@xbrowser/cli 1.27.0 → 1.28.0

package/dist/{chunk-YITK42HZ.js → chunk-OEC7ZSFW.js}

@@ -1285,9 +1285,12 @@ var visionTaskCommand = registerCommand({
         } catch {
         }
       }
-      const buffer = await page.screenshot({ type: "png" });
+      const shotQuality = Math.max(30, Math.min(95, parseInt(process.env.XBROWSER_VLM_SHOT_QUALITY ?? "75", 10) || 75));
+      const buffer = await page.screenshot({ type: "jpeg", quality: shotQuality });
       const b64 = buffer.toString("base64");
-      const { w, h } = pngDims(b64);
+      const vp = typeof page.viewportSize === "function" ? page.viewportSize() : null;
+      const w = vp?.width ?? pngDims(b64).w;
+      const h = vp?.height ?? pngDims(b64).h;
       const obsText = [
         `\u4EFB\u52A1\uFF1A${p.task}`,
         `\u622A\u56FE\u5C3A\u5BF8\uFF1A${w}x${h}\uFF08\u5750\u6807\u539F\u70B9\uFF1A\u5DE6\u4E0A\u89D2\uFF09`,
@@ -1296,7 +1299,7 @@ var visionTaskCommand = registerCommand({
 ${history.slice(-6).join("\n")}` : "\u5C1A\u65E0\u52A8\u4F5C\u3002"
       ].join("\n");
       const reply = await vlmAskRetry(creds, [
-        { type: "image", source: { type: "base64", media_type: "image/png", data: b64 } },
+        { type: "image", source: { type: "base64", media_type: shotQuality >= 100 ? "image/png" : "image/jpeg", data: b64 } },
         { type: "text", text: `${SYSTEM_PROMPT}
 
 ${obsText}` }
@@ -4740,6 +4743,25 @@ function getRefTarget(sessionKey2, ref) {
   if (!session || !target) return null;
   return { screenHash: session.screenHash, target };
 }
+function inheritStableRefs(key, rawTargets) {
+  const prev = sessions.get(key);
+  if (!prev || prev.targets.size === 0) {
+    return rawTargets.map((target, index) => ({ ...target, ref: `e${index + 1}` }));
+  }
+  const bySelector = /* @__PURE__ */ new Map();
+  let maxId = 0;
+  for (const [ref, target] of prev.targets) {
+    bySelector.set(target.selector, ref);
+    const id = Number(ref.slice(1));
+    if (Number.isFinite(id) && id > maxId) maxId = id;
+  }
+  let next = maxId + 1;
+  return rawTargets.map((target) => {
+    const inherited = bySelector.get(target.selector);
+    if (inherited) return { ...target, ref: inherited };
+    return { ...target, ref: `e${next++}` };
+  });
+}
 
 // src/utils/resolve-selector.ts
 function buildElementSelector(el) {
@@ -5364,10 +5386,7 @@ async function observePage(page, sessionId, options = {}) {
       { includeHidden: !!options.includeHidden, limit: options.limit ?? 80 }
     )
   ]);
-  const targets = raw.targets.map((target, index) => ({
-    ref: `e${index + 1}`,
-    ...target
-  }));
+  const targets = inheritStableRefs(sessionKey(sessionId), raw.targets);
   replaceRefs(sessionKey(sessionId), raw.screenHash, targets);
   return {
     url: page.url(),

package/dist/{chunk-IDXOIXQR.js → chunk-WK3INZLA.js}

@@ -74,6 +74,25 @@ function getRefTarget(sessionKey2, ref) {
   if (!session || !target) return null;
   return { screenHash: session.screenHash, target };
 }
+function inheritStableRefs(key, rawTargets) {
+  const prev = sessions.get(key);
+  if (!prev || prev.targets.size === 0) {
+    return rawTargets.map((target, index) => ({ ...target, ref: `e${index + 1}` }));
+  }
+  const bySelector = /* @__PURE__ */ new Map();
+  let maxId = 0;
+  for (const [ref, target] of prev.targets) {
+    bySelector.set(target.selector, ref);
+    const id = Number(ref.slice(1));
+    if (Number.isFinite(id) && id > maxId) maxId = id;
+  }
+  let next = maxId + 1;
+  return rawTargets.map((target) => {
+    const inherited = bySelector.get(target.selector);
+    if (inherited) return { ...target, ref: inherited };
+    return { ...target, ref: `e${next++}` };
+  });
+}
 
 // src/runtime/agent-runtime.ts
 function sessionKey(sessionId) {
@@ -225,10 +244,7 @@ async function observePage(page, sessionId, options = {}) {
       { includeHidden: !!options.includeHidden, limit: options.limit ?? 80 }
     )
   ]);
-  const targets = raw.targets.map((target, index) => ({
-    ref: `e${index + 1}`,
-    ...target
-  }));
+  const targets = inheritStableRefs(sessionKey(sessionId), raw.targets);
   replaceRefs(sessionKey(sessionId), raw.screenHash, targets);
   return {
     url: page.url(),
@@ -1637,9 +1653,12 @@ var visionTaskCommand = registerCommand({
         } catch {
         }
       }
-      const buffer = await page.screenshot({ type: "png" });
+      const shotQuality = Math.max(30, Math.min(95, parseInt(process.env.XBROWSER_VLM_SHOT_QUALITY ?? "75", 10) || 75));
+      const buffer = await page.screenshot({ type: "jpeg", quality: shotQuality });
       const b64 = buffer.toString("base64");
-      const { w, h } = pngDims(b64);
+      const vp = typeof page.viewportSize === "function" ? page.viewportSize() : null;
+      const w = vp?.width ?? pngDims(b64).w;
+      const h = vp?.height ?? pngDims(b64).h;
       const obsText = [
         `\u4EFB\u52A1\uFF1A${p.task}`,
         `\u622A\u56FE\u5C3A\u5BF8\uFF1A${w}x${h}\uFF08\u5750\u6807\u539F\u70B9\uFF1A\u5DE6\u4E0A\u89D2\uFF09`,
@@ -1648,7 +1667,7 @@ var visionTaskCommand = registerCommand({
 ${history.slice(-6).join("\n")}` : "\u5C1A\u65E0\u52A8\u4F5C\u3002"
       ].join("\n");
       const reply = await vlmAskRetry(creds, [
-        { type: "image", source: { type: "base64", media_type: "image/png", data: b64 } },
+        { type: "image", source: { type: "base64", media_type: shotQuality >= 100 ? "image/png" : "image/jpeg", data: b64 } },
         { type: "text", text: `${SYSTEM_PROMPT}
 
 ${obsText}` }

package/dist/cli.js

@@ -17,7 +17,7 @@ import {
   loadHooks,
   parsePluginParams,
   readJsonFile
-} from "./chunk-YITK42HZ.js";
+} from "./chunk-OEC7ZSFW.js";
 import "./chunk-JKVUFP3G.js";
 import {
   forwardCommandLog,
@@ -6297,7 +6297,7 @@ async function handleRemote(args, options, mode) {
   }
 }
 async function startMcpServer() {
-  const { startMcpStdio } = await import("./server-YKMRFDEM.js");
+  const { startMcpStdio } = await import("./server-JXF3JOAM.js");
   startMcpStdio();
 }
 

package/dist/daemon-main.js

@@ -1305,9 +1305,12 @@ var visionTaskCommand = registerCommand({
         } catch {
         }
       }
-      const buffer = await page.screenshot({ type: "png" });
+      const shotQuality = Math.max(30, Math.min(95, parseInt(process.env.XBROWSER_VLM_SHOT_QUALITY ?? "75", 10) || 75));
+      const buffer = await page.screenshot({ type: "jpeg", quality: shotQuality });
       const b64 = buffer.toString("base64");
-      const { w, h } = pngDims(b64);
+      const vp = typeof page.viewportSize === "function" ? page.viewportSize() : null;
+      const w = vp?.width ?? pngDims(b64).w;
+      const h = vp?.height ?? pngDims(b64).h;
       const obsText = [
         `\u4EFB\u52A1\uFF1A${p.task}`,
         `\u622A\u56FE\u5C3A\u5BF8\uFF1A${w}x${h}\uFF08\u5750\u6807\u539F\u70B9\uFF1A\u5DE6\u4E0A\u89D2\uFF09`,
@@ -1316,7 +1319,7 @@ var visionTaskCommand = registerCommand({
 ${history.slice(-6).join("\n")}` : "\u5C1A\u65E0\u52A8\u4F5C\u3002"
       ].join("\n");
       const reply = await vlmAskRetry(creds, [
-        { type: "image", source: { type: "base64", media_type: "image/png", data: b64 } },
+        { type: "image", source: { type: "base64", media_type: shotQuality >= 100 ? "image/png" : "image/jpeg", data: b64 } },
         { type: "text", text: `${SYSTEM_PROMPT}
 
 ${obsText}` }
@@ -4760,6 +4763,25 @@ function getRefTarget(sessionKey2, ref) {
   if (!session || !target) return null;
   return { screenHash: session.screenHash, target };
 }
+function inheritStableRefs(key, rawTargets) {
+  const prev = sessions.get(key);
+  if (!prev || prev.targets.size === 0) {
+    return rawTargets.map((target, index) => ({ ...target, ref: `e${index + 1}` }));
+  }
+  const bySelector = /* @__PURE__ */ new Map();
+  let maxId = 0;
+  for (const [ref, target] of prev.targets) {
+    bySelector.set(target.selector, ref);
+    const id = Number(ref.slice(1));
+    if (Number.isFinite(id) && id > maxId) maxId = id;
+  }
+  let next = maxId + 1;
+  return rawTargets.map((target) => {
+    const inherited = bySelector.get(target.selector);
+    if (inherited) return { ...target, ref: inherited };
+    return { ...target, ref: `e${next++}` };
+  });
+}
 
 // src/utils/resolve-selector.ts
 function buildElementSelector(el) {
@@ -5384,10 +5406,7 @@ async function observePage(page, sessionId, options = {}) {
       { includeHidden: !!options.includeHidden, limit: options.limit ?? 80 }
     )
   ]);
-  const targets = raw.targets.map((target, index) => ({
-    ref: `e${index + 1}`,
-    ...target
-  }));
+  const targets = inheritStableRefs(sessionKey(sessionId), raw.targets);
   replaceRefs(sessionKey(sessionId), raw.screenHash, targets);
   return {
     url: page.url(),

package/dist/index.js

@@ -48,7 +48,7 @@ import {
   splitCommand,
   waitForAIResponse,
   waitForPage
-} from "./chunk-IDXOIXQR.js";
+} from "./chunk-WK3INZLA.js";
 import {
   SessionRecorder
 } from "./chunk-IR6C24P7.js";
@@ -6392,7 +6392,7 @@ async function handleRemote(args, options, mode) {
   }
 }
 async function startMcpServer() {
-  const { startMcpStdio } = await import("./server-WDDGIV3C.js");
+  const { startMcpStdio } = await import("./server-26SE6YZR.js");
   startMcpStdio();
 }
 

package/dist/{server-WDDGIV3C.js → server-26SE6YZR.js}

@@ -1,7 +1,7 @@
 import {
   executeChain,
   executeCommand
-} from "./chunk-IDXOIXQR.js";
+} from "./chunk-WK3INZLA.js";
 import "./chunk-KECYFWJO.js";
 import "./chunk-JKVUFP3G.js";
 import "./chunk-3OD76SUE.js";

package/dist/{server-YKMRFDEM.js → server-JXF3JOAM.js}

@@ -1,7 +1,7 @@
 import {
   executeChain,
   executeCommand
-} from "./chunk-YITK42HZ.js";
+} from "./chunk-OEC7ZSFW.js";
 import "./chunk-JKVUFP3G.js";
 import "./chunk-3OD76SUE.js";
 import "./chunk-ZTHE5RBZ.js";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xbrowser/cli",
-  "version": "1.27.0",
+  "version": "1.28.0",
   "description": "Browser automation CLI for web scraping, headless browsing, SEO analysis, and AI agent workflows. A command-line alternative to Playwright, Puppeteer, and Selenium.",
   "type": "module",
   "bin": {