finch-ego-lite 0.1.2 → 0.1.4

package/dist/index.js

@@ -15,12 +15,18 @@ function egoBinary() {
 }
 async function processRunning() {
   if (process.platform !== "darwin") return false;
-  try {
-    await execFileAsync("pgrep", ["-f", "/ego lite.app/Contents/MacOS/ego lite"], { timeout: 2e3 });
-    return true;
-  } catch {
-    return false;
-  }
+  const probe = async () => {
+    try {
+      await execFileAsync("pgrep", ["-f", "/ego lite.app/Contents/MacOS/ego lite"], { timeout: 2e3 });
+      return true;
+    } catch {
+      return false;
+    }
+  };
+  const first = await probe();
+  if (first) return true;
+  await new Promise((resolve) => setTimeout(resolve, 300));
+  return probe();
 }
 async function runCommand(file, args, timeout = 2e4) {
   const execute = () => new Promise((resolve, reject) => {
@@ -121,6 +127,11 @@ async function launchEgo() {
   if (process.platform !== "darwin") throw new Error("Ego Lite currently supports macOS only.");
   await execFileAsync("open", ["-a", "ego lite"], { timeout: 5e3 });
 }
+async function focusEgo() {
+  if (process.platform !== "darwin") return;
+  await execFileAsync("open", ["-a", "ego lite"], { timeout: 5e3 }).catch(() => {
+  });
+}
 async function openWebsite() {
   const command = process.platform === "darwin" ? "open" : process.platform === "win32" ? "cmd" : "xdg-open";
   const args = process.platform === "win32" ? ["/c", "start", "", WEBSITE] : [WEBSITE];
@@ -136,6 +147,7 @@ var ACTION_STAGES = {
   press_key: "acting",
   scroll: "scrolling",
   screenshot: "shooting",
+  switch_tab: "switching",
   close_space: "closing",
   run: "running"
 };
@@ -159,7 +171,14 @@ function buildScript(action, input, spaceId) {
     case "open_url": {
       const name = input.space_name || "web task";
       const timeout = input.timeout_seconds ?? 20;
+      if (spaceId !== void 0) {
+        return `await switchTaskSpace(${spaceId});
+await openOrReuseTab(${q(input.url ?? "")}, { wait: true, timeout: ${timeout} });
+const tab = await currentTab();
+cliLog(JSON.stringify({ spaceId: ${spaceId}, url: tab.url, title: tab.title }));`;
+      }
       return `const space = await useOrCreateTaskSpace(${q(name)});
+await switchTaskSpace(space.id);
 await openOrReuseTab(${q(input.url ?? "")}, { wait: true, timeout: ${timeout} });
 const tab = await currentTab();
 cliLog(JSON.stringify({ spaceId: space.id, url: tab.url, title: tab.title }));`;
@@ -206,6 +225,12 @@ cliLog(JSON.stringify({ ok: true, sy: info.sy }));`;
       return `${spaceLine}
 const path = await captureScreenshot();
 cliLog('__EGO_SHOT__' + path);`;
+    case "switch_tab":
+      return `await switchTaskSpace(${spaceId});
+await switchTab(${q(input.target_id ?? "")});
+await waitForLoad({ timeout: 15 }).catch(() => {});
+const info = await pageInfo();
+cliLog(JSON.stringify({ ok: true, spaceId: ${spaceId}, url: info.url, title: info.title }));`;
     case "close_space":
       return `await completeTaskSpace(${spaceId}, { keep: false });
 cliLog('ok');`;
@@ -274,9 +299,26 @@ function activate(ctx) {
     }
   }));
   let cachedStatus;
+  let lastRefreshAt = 0;
   let refreshInFlight;
   let pendingDetailedRefresh = false;
-  let refreshStatus = () => {
+  const refreshStatus = (detailed = false) => {
+    if (refreshInFlight) {
+      if (detailed) pendingDetailedRefresh = true;
+      return refreshInFlight;
+    }
+    refreshInFlight = getStatus(detailed).then((status) => {
+      cachedStatus = status;
+      lastRefreshAt = Date.now();
+      action.notifyUpdate();
+    }).catch((error) => ctx.logger.warn("Ego background status refresh failed", error)).finally(() => {
+      refreshInFlight = void 0;
+      if (pendingDetailedRefresh) {
+        pendingDetailedRefresh = false;
+        refreshStatus(true);
+      }
+    });
+    return refreshInFlight;
   };
   const action = ctx.composerActions.register("ego-browser", {
     async getBadge() {
@@ -285,7 +327,9 @@ function activate(ctx) {
       return online ? { text: t("badge.online"), active: true } : { text: t("badge.offline"), active: false };
     },
     async getMenu() {
-      refreshStatus(true);
+      if (cachedStatus === void 0 || Date.now() - lastRefreshAt > 3e3) {
+        await refreshStatus(true);
+      }
       if (!cachedStatus) {
         return [{
           id: "__checking__",
@@ -452,7 +496,6 @@ ${domainOf(tab.url, t("site.unknown"))}`,
           if (!binary || !await installedVersion(binary)) throw new Error(t("error.notInstalled"));
           const script = `const result = await completeTaskSpace(${JSON.stringify(spaceId)}, { keep: false }); cliLog(JSON.stringify(result));`;
           await runCommand(binary, ["nodejs", "-e", script], 15e3);
-          cachedStatus = void 0;
           action.notifyUpdate();
           refreshStatus(true);
           await ctx.ui.showToast({ title: t("toast.closed"), variant: "success", position: "TC" });
@@ -492,6 +535,7 @@ action:
   press_key   \u2014 press a key: Enter, Tab, Escape, ArrowDown... (space_id?, key)
   scroll      \u2014 scroll down by pixels (space_id?, delta_y?)
   screenshot  \u2014 capture the page and return it as an image (space_id?)
+  switch_tab  \u2014 switch to a task space and activate one of its pages (space_id, target_id from status)
   extract     \u2014 extract innerText of a CSS selector, or the whole page (space_id?, selector?)
   close_space \u2014 close a task space (space_id)
   run         \u2014 advanced: run a raw Ego Browser script (script, timeout_seconds?)
@@ -500,10 +544,11 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
     inputSchema: {
       type: "object",
       properties: {
-        action: { type: "string", enum: ["status", "open_url", "snapshot", "click", "fill", "type", "press_key", "scroll", "screenshot", "extract", "close_space", "run"] },
+        action: { type: "string", enum: ["status", "open_url", "snapshot", "click", "fill", "type", "press_key", "scroll", "screenshot", "switch_tab", "extract", "close_space", "run"] },
         url: { type: "string", description: "URL to open (action=open_url)." },
         space_id: { type: "number", description: "Task space id from status/open_url. Optional when only one agent space exists." },
         space_name: { type: "string", description: "Semantic name for the task space (action=open_url)." },
+        target_id: { type: "string", description: "Tab target id from action=status (action=switch_tab)." },
         selector: { type: "string", description: "CSS selector of the target element (click/fill/type/extract)." },
         text: { type: "string", description: "Text to type (action=type)." },
         value: { type: "string", description: "Value to fill into the field (action=fill)." },
@@ -514,7 +559,7 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
       },
       required: ["action"]
     },
-    risk: "high",
+    risk: "medium",
     defaultEnabled: true,
     async execute(input, exec) {
       const request = input;
@@ -548,6 +593,9 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
       if (actionName === "type" && !request.text) {
         return { content: [{ type: "text", text: "action=type requires text." }], isError: true };
       }
+      if (actionName === "switch_tab" && !request.target_id) {
+        return { content: [{ type: "text", text: "action=switch_tab requires target_id from action=status." }], isError: true };
+      }
       if (process.platform !== "darwin") {
         return { content: [{ type: "text", text: t("status.unsupported", { platform: process.platform }) }], isError: true };
       }
@@ -560,9 +608,9 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
         };
       }
       let spaceId;
-      if (actionName !== "open_url" && actionName !== "run") {
+      if (actionName !== "run") {
         spaceId = await resolveSpaceId(request.space_id);
-        if (spaceId === void 0) {
+        if (spaceId === void 0 && actionName !== "open_url") {
           return {
             content: [{ type: "text", text: "No task space resolved. Call action=status or action=open_url first, then pass its space_id." }],
             isError: true
@@ -573,15 +621,16 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
         exec.progress.report({ stage: "launching", message: pickProgress(t, "launching") });
         await launchEgo();
         await new Promise((resolve) => setTimeout(resolve, 1800));
+      } else {
+        await focusEgo();
       }
       const script = buildScript(actionName, request, spaceId);
       const timeout = Math.min(300, Math.max(1, request.timeout_seconds ?? 60)) * 1e3;
       exec.progress.report({ stage: ACTION_STAGES[actionName], message: pickProgress(t, ACTION_STAGES[actionName] ?? "browsing") });
       try {
         const output = await runCommand(binary, ["nodejs", "-e", script], timeout);
-        cachedStatus = void 0;
         action.notifyUpdate();
-        refreshStatus(false);
+        refreshStatus(true);
         if (actionName === "screenshot") {
           const path = output.split("__EGO_SHOT__").pop()?.split("\n")[0]?.trim();
           if (!path || !existsSync(path)) {
@@ -597,32 +646,15 @@ Reuse task spaces, respect user ownership, verify meaningful actions, and comple
         }
         return { content: [{ type: "text", text: output || "ok" }] };
       } catch (error) {
-        cachedStatus = void 0;
         action.notifyUpdate();
-        refreshStatus(false);
+        refreshStatus(true);
         const message = error instanceof Error ? error.message : String(error);
         return { content: [{ type: "text", text: message }], isError: true };
       }
     }
   }));
-  refreshStatus = (detailed = false) => {
-    if (refreshInFlight) {
-      pendingDetailedRefresh ||= detailed;
-      return;
-    }
-    refreshInFlight = getStatus(detailed).then((status) => {
-      cachedStatus = status;
-      action.notifyUpdate();
-    }).catch((error) => ctx.logger.warn("Ego background status refresh failed", error)).finally(() => {
-      refreshInFlight = void 0;
-      if (pendingDetailedRefresh) {
-        pendingDetailedRefresh = false;
-        refreshStatus(true);
-      }
-    });
-  };
   refreshStatus(true);
-  const timer = setInterval(() => refreshStatus(true), 1e4);
+  const timer = setInterval(() => refreshStatus(true), 5e3);
   ctx.subscriptions.push({ dispose: () => clearInterval(timer) });
   ctx.logger.info("ego-browser mini tool activated", { cached: Boolean(cachedStatus) });
 }

package/i18n/en-US.json

@@ -104,6 +104,7 @@
     "acting": "Clicking gently…|Tapping the button…|Filling things in…",
     "scrolling": "Scrolling down…|Rolling the wheel…",
     "shooting": "Saying cheese! 📸|Freezing the moment…",
+    "switching": "Switching windows…|Jumping to another page…|Ego is changing its view…",
     "closing": "Tidying up…|Closing the tab circus…|Wrapping up…",
     "running": "Ego is doing its thing…|Browser magic in progress…",
     "browsing": "Agent is working in Ego Browser…"

package/i18n/zh-CN.json

@@ -104,6 +104,7 @@
     "acting": "轻轻点了一下…|正在按下按钮…|正在填写…",
     "scrolling": "正在往下滚动…|鼠标滚轮转起来了…",
     "shooting": "咔嚓！拍照留念 📸|定格这一刻…",
+    "switching": "换个窗口看看…|跳到另一个页面…|Ego 正在切换视野…",
     "closing": "正在收摊…|正在收拾战场…|任务完成，整理中…",
     "running": "Ego 正在忙活…|浏览器魔法进行中…",
     "browsing": "Agent 正在操作 Ego Browser…"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "finch-ego-lite",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "description": "Let Finch browse the web in isolated Ego Lite task spaces while you follow every page it opens.",
   "author": {
     "name": "PuterJam",

package/skills/ego-browser/SKILL.md

@@ -2,7 +2,7 @@
 name: ego-browser
 description: Use Ego Browser whenever the user needs to open a website, navigate pages, click, type, upload, take screenshots, extract page data, test web apps, log in, or automate browser actions. Ego gives Agents isolated task spaces that reuse the user's login state without competing with normal browser windows.
 metadata:
-  version: "0.1.2"
+  version: "0.1.3"
 ---
 
 # Ego Browser 工具使用指南
@@ -31,6 +31,7 @@ metadata:
 | press_key | space_id?, key | 按键：Enter/Tab/Escape… |
 | scroll | space_id?, delta_y? | 向下滚动（默认 300px） |
 | screenshot | space_id? | 截图并作为图片返回 |
+| switch_tab | space_id, target_id | 先切到任务空间，再激活其中某个页面（target_id 来自 status 的 tabs） |
 | extract | space_id?, selector? | 提取区域 innerText 或整页文本 |
 | close_space | space_id | 关闭任务空间 |
 | run | script, timeout_seconds? | 高级：执行原始 Ego 脚本 |