diff --git a/apps/desktop-client/src/main/adapters/wenxin.test.ts b/apps/desktop-client/src/main/adapters/wenxin.test.ts
index 9232001..b9b5aad 100644
--- a/apps/desktop-client/src/main/adapters/wenxin.test.ts
+++ b/apps/desktop-client/src/main/adapters/wenxin.test.ts
@@ -3,8 +3,10 @@ import { describe, expect, it } from "vitest";
 import { __wenxinTestUtils } from "./wenxin";
 
 const {
+  buildSourceItem,
   htmlTableToMarkdown,
   isObservationComplete,
+  normalizeUrl,
   parseWenxinHistoryResponse,
   parseWenxinSSEBody,
 } = __wenxinTestUtils;
@@ -163,6 +165,62 @@ describe("wenxin adapter helpers", () => {
     expect(history.citations[0]?.url).toBe("https://fallback.example.com/article");
   });
 
+  it("unwraps Wenxin reference redirect urls", () => {
+    expect(normalizeUrl("https://yiyan.baidu.com/eb/link?target=https%3A%2F%2Fwww.example.com%2Farticle%23ref")).toBe(
+      "https://www.example.com/article",
+    );
+
+    expect(buildSourceItem({
+      pageUrl: "https://yiyan.baidu.com/eb/link?url=https%3A%2F%2Fsource.example.com%2Fnews%3Fid%3D1",
+      title: "引用来源标题",
+      source: "来源站点",
+    })).toMatchObject({
+      url: "https://source.example.com/news?id=1",
+      title: "引用来源标题",
+      site_name: "来源站点",
+      host: "source.example.com",
+    });
+  });
+
+  it("extracts web page references from Wenxin reference list fields", () => {
+    const history = parseWenxinHistoryResponse({
+      code: 0,
+      msg: "success",
+      data: {
+        state: 1,
+        chats: [
+          {
+            id: "bot-1",
+            role: "robot",
+            stop: 1,
+            modelSign: "EB45T",
+            message: [
+              {
+                contentType: "text",
+                content: "这是答案",
+              },
+            ],
+            webPages: [
+              {
+                targetUrl: "https://yiyan.baidu.com/eb/link?target=https%3A%2F%2Fwww.reference-source.com%2Fa%23section",
+                title: "参考网页 A",
+                siteName: "reference-source.com",
+              },
+            ],
+          },
+        ],
+      },
+    }, "session-3");
+
+    expect(history.citations).toHaveLength(1);
+    expect(history.searchResults).toHaveLength(1);
+    expect(history.citations[0]).toMatchObject({
+      url: "https://www.reference-source.com/a",
+      title: "参考网页 A",
+      site_name: "reference-source.com",
+    });
+  });
+
   it("does not treat error-only observations as complete", () => {
     expect(isObservationComplete({
       sessionId: "session-1",
diff --git a/apps/desktop-client/src/main/adapters/wenxin.ts b/apps/desktop-client/src/main/adapters/wenxin.ts
index 5d28fe9..b26e655 100644
--- a/apps/desktop-client/src/main/adapters/wenxin.ts
+++ b/apps/desktop-client/src/main/adapters/wenxin.ts
@@ -29,8 +29,28 @@ const WENXIN_LOGIN_REQUIRED_PATTERN = /(请先登录|未登录|登录后|登录
 const WENXIN_PLACEHOLDER_PATTERN = /^(?:正在生成中(?:\.\.\.)?|思考中(?:\.\.\.)?|内容由AI生成，仅供参考，请仔细甄别|全选)$/;
 const WENXIN_TABLE_HTML_PATTERN = /<table\b[\s\S]*?<\/table>/gi;
 const WENXIN_REFERENCE_TRIGGER_PATTERN = /(参考\s*\d+\s*(?:个网页|条网页信息源)|found\s*\d+\s*web pages)/i;
-const WENXIN_CITATION_FIELD_PATTERN = /^(?:citations?|citationInfo|pluginCitations?|pluginCitation|reference(?:s|List|_list|Cards|_cards|Docs|_docs)?|source(?:s|List|_list|Cards|_cards)?)$/i;
-const WENXIN_SEARCH_FIELD_PATTERN = /^(?:searchCitations?|search_results?|searchResultList|search_result_list|webListData|webList|browseReferences?|browse_results?|grounding(?:_results?)?)$/i;
+const WENXIN_CITATION_FIELD_PATTERN = /^(?:citations?|citationInfo|pluginCitations?|pluginCitation|reference(?:s|List|_list|Cards|_cards|Docs|_docs|Info|_info)?|source(?:s|List|_list|Cards|_cards|Info|_info)?|web(?:References?|Sources?|Pages?|Docs?|Results?)|page(?:References?|Sources?))$/i;
+const WENXIN_SEARCH_FIELD_PATTERN = /^(?:searchCitations?|search_results?|searchResultList|search_result_list|webListData|webList|web(?:References?|Sources?|Pages?|Docs?|Results?)|browseReferences?|browse_results?|grounding(?:_results?)?)$/i;
+const WENXIN_REDIRECT_QUERY_KEYS = [
+  "url",
+  "u",
+  "target",
+  "targetUrl",
+  "target_url",
+  "link",
+  "href",
+  "sourceUrl",
+  "source_url",
+  "originUrl",
+  "origin_url",
+  "realUrl",
+  "real_url",
+  "redirect",
+  "redirect_url",
+  "jump",
+  "jump_url",
+  "to",
+];
 
 type WenxinPageSourceLink = {
   url: string;
@@ -165,19 +185,85 @@ function normalizeBlockText(value: string | null): string | null {
     .trim() || null;
 }
 
+function decodeUrlCandidate(value: string): string {
+  let decoded = value.trim();
+  for (let index = 0; index < 3; index += 1) {
+    try {
+      const next = decodeURIComponent(decoded);
+      if (next === decoded) {
+        break;
+      }
+      decoded = next;
+    } catch {
+      break;
+    }
+  }
+  return decoded;
+}
+
+function coercePotentialUrl(value: string): string {
+  const trimmed = decodeUrlCandidate(value);
+  if (/^\/\//.test(trimmed)) {
+    return `https:${trimmed}`;
+  }
+  if (/^[a-z0-9-]+(?:\.[a-z0-9-]+)+(?:[/?#].*)?$/i.test(trimmed)) {
+    return `https://${trimmed}`;
+  }
+  return trimmed;
+}
+
+function normalizeAbsoluteUrl(value: string): string | null {
+  try {
+    const url = new URL(coercePotentialUrl(value));
+    url.hash = "";
+    return url.toString();
+  } catch {
+    return null;
+  }
+}
+
+function extractRedirectTargetUrl(value: string): string | null {
+  const direct = normalizeAbsoluteUrl(value);
+  if (!direct) {
+    return null;
+  }
+
+  try {
+    const url = new URL(direct);
+    for (const key of WENXIN_REDIRECT_QUERY_KEYS) {
+      const raw = url.searchParams.get(key);
+      if (!raw) {
+        continue;
+      }
+      const normalized = normalizeAbsoluteUrl(raw);
+      if (normalized) {
+        return normalized;
+      }
+    }
+  } catch {
+    return null;
+  }
+
+  const encodedMatch = direct.match(/https?%3A%2F%2F[^&\s"'<>]+/i);
+  if (encodedMatch) {
+    return normalizeAbsoluteUrl(encodedMatch[0]);
+  }
+
+  const plainMatch = direct.match(/https?:\/\/[^&\s"'<>]+/i);
+  if (plainMatch && plainMatch[0] !== direct) {
+    return normalizeAbsoluteUrl(plainMatch[0]);
+  }
+
+  return null;
+}
+
 function normalizeUrl(value: unknown): string | null {
   const input = normalizeText(value);
   if (!input) {
     return null;
   }
 
-  try {
-    const url = new URL(input);
-    url.hash = "";
-    return url.toString();
-  } catch {
-    return null;
-  }
+  return extractRedirectTargetUrl(input) ?? normalizeAbsoluteUrl(input);
 }
 
 function isLikelyWenxinSourceUrl(url: string | null): boolean {
@@ -351,6 +437,14 @@ function buildSourceItem(input: unknown): MonitoringSourceItem | null {
     input.url
     ?? input.href
     ?? input.link
+    ?? input.pageUrl
+    ?? input.page_url
+    ?? input.webUrl
+    ?? input.web_url
+    ?? input.realUrl
+    ?? input.real_url
+    ?? input.displayUrl
+    ?? input.display_url
     ?? input.originUrl
     ?? input.origin_url
     ?? input.targetUrl
@@ -1100,7 +1194,7 @@ async function ensurePageOnWenxinChat(page: PlaywrightPage, signal: AbortSignal)
 
 async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageSnapshot> {
   return await page.evaluate(
-    ({ editorSelector, loginSignals, triggerPatternSource, citationFieldPatternSource, searchFieldPatternSource }) => {
+    ({ editorSelector, loginSignals, triggerPatternSource, citationFieldPatternSource, searchFieldPatternSource, redirectQueryKeys }) => {
       const isObjectRecord = (value: unknown): value is Record<string, unknown> =>
         typeof value === "object" && value !== null && !Array.isArray(value);
 
@@ -1116,6 +1210,73 @@ async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageS
         return trimmed || null;
       };
 
+      const decodeUrlCandidate = (value: string): string => {
+        let decoded = value.trim();
+        for (let index = 0; index < 3; index += 1) {
+          try {
+            const next = decodeURIComponent(decoded);
+            if (next === decoded) {
+              break;
+            }
+            decoded = next;
+          } catch {
+            break;
+          }
+        }
+        return decoded;
+      };
+
+      const coercePotentialUrl = (value: string): string => {
+        const trimmed = decodeUrlCandidate(value);
+        if (/^\/\//.test(trimmed)) {
+          return `https:${trimmed}`;
+        }
+        if (/^[a-z0-9-]+(?:\.[a-z0-9-]+)+(?:[/?#].*)?$/i.test(trimmed)) {
+          return `https://${trimmed}`;
+        }
+        return trimmed;
+      };
+
+      const normalizeAbsoluteHref = (value: string): string | null => {
+        try {
+          const url = new URL(coercePotentialUrl(value), window.location.href);
+          url.hash = "";
+          return url.toString();
+        } catch {
+          return null;
+        }
+      };
+
+      const extractRedirectTargetHref = (value: string): string | null => {
+        const direct = normalizeAbsoluteHref(value);
+        if (!direct) {
+          return null;
+        }
+
+        try {
+          const url = new URL(direct);
+          for (const key of redirectQueryKeys) {
+            const raw = url.searchParams.get(key);
+            if (!raw) {
+              continue;
+            }
+            const normalized = normalizeAbsoluteHref(raw);
+            if (normalized) {
+              return normalized;
+            }
+          }
+        } catch {
+          return null;
+        }
+
+        const encodedMatch = direct.match(/https?%3A%2F%2F[^&\s"'<>]+/i);
+        if (encodedMatch) {
+          return normalizeAbsoluteHref(encodedMatch[0]);
+        }
+
+        return null;
+      };
+
       const normalizeHref = (value: unknown): string | null => {
         if (typeof value !== "string") {
           return null;
@@ -1126,13 +1287,7 @@ async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageS
           return null;
         }
 
-        try {
-          const url = new URL(trimmed, window.location.href);
-          url.hash = "";
-          return url.toString();
-        } catch {
-          return null;
-        }
+        return extractRedirectTargetHref(trimmed) ?? normalizeAbsoluteHref(trimmed);
       };
 
       const isLikelySourceUrl = (value: string | null): value is string => {
@@ -1169,6 +1324,14 @@ async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageS
           input.url
           ?? input.href
           ?? input.link
+          ?? input.pageUrl
+          ?? input.page_url
+          ?? input.webUrl
+          ?? input.web_url
+          ?? input.realUrl
+          ?? input.real_url
+          ?? input.displayUrl
+          ?? input.display_url
           ?? input.originUrl
           ?? input.origin_url
           ?? input.targetUrl
@@ -1331,39 +1494,225 @@ async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageS
         };
       };
 
+      const urlAttributeNames = [
+        "href",
+        "data-href",
+        "data-url",
+        "data-link",
+        "data-target-url",
+        "data-source-url",
+        "data-origin-url",
+        "data-real-url",
+        "data-page-url",
+        "data-web-url",
+        "data-redirect",
+        "data-jump-url",
+        "data-log",
+        "data-click",
+        "data-extra",
+        "data-info",
+        "data-value",
+      ];
+
+      const firstUrlFromText = (value: string | null): string | null => {
+        if (!value) {
+          return null;
+        }
+
+        const httpMatch = value.match(/https?:\/\/[^\s"'<>，。；、]+/i);
+        if (httpMatch) {
+          return normalizeHref(httpMatch[0]);
+        }
+
+        const domainMatch = value.match(/\b[a-z0-9-]+(?:\.[a-z0-9-]+)+(?:\/[^\s"'<>，。；、]*)?/i);
+        return domainMatch ? normalizeHref(domainMatch[0]) : null;
+      };
+
+      const readElementUrl = (element: Element): string | null => {
+        const candidates: string[] = [];
+        for (const name of urlAttributeNames) {
+          const value = element.getAttribute(name);
+          if (value) {
+            candidates.push(value);
+          }
+        }
+
+        if (element instanceof HTMLElement) {
+          candidates.push(...Object.values(element.dataset).filter((value): value is string => Boolean(value)));
+        }
+
+        for (const candidate of candidates) {
+          const url = normalizeHref(candidate) ?? firstUrlFromText(candidate);
+          if (isLikelySourceUrl(url)) {
+            return url;
+          }
+        }
+
+        const text = normalize(element.textContent ?? "");
+        if (text && text.length <= 2_000) {
+          const url = firstUrlFromText(text);
+          if (isLikelySourceUrl(url)) {
+            return url;
+          }
+        }
+
+        return null;
+      };
+
+      const normalizeShortText = (value: unknown, maxLength = 220): string | null => {
+        const text = normalize(value);
+        if (!text) {
+          return null;
+        }
+        return text.length <= maxLength ? text : null;
+      };
+
+      const extractElementSourceLink = (element: Element) => {
+        if (!isVisible(element)) {
+          return null;
+        }
+
+        const anchorLink = extractAnchorLink(element);
+        if (anchorLink) {
+          return anchorLink;
+        }
+
+        const url = readElementUrl(element);
+        if (!isLikelySourceUrl(url)) {
+          return null;
+        }
+
+        const card = element.closest([
+          "a",
+          "li",
+          "article",
+          "section",
+          "[role=\"listitem\"]",
+          "[class*=\"item\"]",
+          "[class*=\"Item\"]",
+          "[class*=\"card\"]",
+          "[class*=\"Card\"]",
+          "[class*=\"source\"]",
+          "[class*=\"Source\"]",
+          "[class*=\"reference\"]",
+          "[class*=\"Reference\"]",
+          "[class*=\"web\"]",
+          "[class*=\"Web\"]",
+        ].join(",")) ?? element;
+
+        const title = normalizeShortText(
+          card.querySelector("[class*=\"title\"], [class*=\"Title\"], h1, h2, h3, h4")?.textContent
+          ?? element.getAttribute("title")
+          ?? element.getAttribute("aria-label")
+          ?? card.textContent,
+        );
+        const siteName = normalizeShortText(
+          card.querySelector("[class*=\"site\"], [class*=\"Site\"], [class*=\"domain\"], [class*=\"Domain\"], [class*=\"source\"], [class*=\"Source\"]")?.textContent
+          ?? element.getAttribute("data-site-name")
+          ?? (element instanceof HTMLElement ? element.dataset.siteName : null),
+          80,
+        );
+
+        return {
+          url,
+          title,
+          text: title ?? siteName,
+          siteName,
+        };
+      };
+
+      const collectSourceLinksFromElement = (root: HTMLElement) => {
+        const candidates = [
+          root,
+          ...Array.from(root.querySelectorAll([
+            "a[href]",
+            "[data-href]",
+            "[data-url]",
+            "[data-link]",
+            "[data-target-url]",
+            "[data-source-url]",
+            "[data-origin-url]",
+            "[data-real-url]",
+            "[data-page-url]",
+            "[data-web-url]",
+            "[data-log]",
+            "[data-click]",
+            "[data-extra]",
+            "[data-info]",
+            "li",
+            "article",
+            "[role=\"listitem\"]",
+            "[class*=\"item\"]",
+            "[class*=\"Item\"]",
+            "[class*=\"card\"]",
+            "[class*=\"Card\"]",
+            "[class*=\"source\"]",
+            "[class*=\"Source\"]",
+            "[class*=\"reference\"]",
+            "[class*=\"Reference\"]",
+            "[class*=\"web\"]",
+            "[class*=\"Web\"]",
+          ].join(","))),
+        ];
+
+        return dedupeLinks(
+          candidates
+            .map((node) => extractElementSourceLink(node))
+            .filter((node): node is NonNullable<typeof node> => Boolean(node)),
+        );
+      };
+
       const collectVisiblePanelLinks = (triggerText: string | null) => {
         const panelSelectors = [
+          "[role=\"dialog\"]",
+          "[role=\"list\"]",
+          "[class*=\"modal\"]",
+          "[class*=\"Modal\"]",
+          "[class*=\"dialog\"]",
+          "[class*=\"Dialog\"]",
           "[class*=\"drawer\"]",
+          "[class*=\"Drawer\"]",
           "[class*=\"panel\"]",
+          "[class*=\"Panel\"]",
           "[class*=\"popover\"]",
+          "[class*=\"Popover\"]",
+          "[class*=\"popup\"]",
+          "[class*=\"Popup\"]",
           "[class*=\"webList\"]",
+          "[class*=\"WebList\"]",
           "[class*=\"reference\"]",
+          "[class*=\"Reference\"]",
           "[class*=\"material\"]",
           "[class*=\"cite\"]",
+          "[class*=\"source\"]",
+          "[class*=\"Source\"]",
+          "[class*=\"web\"]",
+          "[class*=\"Web\"]",
           "aside",
           "section",
           "article",
+          "body > div",
         ];
 
+        const rootLinks = new Map<HTMLElement, ReturnType<typeof collectSourceLinksFromElement>>();
         const roots = Array.from(document.querySelectorAll(panelSelectors.join(",")))
           .filter((element): element is HTMLElement => isVisible(element))
           .filter((element) => {
             const text = normalize(element.innerText) ?? "";
-            const externalAnchorCount = Array.from(element.querySelectorAll("a[href]"))
-              .filter((node) => extractAnchorLink(node) !== null)
-              .length;
-            return externalAnchorCount > 0
+            const className = String(element.className || "");
+            const links = collectSourceLinksFromElement(element);
+            rootLinks.set(element, links);
+            return links.length > 0
               && (
                 /参考|网页|信息源|引用|web pages|references/i.test(text)
                 || Boolean(triggerText && text.includes(triggerText))
+                || element.getAttribute("role") === "dialog"
+                || /(drawer|panel|popover|popup|modal|dialog|web|source|reference|cite)/i.test(className)
               );
           });
 
         const links = roots.length
-          ? roots.flatMap((root) =>
-            Array.from(root.querySelectorAll("a[href]"))
-              .map((node) => extractAnchorLink(node))
-              .filter((node): node is NonNullable<typeof node> => Boolean(node)))
+          ? roots.flatMap((root) => rootLinks.get(root) ?? collectSourceLinksFromElement(root))
           : [];
 
         return dedupeLinks(links);
@@ -1464,6 +1813,7 @@ async function readWenxinPageSnapshot(page: PlaywrightPage): Promise<WenxinPageS
       triggerPatternSource: WENXIN_REFERENCE_TRIGGER_PATTERN.source,
       citationFieldPatternSource: WENXIN_CITATION_FIELD_PATTERN.source,
       searchFieldPatternSource: WENXIN_SEARCH_FIELD_PATTERN.source,
+      redirectQueryKeys: WENXIN_REDIRECT_QUERY_KEYS,
     },
   );
 }
@@ -1500,17 +1850,58 @@ async function maybeOpenWenxinReferencePanel(page: PlaywrightPage): Promise<void
     };
 
     const triggerPattern = new RegExp(triggerPatternSource, "i");
-    const trigger = Array.from(document.querySelectorAll("button, [role=\"button\"], a, div, span, p"))
+    const rawCandidates = Array.from(document.querySelectorAll("button, [role=\"button\"], a, [tabindex], div, span, p"))
       .filter((element): element is HTMLElement => isVisible(element))
-      .find((element) => {
+      .filter((element) => {
         const text = normalize(element.innerText || element.textContent || "");
         return Boolean(text && triggerPattern.test(text));
       });
+    const candidates = rawCandidates
+      .map((element) => {
+        const target = element.closest("button, [role=\"button\"], a, [tabindex]") ?? element;
+        if (!(target instanceof HTMLElement) || !isVisible(target)) {
+          return null;
+        }
+
+        const rect = target.getBoundingClientRect();
+        const text = normalize(target.innerText || target.textContent || "") ?? "";
+        const style = window.getComputedStyle(target);
+        const className = String(target.className || "");
+        let score = 0;
+        if (target.matches("button, [role=\"button\"], a, [tabindex]")) {
+          score += 80;
+        }
+        if (style.cursor === "pointer" || typeof target.onclick === "function") {
+          score += 40;
+        }
+        if (/(reference|cite|source|web)/i.test(className)) {
+          score += 30;
+        }
+        if (text.length <= 80) {
+          score += 20;
+        }
+        score -= Math.min(rect.width * rect.height / 10_000, 50);
+        return { target, score };
+      })
+      .filter((item): item is { target: HTMLElement; score: number } => Boolean(item));
+
+    const seen = new Set<HTMLElement>();
+    const trigger = candidates
+      .sort((left, right) => right.score - left.score)
+      .map((item) => item.target)
+      .find((target) => {
+        if (seen.has(target)) {
+          return false;
+        }
+        seen.add(target);
+        return true;
+      });
 
     if (!trigger) {
       return false;
     }
 
+    trigger.scrollIntoView({ block: "center", inline: "center" });
     trigger.click();
     return true;
   }, {
@@ -1518,7 +1909,8 @@ async function maybeOpenWenxinReferencePanel(page: PlaywrightPage): Promise<void
   }).catch(() => false);
 
   if (opened) {
-    await page.waitForTimeout(350).catch(() => undefined);
+    await page.waitForTimeout(1_000).catch(() => undefined);
+    await page.waitForLoadState("networkidle", { timeout: 1_500 }).catch(() => undefined);
   }
 }
 
@@ -2153,8 +2545,10 @@ export const wenxinAdapter: MonitorAdapter = {
 };
 
 export const __wenxinTestUtils = {
+  buildSourceItem,
   htmlTableToMarkdown,
   isObservationComplete,
+  normalizeUrl,
   parseWenxinHistoryResponse,
   parseWenxinSSEBody,
 };