${JSON.stringify(active["placeholder"])}` : ""}${active["ariaLabel"] ? ` aria-label=${JSON.stringify(active["ariaLabel"])}` : ""}.`, start2);
             }
             case "type": {
               if (text === void 0)
@@ -10760,30 +10998,14 @@ var init_playwright_browser = __esm({
                 await page.type(selector, text, { timeout: timeout2, delay: typingDelay });
                 return ok(`Typed "${text}" into ${selector}`, start2);
               }
-              const active = await page.evaluate(`(() => {
-            const el = document.activeElement;
-            if (!el) return null;
-            const rect = el.getBoundingClientRect();
-            return {
-              tag: String(el.tagName || "").toLowerCase(),
-              id: el.id || "",
-              name: el.getAttribute("name") || "",
-              role: el.getAttribute("role") || "",
-              ariaLabel: el.getAttribute("aria-label") || "",
-              type: el.getAttribute("type") || "",
-              placeholder: el.getAttribute("placeholder") || "",
-              text: String(el.textContent || "").trim().slice(0, 120),
-              isEditable: el.matches("input, textarea, [contenteditable=''], [contenteditable='true']")
-                || ["textbox", "searchbox", "combobox"].includes((el.getAttribute("role") || "").toLowerCase()),
-              rect: { x: rect.x, y: rect.y, width: rect.width, height: rect.height },
-            };
-          })()`);
+              const active = await describeFocusedEditable(page);
               if (!active || typeof active !== "object" || active.isEditable !== true) {
                 return fail("No editable focused element is active; use visual_click on a form field or pass a selector to type.", start2);
               }
               await page.keyboard.type(text, { delay: typingDelay });
               const label = active && typeof active === "object" ? `<${active.tag || "element"}>${active.id ? `#${active.id}` : ""}` : "focused element";
-              return ok(`Typed "${text}" into ${label}`, start2);
+              const frame = active["frame"];
+              return ok(`Typed "${text}" into ${label}${frame?.kind === "frame" ? ` in frame ${frame.index}` : ""}`, start2);
             }
             case "press": {
               const key = text || "Enter";
@@ -10824,7 +11046,14 @@ var init_playwright_browser = __esm({
               return ok(`Hovered: ${resolvedSelector}${resolvedSelector !== selector ? ` (from ${selector})` : ""}`, start2);
             }
             // ── Waiting ──
-            case "wait":
+            case "wait": {
+              if (selector) {
+                await page.waitForSelector(selector, { timeout: timeout2 });
+                return ok(`Element appeared: ${selector}`, start2);
+              }
+              await page.waitForTimeout(timeout2);
+              return ok(`Waited ${timeout2}ms`, start2);
+            }
             case "waitForSelector": {
               if (!selector)
                 return fail("selector is required", start2);
@@ -10893,9 +11122,13 @@ var init_playwright_browser = __esm({
             case "evaluate": {
               if (!text)
                 return fail("text (JavaScript code) is required", start2);
-              const result = await page.evaluate(text);
-              const serialized = typeof result === "string" ? result : JSON.stringify(result, null, 2);
-              return ok(serialized?.slice(0, 15e3) ?? "undefined", start2);
+              try {
+                const result = await page.evaluate(text);
+                const serialized = typeof result === "string" ? result : JSON.stringify(result, null, 2);
+                return ok(serialized?.slice(0, 15e3) ?? "undefined", start2);
+              } catch (err2) {
+                return fail(evaluateFailureMessage(err2, text), start2);
+              }
             }
             // ── Screenshot / PDF ──
             case "screenshot": {
@@ -11273,8 +11506,13 @@ ${JSON.stringify(data, null, 2)}`, start2);
           })()`);
               let clickSource = point.source || pointResult?.source || "vision";
               const candidate = await findBrowserVisualCandidate(page, visualTarget, cssX, cssY, false);
-              if (candidate) {
-                const candidateRecord = candidate;
+              let candidateRecord = candidate;
+              if (!candidateRecord) {
+                candidateRecord = await findBrowserVisualCandidate(page, visualTarget, cssX, cssY, true, true, true);
+                if (candidateRecord)
+                  await page.waitForTimeout(150);
+              }
+              if (candidateRecord) {
                 const center = candidateRecord["center"];
                 const nextX = Number(center?.x);
                 const nextY = Number(center?.y);
@@ -11282,7 +11520,7 @@ ${JSON.stringify(data, null, 2)}`, start2);
                   cssX = Math.max(0, Math.min(viewport.width, nextX));
                   cssY = Math.max(0, Math.min(viewport.height, nextY));
                   elementInfo = candidateRecord;
-                  clickSource = `${clickSource}+dom-candidate`;
+                  clickSource = `${clickSource}+dom-candidate${candidateRecord["scrolledIntoView"] === true ? "+scroll" : ""}`;
                 }
               }
               await page.mouse.move(cssX, cssY, { steps: 12 });
@@ -23994,8 +24232,8 @@ var init_explore_tools = __esm({
       enter_worktree: "Create isolated git worktree for safe parallel file modifications",
       exit_worktree: "Exit and optionally remove a git worktree (keep for merge or discard)",
       notebook_edit: "Edit Jupyter .ipynb notebooks at cell level (list, replace, insert, delete cells)",
-      browser_action: "Interactive browser: login, fill forms, click buttons, screenshot — session persists between calls; for console/page-error/network diagnostics prefer playwright_browser",
-      playwright_browser: "Full browser verification and visual action loop: observe_bundle, visual_click via Moondream pointing, focused-element typing for visual form filling, screenshot, page_errors, console_logs, network_log, DOM/accessibility, storage",
+      browser_action: "Interactive Selenium browser: login, fill forms, click buttons, screenshot — session persists between browser_action calls only; separate runtime from playwright_browser",
+      playwright_browser: "Full browser verification and visual action loop: observe_bundle, visual_click via Moondream pointing, selector/target fill, focused-element typing, screenshot, page_errors, console_logs, network_log, DOM/accessibility, storage",
       carbonyl_browser: "Terminal-rendered real browser automation via Carbonyl: navigate, read rendered text, click/type, sessions, daemon mode",
       scheduler: "Schedule tasks for automatic future execution via OS cron",
       cronjob: "Alias for scheduler: OS cron-backed time triggers",
@@ -284492,6 +284730,7 @@ async function ensureSession(options2 = {}) {
       }
       activeSessionId = null;
       activeSessionHeadless = null;
+      activeSessionUrl = null;
     }
   }
   if (activeSessionId) {
@@ -284503,6 +284742,13 @@ async function ensureSession(options2 = {}) {
     }
     activeSessionId = null;
     activeSessionHeadless = null;
+    activeSessionUrl = null;
+  }
+  if (options2.allowCreate === false) {
+    return {
+      error: "No active browser_action Selenium session exists for this action. browser_action is a separate browser/runtime from playwright_browser; continue the current page with playwright_browser, or call browser_action({action:'navigate', url: ...}) first.",
+      sessionId: ""
+    };
   }
   const headless = options2.headless ?? defaultBrowserHeadless();
   const res = await fetch(`${BASE_URL}/session/start`, {
@@ -284520,8 +284766,16 @@ async function ensureSession(options2 = {}) {
     return { error: String(data.message ?? "Failed to start browser session"), sessionId: "" };
   activeSessionId = data.session_id;
   activeSessionHeadless = headless;
+  activeSessionUrl = null;
   return { sessionId: activeSessionId };
 }
+function browserActionRuntimeHint() {
+  return [
+    "browser_action is a separate browser/runtime from playwright_browser and uses its own Selenium/Chrome session; it does not share page state, cookies, focus, or navigation.",
+    activeSessionUrl ? `Current browser_action URL: ${activeSessionUrl}` : "Current browser_action URL: unknown or not navigated.",
+    "If this page was opened with playwright_browser, keep using playwright_browser actions such as dom_summary, fill, type, press, visual_click, and observe_bundle."
+  ].join(" ");
+}
 async function apiCall(endpoint, method = "POST", body) {
   const options2 = {
     method,
@@ -284544,7 +284798,7 @@ async function apiCall(endpoint, method = "POST", body) {
   const res = await fetch(url, options2);
   return await res.json();
 }
-var __dirname3, DEFAULT_PORT, SCRAPE_SCRIPT, BASE_URL, serviceProcess, activeSessionId, activeSessionHeadless, BrowserActionTool;
+var __dirname3, DEFAULT_PORT, SCRAPE_SCRIPT, BASE_URL, serviceProcess, activeSessionId, activeSessionHeadless, activeSessionUrl, BrowserActionTool;
 var init_browser_action = __esm({
   "packages/execution/dist/tools/browser-action.js"() {
     "use strict";
@@ -284557,9 +284811,10 @@ var init_browser_action = __esm({
     serviceProcess = null;
     activeSessionId = null;
     activeSessionHeadless = null;
+    activeSessionUrl = null;
     BrowserActionTool = class {
       name = "browser_action";
-      description = "Control a persistent headless Chrome browser session for interactive web tasks. The browser stays open between calls, maintaining cookies, login state, and history. Use this (not web_fetch/web_crawl) when you need to: (1) log into a website, (2) fill and submit forms, (3) click buttons or links interactively, (4) take screenshots of rendered pages, (5) navigate multi-step workflows (checkout, signup, dashboards), (6) interact with elements that require JavaScript (dropdowns, modals, infinite scroll). Actions: navigate, click, click_xy, type, screenshot, dom, scroll, scroll_up, scroll_down, back, forward, close. For browser visuals, use browser_action({action:'screenshot', width, height, output_path}) — this captures the headless browser viewport, not the desktop. Use the desktop screenshot tool only when the actual OS screen is the target. For verification of browser runtime failures, prefer playwright_browser because it exposes page_errors, console_logs, network_log, DOM/accessibility, and screenshots from the same session. IMPORTANT: Start by calling navigate with the URL — do NOT ask the user for credentials or info first. Loopback URLs (localhost, 127.0.0.1, ::1) are allowed for local development servers; private LAN and metadata URLs remain blocked. Navigate to the page, then use dom/screenshot to see what's there, then type/click to interact. Call 'close' when done to free resources. This tool does not save or download arbitrary rendered files (PDFs, archives, media) to disk — clicking a 'Download' link inside the browser does not produce a local file path for the agent. For file acquisition, use the dedicated download/file tool and validate the resulting content-type and size before treating the result as success.";
+      description = "Control a persistent headless Chrome browser session for interactive web tasks. The browser stays open between calls, maintaining cookies, login state, and history. This is a separate Selenium/Chrome runtime from playwright_browser; do not switch between the two mid-workflow unless you intentionally navigate the second tool to the same URL. Use this (not web_fetch/web_crawl) when you need to: (1) log into a website, (2) fill and submit forms, (3) click buttons or links interactively, (4) take screenshots of rendered pages, (5) navigate multi-step workflows (checkout, signup, dashboards), (6) interact with elements that require JavaScript (dropdowns, modals, infinite scroll). Actions: navigate, click, click_xy, type, screenshot, dom, scroll, scroll_up, scroll_down, back, forward, close. For browser visuals, use browser_action({action:'screenshot', width, height, output_path}) — this captures the headless browser viewport, not the desktop. Use the desktop screenshot tool only when the actual OS screen is the target. For verification of browser runtime failures, prefer playwright_browser because it exposes page_errors, console_logs, network_log, DOM/accessibility, and screenshots from the same session. IMPORTANT: Start by calling navigate with the URL — do NOT ask the user for credentials or info first. Loopback URLs (localhost, 127.0.0.1, ::1) are allowed for local development servers; private LAN and metadata URLs remain blocked. Navigate to the page, then use dom/screenshot to see what's there, then type/click to interact. Call 'close' when done to free resources. This tool does not save or download arbitrary rendered files (PDFs, archives, media) to disk — clicking a 'Download' link inside the browser does not produce a local file path for the agent. For file acquisition, use the dedicated download/file tool and validate the resulting content-type and size before treating the result as success.";
       parameters = {
         type: "object",
         properties: {
@@ -284629,27 +284884,38 @@ var init_browser_action = __esm({
         const requestedWidth = args.width == null ? void 0 : asPositiveInt2(args.width, 1280, 320, 3840);
         const requestedHeight = args.height == null ? void 0 : asPositiveInt2(args.height, 720, 240, 2160);
         const requestedScale = args.device_scale_factor == null ? void 0 : asPositiveNumber(args.device_scale_factor, 1, 0.25, 3);
-        const launchErr = await launchService();
-        if (launchErr) {
-          return { success: false, output: "", error: launchErr, durationMs: Date.now() - start2 };
-        }
         if (action === "close") {
-          if (activeSessionId) {
+          if (activeSessionId || await probeService()) {
             try {
               await apiCall("/session/close");
             } catch {
             }
             activeSessionId = null;
             activeSessionHeadless = null;
+            activeSessionUrl = null;
           }
           return { success: true, output: "Browser session closed.", durationMs: Date.now() - start2 };
         }
+        const actionStartsSession = action === "navigate";
+        if (!actionStartsSession && !activeSessionId) {
+          return {
+            success: false,
+            output: "",
+            error: `browser_action ${action || "(missing action)"} requires an active browser_action session. ` + browserActionRuntimeHint(),
+            durationMs: Date.now() - start2
+          };
+        }
+        const launchErr = await launchService();
+        if (launchErr) {
+          return { success: false, output: "", error: launchErr, durationMs: Date.now() - start2 };
+        }
         const session = await ensureSession({
           width: requestedWidth,
           height: requestedHeight,
           deviceScaleFactor: requestedScale,
           headless: asOptionalBoolean2(args.headless),
-          forceNew: asOptionalBoolean2(args.force_new) === true
+          forceNew: asOptionalBoolean2(args.force_new) === true,
+          allowCreate: actionStartsSession
         });
         if (session.error) {
           return { success: false, output: "", error: session.error, durationMs: Date.now() - start2 };
@@ -284667,7 +284933,13 @@ var init_browser_action = __esm({
               }
               result = await apiCall("/navigate", "POST", { url: args.url });
               if (result.ok) {
-                return { success: true, output: `Navigated to ${args.url}`, durationMs: Date.now() - start2 };
+                activeSessionUrl = args.url;
+                return {
+                  success: true,
+                  output: `Navigated to ${args.url}
+Runtime: browser_action Selenium/Chrome session. Continue with browser_action for this page, or use playwright_browser separately after navigating it.`,
+                  durationMs: Date.now() - start2
+                };
               }
               const navMsg = String(result.message ?? "Navigation failed");
               const navHint = navMsg.toLowerCase().includes("connection") || navMsg.toLowerCase().includes("refused") || navMsg.toLowerCase().includes("err_connection") ? " (the URL appears unreachable — check if the target server is running and accepting connections)" : navMsg.toLowerCase().includes("timeout") ? " (page load timed out — try again or use a different URL)" : "";
@@ -284689,7 +284961,7 @@ var init_browser_action = __esm({
               return {
                 success: false,
                 output: `Click on ${args.selector} failed: ${clickMsg}`,
-                error: `browser_action click failed: ${clickMsg}. Try dom_summary first to see what selectors exist on the page.`,
+                error: `browser_action click failed: ${clickMsg}. Try dom_summary first to see what selectors exist on the page. ${browserActionRuntimeHint()}`,
                 durationMs: Date.now() - start2
               };
             }
@@ -284731,7 +285003,7 @@ var init_browser_action = __esm({
               return {
                 success: false,
                 output: `Type into ${args.selector} failed: ${typeMsg}`,
-                error: `browser_action type failed: ${typeMsg}. Verify the element is visible and is an input/textarea — use dom_summary to check.`,
+                error: `browser_action type failed: ${typeMsg}. Verify the element is visible and is an input/textarea — use dom_summary to check. ${browserActionRuntimeHint()}`,
                 durationMs: Date.now() - start2
               };
             }
@@ -284872,7 +285144,7 @@ var init_browser_action = __esm({
                 if (!pointResult || pointResult.points.length === 0) {
                   return {
                     success: false,
-                    output: `Vision could not find "${target}" on the page. Try using dom_summary to find the CSS selector instead.`,
+                    output: `Vision could not find "${target}" on the page. Try using dom_summary to find the CSS selector instead. ${browserActionRuntimeHint()}`,
                     error: "No point backend returned normalized coordinates.",
                     durationMs: Date.now() - start2
                   };

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.206",
+  "version": "1.0.208",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.206",
+      "version": "1.0.208",
       "bundleDependencies": [
         "image-to-ascii"
       ],
@@ -4565,9 +4565,19 @@
       }
     },
     "node_modules/js-yaml": {
-      "version": "4.1.1",
-      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.1.1.tgz",
-      "integrity": "sha512-qQKT4zQxXl8lLwBtHMWwaTcGfFOZviOJet3Oy/xmGk2gZH677CJM9EvtfdSkgWcATZhj/55JZ0rmy3myCT5lsA==",
+      "version": "4.2.0",
+      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.2.0.tgz",
+      "integrity": "sha512-ePWsvanv0DWuDRsW8dnt+R4jQ31SCRCQ7hhNcPXZPsoBZiemuZNYGf7adZdqX2D86j6rvKp3RpCxVTSb8WQlOw==",
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/puzrin"
+        },
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/nodeca"
+        }
+      ],
       "license": "MIT",
       "dependencies": {
         "argparse": "^2.0.1"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.206",
+  "version": "1.0.208",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",