diff --git a/admin/dashboard/blueprints/contacts.py b/admin/dashboard/blueprints/contacts.py
index 0cc71e8..19909ed 100644
--- a/admin/dashboard/blueprints/contacts.py
+++ b/admin/dashboard/blueprints/contacts.py
@@ -1,4 +1,5 @@
 import asyncio
+import base64
 import os
 import re
 import threading
@@ -22,6 +23,23 @@ contacts_refresh_lock = threading.Lock()
 contacts_refresh_running = False
 _EMOJI_MD5_RE = re.compile(r'md5\s*=\s*[\"\']([0-9a-fA-F]{16,64})[\"\']', re.IGNORECASE)
 _EMOJI_TOTALLEN_RE = re.compile(r'(?:totallen|total_len|len)\s*=\s*[\"\'](\d+)[\"\']', re.IGNORECASE)
+_EMOJI_BASE64_RE = re.compile(r"^[A-Za-z0-9+/=]+$")
+_EMOJI_LOCALE_KEYS = {"zh_cn", "zh_tw", "zh_hk", "default", "en", "ja", "ko"}
+_EMOJI_SEMANTIC_STOPWORDS = {
+    "default",
+    "zh_cn",
+    "zh_tw",
+    "zh_hk",
+    "en",
+    "ja",
+    "ko",
+    "opus",
+    "gif",
+    "png",
+    "jpg",
+    "jpeg",
+    "webp",
+}
 
 def get_or_create_loop():
     """获取或创建共享的事件循环"""
@@ -203,6 +221,244 @@ def _extract_emoji_meta(attachment_url: str, image_path: str):
     return md5, total_length
 
 
+def _read_protobuf_varint(payload: bytes, offset: int):
+    """读取 protobuf varint。
+
+    说明：
+    1. 微信表情的 desc / emojiattr 经常是 base64 后的 protobuf 片段；
+    2. 这里不依赖 schema，只做最小化的通用 varint 解析，便于递归提取字符串字段；
+    3. 一旦遇到异常字节，直接抛错交给上层兜底，避免误读出脏语义。
+    """
+    result = 0
+    shift = 0
+    index = offset
+    while index < len(payload) and shift <= 63:
+        current = payload[index]
+        index += 1
+        result |= (current & 0x7F) << shift
+        if not (current & 0x80):
+            return result, index
+        shift += 7
+    raise ValueError("protobuf varint 读取失败")
+
+
+def _extract_protobuf_strings(payload: bytes, depth: int = 0):
+    """递归提取 protobuf length-delimited 字段中的 UTF-8 文本。
+
+    说明：
+    1. 这里的目标不是完整反序列化，而是从未知结构中尽量稳定地把“可读文本”捞出来；
+    2. desc 常见格式是 zh_cn/default 语言包嵌套结构，递归 2 层就足够覆盖；
+    3. 如果字段本身是纯文本，递归会自然停掉，不会影响最终结果。
+    """
+    if not payload:
+        return []
+    results = []
+    index = 0
+    while index < len(payload):
+        try:
+            tag, index = _read_protobuf_varint(payload, index)
+        except Exception:
+            break
+        if tag <= 0:
+            break
+
+        wire_type = tag & 0x07
+        if wire_type == 0:
+            try:
+                _, index = _read_protobuf_varint(payload, index)
+            except Exception:
+                break
+            continue
+
+        if wire_type == 1:
+            index += 8
+            continue
+
+        if wire_type == 5:
+            index += 4
+            continue
+
+        if wire_type != 2:
+            break
+
+        try:
+            length, index = _read_protobuf_varint(payload, index)
+        except Exception:
+            break
+        if length < 0 or index + length > len(payload):
+            break
+
+        chunk = payload[index:index + length]
+        index += length
+        if not chunk:
+            continue
+
+        try:
+            decoded = chunk.decode("utf-8")
+        except Exception:
+            decoded = ""
+        if decoded:
+            results.append(decoded)
+
+        if depth < 2:
+            results.extend(_extract_protobuf_strings(chunk, depth + 1))
+    return results
+
+
+def _sanitize_emoji_semantic_text(value: str):
+    """清洗候选语义文本，去掉控制字符和多余空白。"""
+    text = "".join(ch for ch in _safe_text(value) if ch.isprintable()).strip()
+    text = re.sub(r"\s+", " ", text)
+    return text.strip()
+
+
+def _is_emoji_semantic_candidate(value: str):
+    """判断一个候选文本是否像“可读的表情语义”。
+
+    说明：
+    1. 过滤 locale key、文件扩展名、产品 ID 这类元数据；
+    2. 只保留包含中文或英文字母的短文本，避免把长链接、哈希、协议字段误当语义；
+    3. 单字语义也允许保留，例如“害”这类表情实际就有意义。
+    """
+    text = _sanitize_emoji_semantic_text(value)
+    if not text:
+        return False
+    lowered = text.lower()
+    if lowered in _EMOJI_LOCALE_KEYS or lowered in _EMOJI_SEMANTIC_STOPWORDS:
+        return False
+    if any(locale_key in lowered for locale_key in _EMOJI_LOCALE_KEYS):
+        return False
+    if lowered.startswith("com.tencent.") or lowered.startswith("finder:"):
+        return False
+    if re.fullmatch(r"[0-9a-f]{16,64}", lowered):
+        return False
+    if len(text) >= 8 and _EMOJI_BASE64_RE.fullmatch(text):
+        return False
+    if len(text) > 40:
+        return False
+    return bool(re.search(r"[\u4e00-\u9fffA-Za-z]", text))
+
+
+def _dedupe_emoji_semantic_candidates(values):
+    """按出现顺序去重候选语义文本。"""
+    seen = set()
+    results = []
+    for item in values or []:
+        text = _sanitize_emoji_semantic_text(item)
+        if not _is_emoji_semantic_candidate(text):
+            continue
+        key = text.lower()
+        if key in seen:
+            continue
+        seen.add(key)
+        results.append(text)
+    return results
+
+
+def _maybe_decode_base64_payload(value: str):
+    """尽量把字段值解成 base64 原始字节，失败时返回空字节。
+
+    说明：
+    1. 微信的 desc / emojiattr 并不总是明文，有不少是 base64 包起来的 protobuf；
+    2. 这里先做格式筛选，避免把普通中文直接当 base64 解坏；
+    3. 允许缺省 padding，兼容历史数据里的非标准尾部。
+    """
+    normalized = re.sub(r"\s+", "", _safe_text(value))
+    if len(normalized) < 4 or not _EMOJI_BASE64_RE.fullmatch(normalized):
+        return b""
+    normalized += "=" * (-len(normalized) % 4)
+    try:
+        return base64.b64decode(normalized, validate=False)
+    except Exception:
+        return b""
+
+
+def _decode_emoji_semantic_value(value: str):
+    """解析单个表情语义字段，输出候选语义文本列表。
+
+    说明：
+    1. 若字段本身就是明文中文，直接保留；
+    2. 若字段是 base64，则先尝试整段 UTF-8，再递归提取 protobuf 内嵌字符串；
+    3. 最终统一做去重和脏值过滤，避免把 locale key 一起带回前端。
+    """
+    raw_text = _safe_text(value).strip()
+    if not raw_text:
+        return []
+
+    candidates = []
+    if _is_emoji_semantic_candidate(raw_text):
+        candidates.append(raw_text)
+
+    decoded_bytes = _maybe_decode_base64_payload(raw_text)
+    if not decoded_bytes:
+        return _dedupe_emoji_semantic_candidates(candidates)
+
+    protobuf_texts = _extract_protobuf_strings(decoded_bytes)
+    candidates.extend(protobuf_texts)
+
+    # 某些 emojiattr 不是 protobuf，而是“base64 后的纯文本”。
+    # 只有在 protobuf 路径没抽到结果时，才退回整段 UTF-8 解码，避免把外层语言包拼接串带进来。
+    if not _dedupe_emoji_semantic_candidates(candidates):
+        try:
+            decoded_text = decoded_bytes.decode("utf-8")
+        except Exception:
+            decoded_text = ""
+        if decoded_text:
+            candidates.append(decoded_text)
+    return _dedupe_emoji_semantic_candidates(candidates)
+
+
+def _extract_emoji_semantic_info(attachment_url: str):
+    """从表情 XML 中提取“可读语义”。
+
+    说明：
+    1. 当前表情库主要只有 md5/len，不方便后续让 AI 直接利用；
+    2. 这里优先解析 desc、attachedtext、emojiattr 这些潜在语义字段；
+    3. 返回主语义 + 别名列表 + 来源，后续无论是后台展示还是自动回复匹配都能复用。
+    """
+    text = _safe_text(attachment_url).strip()
+    if not text.startswith("<"):
+        return {
+            "semantic_text": "",
+            "semantic_aliases": [],
+            "semantic_source": "",
+        }
+
+    field_values = []
+    try:
+        root = ET.fromstring(text)
+        emoji_node = root.find(".//emoji")
+        if emoji_node is not None:
+            for field_name in ("desc", "attachedtext", "emojiattr"):
+                field_values.append((field_name, _safe_text(emoji_node.attrib.get(field_name, "")).strip()))
+    except Exception:
+        for field_name in ("desc", "attachedtext", "emojiattr"):
+            match = re.search(rf'{field_name}\s*=\s*[\"\']([^\"\']+)[\"\']', text, re.IGNORECASE)
+            field_values.append((field_name, _safe_text(match.group(1) if match else "").strip()))
+
+    aliases = []
+    sources = []
+    for field_name, field_value in field_values:
+        decoded_candidates = _decode_emoji_semantic_value(field_value)
+        if not decoded_candidates:
+            continue
+        sources.append(field_name)
+        aliases.extend(decoded_candidates)
+
+    semantic_aliases = _dedupe_emoji_semantic_candidates(aliases)
+    semantic_text = ""
+    if semantic_aliases:
+        # 优先选中文最明显的候选，尽量把“哈哈哈”“害”这类直观语义放到第一位。
+        chinese_first = [item for item in semantic_aliases if re.search(r"[\u4e00-\u9fff]", item)]
+        semantic_text = chinese_first[0] if chinese_first else semantic_aliases[0]
+
+    return {
+        "semantic_text": semantic_text,
+        "semantic_aliases": semantic_aliases,
+        "semantic_source": ",".join(sources),
+    }
+
+
 def _parse_positive_int(value):
     """将任意输入尽量解析为正整数，失败时返回 0。
 
@@ -781,24 +1037,47 @@ def api_emoji_library():
 
         dedup = {}
         for item in records:
+            attachment_url = _safe_text(item.get("attachment_url"))
             image_path = _safe_text(item.get("image_path")).strip()
-            if not image_path:
-                continue
-            md5, total_length = _extract_emoji_meta(_safe_text(item.get("attachment_url")), image_path)
+            md5, total_length = _extract_emoji_meta(attachment_url, image_path)
             if not md5 or total_length <= 0:
                 continue
-            if md5 in dedup:
-                continue
-            dedup[md5] = {
+            semantic_info = _extract_emoji_semantic_info(attachment_url)
+
+            # 同一个 md5 可能在多条历史里反复出现：
+            # 1. 有的记录有预览图但没有语义；
+            # 2. 有的记录有语义但图片还没落盘；
+            # 3. 因此这里按 md5 聚合，尽量把“发送参数 + 预览图 + 中文语义”拼成一条完整资产。
+            target = dedup.setdefault(md5, {
                 "md5": md5,
                 "total_length": total_length,
-                "preview_url": image_path,
+                "preview_url": "",
                 "timestamp": _safe_text(item.get("timestamp")),
                 "group_id": _safe_text(item.get("group_id")),
                 "message_id": _safe_text(item.get("message_id")),
-            }
+                "semantic_text": "",
+                "semantic_aliases": [],
+                "semantic_source": "",
+            })
 
-        emojis = list(dedup.values())
+            if not target.get("preview_url") and image_path:
+                target["preview_url"] = image_path
+            if not target.get("total_length") and total_length > 0:
+                target["total_length"] = total_length
+
+            target_aliases = target.get("semantic_aliases") or []
+            merged_aliases = _dedupe_emoji_semantic_candidates(target_aliases + (semantic_info.get("semantic_aliases") or []))
+            target["semantic_aliases"] = merged_aliases
+            if not target.get("semantic_text") and semantic_info.get("semantic_text"):
+                target["semantic_text"] = semantic_info.get("semantic_text")
+            if not target.get("semantic_source") and semantic_info.get("semantic_source"):
+                target["semantic_source"] = semantic_info.get("semantic_source")
+
+        # 只有带预览图的表情才回给前端弹窗：
+        # 1. 目前弹窗主要承担“人工挑选并发送”的作用，没有缩略图会很难用；
+        # 2. 语义可以从其他重复记录补过来，但最终仍要求至少有一条落盘图片；
+        # 3. 后续若要纯语义离线匹配，可再单独开放无预览的内部接口。
+        emojis = [item for item in dedup.values() if item.get("preview_url")]
         return jsonify({
             "success": True,
             "data": {
diff --git a/admin/dashboard/templates/contacts_management.html b/admin/dashboard/templates/contacts_management.html
index 63b0196..8c4b8a1 100644
--- a/admin/dashboard/templates/contacts_management.html
+++ b/admin/dashboard/templates/contacts_management.html
@@ -789,7 +789,7 @@
 
     <el-dialog title="表情库" :visible.sync="emojiDialogVisible" width="52%">
         <div class="emoji-toolbar">
-            <el-input v-model="emojiKeyword" clearable placeholder="搜索 md5..." size="small"></el-input>
+            <el-input v-model="emojiKeyword" clearable placeholder="搜索 md5 / 中文语义..." size="small"></el-input>
             <el-button size="small" icon="el-icon-refresh" :loading="emojiLibraryLoading" @click="loadEmojiLibrary">刷新</el-button>
         </div>
         <div class="emoji-grid" v-loading="emojiLibraryLoading">
@@ -797,6 +797,8 @@
             <div v-if="!filteredEmojiLibrary.length" class="emoji-empty">暂无可用表情，先在群里让媒体下载插件抓取几条表情。</div>
             <div v-for="item in filteredEmojiLibrary" :key="item.md5" class="emoji-card">
                 <img class="emoji-thumb" :src="getChatMediaUrl(item.preview_url)" />
+                <div v-if="item.semantic_text" class="emoji-semantic">{{ item.semantic_text }}</div>
+                <div v-if="item.semantic_aliases && item.semantic_aliases.length > 1" class="emoji-aliases">{{ item.semantic_aliases.join(' / ') }}</div>
                 <div class="emoji-md5">{{ item.md5 }}</div>
                 <div class="emoji-actions">
                     <el-button type="primary" size="mini" @click="sendEmojiItem(item)">发送</el-button>
@@ -905,7 +907,14 @@
             filteredEmojiLibrary() {
                 const keyword = (this.emojiKeyword || '').trim().toLowerCase();
                 if (!keyword) return this.emojiLibrary;
-                return this.emojiLibrary.filter(item => (item.md5 || '').toLowerCase().includes(keyword));
+                return this.emojiLibrary.filter(item => {
+                    // 表情库后续要服务 AI 自动回复，因此这里除了 md5，也支持按主语义和别名检索。
+                    // 这样人工整理映射时，可以直接搜“哈哈/害/难道”之类语义词，不需要反复记 md5。
+                    const md5 = (item.md5 || '').toLowerCase();
+                    const semanticText = (item.semantic_text || '').toLowerCase();
+                    const aliases = Array.isArray(item.semantic_aliases) ? item.semantic_aliases.join(' ').toLowerCase() : '';
+                    return md5.includes(keyword) || semanticText.includes(keyword) || aliases.includes(keyword);
+                });
             },
             previewGroupWelcomeText() {
                 return this.renderWelcomeTemplate(this.groupWelcomeConfig.welcome_text_template);
@@ -1856,6 +1865,8 @@
     }
     .preview-box p{margin:0 0 4px 0}
     .emoji-thumb { width: 72px; height: 72px; object-fit: contain; border-radius: 8px; background: rgba(148,163,184,0.08); }
+    .emoji-semantic { font-size: 13px; font-weight: 600; color: #0f172a; text-align: center; min-height: 18px; }
+    .emoji-aliases { font-size: 11px; color: #475569; text-align: center; line-height: 1.45; max-width: 100%; word-break: break-word; min-height: 16px; }
     .emoji-md5 { font-size: 11px; color: #64748b; word-break: break-all; text-align: center; min-height: 30px; }
     .emoji-actions { width: 100%; display: flex; justify-content: center; }
     .emoji-empty { color: #94a3b8; padding: 12px; }