feat(token): Add consumed mode for quota management (opt-in)

_public/static/admin/js/config.js

@@ -150,7 +150,8 @@ const LOCALE_MAP = {
     "fail_threshold": { title: "失败阈值", desc: "单个 Token 连续失败多少次后被标记为不可用。" },
     "save_delay_ms": { title: "保存延迟", desc: "Token 变更合并写入的延迟（毫秒）。" },
     "usage_flush_interval_sec": { title: "用量落库间隔", desc: "用量类字段写入数据库的最小间隔（秒）。" },
-    "reload_interval_sec": { title: "同步间隔", desc: "多 worker 场景下 Token 状态刷新间隔（秒）。" }
+    "reload_interval_sec": { title: "同步间隔", desc: "多 worker 场景下 Token 状态刷新间隔（秒）。" },
+    "consumed_mode_enabled": { title: "启用消耗模式", desc: "启用新额度管理逻辑：使用本地消耗记录而非 API 返回值，支持更均衡的负载分配。（试验性功能，默认关闭）" }
   },
 
 

_public/static/admin/js/token.js

@@ -1,4 +1,5 @@
 let apiKey = '';
+let consumedModeEnabled = false;
 let allTokens = {};
 let flatTokens = [];
 let isBatchProcessing = false;
@@ -123,9 +124,11 @@ async function loadData() {
     });
     if (res.ok) {
       const data = await res.json();
-      allTokens = data;
-      processTokens(data);
-      updateStats(data);
+      allTokens = data.tokens;
+      consumedModeEnabled = data.consumed_mode_enabled || false;
+      updateQuotaHeader();
+      processTokens(data.tokens);
+      updateStats(data.tokens);
       renderTable();
     } else if (res.status === 401) {
       logout();
@@ -151,6 +154,7 @@ function processTokens(data) {
             token: t.token,
             status: t.status || 'active',
             quota: t.quota || 0,
+            consumed: t.consumed || 0,
             note: t.note || '',
             fail_count: t.fail_count || 0,
             use_count: t.use_count || 0,
@@ -168,6 +172,19 @@ function processTokens(data) {
   });
 }
 
+function updateQuotaHeader() {
+  const thQuota = document.getElementById('th-quota');
+  if (thQuota) {
+    if (consumedModeEnabled) {
+      thQuota.textContent = t('token.tableQuotaConsumed');
+      thQuota.dataset.i18n = 'token.tableQuotaConsumed';
+    } else {
+      thQuota.textContent = t('token.tableQuota');
+      thQuota.dataset.i18n = 'token.tableQuota';
+    }
+  }
+}
+
 function updateStats(data) {
   // Logic same as before, simplified reuse if possible, but let's re-run on flatTokens
   let totalTokens = flatTokens.length;
@@ -197,14 +214,27 @@ function updateStats(data) {
   });
 
   const imageQuota = Math.floor(chatQuota / 2);
+  const totalConsumed = flatTokens.reduce((sum, t) => sum + (t.consumed || 0), 0);
 
+  // 更新统计卡片 (这些不受 consumedMode 影响)
   setText('stat-total', totalTokens.toLocaleString());
   setText('stat-active', activeTokens.toLocaleString());
   setText('stat-cooling', coolingTokens.toLocaleString());
   setText('stat-invalid', invalidTokens.toLocaleString());
 
-  setText('stat-chat-quota', chatQuota.toLocaleString());
-  setText('stat-image-quota', imageQuota.toLocaleString());
+  // 根据配置决定显示消耗还是剩余
+  if (consumedModeEnabled) {
+    setText('stat-chat-quota', totalConsumed.toLocaleString());
+    setText('stat-image-quota', Math.floor(totalConsumed / 2).toLocaleString());
+    const chatLabel = document.querySelector('[data-i18n="token.statChatQuota"]');
+    const imageLabel = document.querySelector('[data-i18n="token.statImageQuota"]');
+    if (chatLabel) chatLabel.textContent = t('token.statChatConsumed');
+    if (imageLabel) imageLabel.textContent = t('token.statImageConsumed');
+  } else {
+    setText('stat-chat-quota', chatQuota.toLocaleString());
+    setText('stat-image-quota', imageQuota.toLocaleString());
+  }
+
   setText('stat-total-calls', totalCalls.toLocaleString());
 
   updateTabCounts({
@@ -293,7 +323,16 @@ function renderTable() {
     // Quota (Center)
     const tdQuota = document.createElement('td');
     tdQuota.className = 'text-center font-mono text-xs';
-    tdQuota.innerText = item.quota;
+    // 根据配置决定显示消耗还是剩余
+    if (consumedModeEnabled) {
+      tdQuota.innerText = item.consumed;
+      tdQuota.title = t('token.tableQuotaConsumed');
+    } else {
+      tdQuota.innerText = item.quota;
+      tdQuota.title = t('token.tableQuota');
+    }
+
+
 
     // Note (Left)
     const tdNote = document.createElement('td');
@@ -503,6 +542,23 @@ function openEditModal(index) {
     byId('edit-pool').value = item.pool;
     byId('edit-quota').value = item.quota;
     byId('edit-note').value = item.note;
+
+    // 根据配置决定是否禁用 quota 编辑
+    const quotaInput = byId('edit-quota');
+    const quotaInputParent = quotaInput?.closest('div');
+    const quotaLabel = quotaInputParent?.previousElementSibling;
+    if (consumedModeEnabled) {
+      quotaInput.disabled = true;
+      quotaInput.classList.add('bg-gray-100', 'text-gray-400');
+      if (quotaLabel) quotaLabel.textContent = t('token.tableQuotaConsumed');
+    } else {
+      quotaInput.disabled = false;
+      quotaInput.classList.remove('bg-gray-100', 'text-gray-400');
+      if (quotaLabel) quotaLabel.textContent = t('token.editQuota');
+    }
+
+    document.querySelector('#edit-modal h3').innerText = t('token.editTitle');
+    byId('edit-note').value = item.note;
     document.querySelector('#edit-modal h3').innerText = t('token.editTitle');
   } else {
     // New Token
@@ -518,6 +574,14 @@ function openEditModal(index) {
     byId('edit-quota').value = getDefaultQuotaForPool('ssoBasic');
     byId('edit-note').value = '';
     document.querySelector('#edit-modal h3').innerText = t('token.addTitle');
+
+    // 新建 Token 时启用 quota 编辑
+    const newQuotaInput = byId('edit-quota');
+    const newQuotaInputParent = newQuotaInput?.closest('div');
+    const newQuotaLabel = newQuotaInputParent?.previousElementSibling;
+    newQuotaInput.disabled = false;
+    newQuotaInput.classList.remove('bg-gray-100', 'text-gray-400');
+    if (newQuotaLabel) newQuotaLabel.textContent = t('token.editQuota');
   }
 
   openModal('edit-modal');

_public/static/admin/pages/token.html

@@ -167,7 +167,7 @@ <h2 class="text-2xl font-semibold tracking-tight" data-i18n="token.title">Token
                 <th class="w-56 text-left" data-i18n="token.tableToken">Token</th>
                 <th class="w-24" data-i18n="token.tableType">类型</th>
                 <th class="w-24" data-i18n="token.tableStatus">状态</th>
-                <th class="w-20" data-i18n="token.tableQuota">额度</th>
+                <th class="w-20" id="th-quota" data-i18n="token.tableQuota">额度</th>
                 <th class="text-left" data-i18n="token.tableNote">备注</th>
                 <th class="w-44 text-center" data-i18n="token.tableActions">操作</th>
               </tr>

_public/static/i18n/locales/en.json

@@ -244,6 +244,8 @@
     "statInvalid": "Invalid",
     "statChatQuota": "Chat Remaining",
     "statImageQuota": "Image Remaining",
+    "statChatConsumed": "Chat Consumed",
+    "statImageConsumed": "Image Consumed",
     "statVideoQuota": "Video Remaining",
     "statVideoUnavailable": "N/A",
     "statTotalCalls": "Total Calls",
@@ -258,6 +260,7 @@
     "tableType": "Type",
     "tableStatus": "Status",
     "tableQuota": "Quota",
+    "tableQuotaConsumed": "Consumed",
     "tableNote": "Note",
     "tableActions": "Actions",
     "refreshStatus": "Refresh status",

_public/static/i18n/locales/zh.json

@@ -244,6 +244,8 @@
     "statInvalid": "Token 失效",
     "statChatQuota": "Chat 剩余",
     "statImageQuota": "Image 剩余",
+    "statChatConsumed": "Chat 已消耗",
+    "statImageConsumed": "Image 已消耗",
     "statVideoQuota": "Video 剩余",
     "statVideoUnavailable": "无法统计",
     "statTotalCalls": "总调用次数",
@@ -258,6 +260,7 @@
     "tableType": "类型",
     "tableStatus": "状态",
     "tableQuota": "额度",
+    "tableQuotaConsumed": "已消耗",
     "tableNote": "备注",
     "tableActions": "操作",
     "refreshStatus": "刷新状态",

app/api/v1/admin/token.py

@@ -48,7 +48,14 @@ async def get_tokens():
     """获取所有 Token"""
     storage = get_storage()
     tokens = await storage.load_tokens()
-    return tokens or {}
+    # 获取消耗模式配置
+    from app.core.config import get_config
+    consumed_mode = get_config("token.consumed_mode_enabled", False)
+
+    return {
+        "tokens": tokens or {},
+        "consumed_mode_enabled": consumed_mode
+    }
 
 
 @router.post("/tokens", dependencies=[Depends(verify_app_key)])

app/services/token/manager.py

@@ -471,7 +471,18 @@ async def consume(
             token = pool.get(raw_token)
             if token:
                 old_status = token.status
-                consumed = token.consume(effort)
+                # 检查是否启用消耗模式
+                consumed_mode = False
+                try:
+                    from app.core.config import get_config
+                    consumed_mode = get_config("token.consumed_mode_enabled", False)
+                except Exception:
+                    pass
+
+                if consumed_mode:
+                    consumed = token.consume_with_consumed(effort)
+                else:
+                    consumed = token.consume(effort)
                 logger.debug(
                     f"Token {raw_token[:10]}...: consumed {consumed} quota, use_count={token.use_count}"
                 )
@@ -662,6 +673,7 @@ async def mark_rate_limited(self, token_str: str) -> bool:
                 old_quota = token.quota
                 token.quota = 0
                 token.status = TokenStatus.COOLING
+                token.consumed = 0  # 进入冷却时重置本轮消耗
                 logger.warning(
                     f"Token {raw_token[:10]}...: marked as rate limited "
                     f"(quota {old_quota} -> 0, status -> cooling)"
@@ -936,6 +948,25 @@ async def _refresh_one(item: tuple[str, TokenInfo]) -> dict:
                     old_status = token_info.status
 
                     token_info.update_quota(new_quota)
+
+                    # 检查是否启用 consumed 模式
+                    consumed_mode = False
+                    try:
+                        from app.core.config import get_config
+                        consumed_mode = get_config("token.consumed_mode_enabled", False)
+                    except Exception:
+                        pass
+
+                    if consumed_mode:
+                        # Consumed 模式：使用新逻辑
+                        token_info.update_quota_with_consumed(new_quota)
+                    else:
+                        # 默认模式：使用旧逻辑
+                        token_info.update_quota(new_quota)
+
+                    # 刷新成功后如果 quota > 0，恢复活跃状态
+                    if new_quota > 0:
+                        token_info.status = TokenStatus.ACTIVE
                     token_info.mark_synced()
 
                     window_size = self._extract_window_size_seconds(result)