feat: multi-channel LLM support with visual channel editor (#494)

* feat: multi-channel LLM support with visual channel editor

- Add three-tier LLM config: LITELLM_CONFIG (YAML) > LLM_CHANNELS (env) > legacy keys
- Each channel gets independent base_url / api_key / models (no OPENAI_BASE_URL conflict)
- Support DEEPSEEK_API_KEY as standalone provider (auto-infers deepseek-chat model)
- Add LLMChannelEditor component with 9 presets (AIHubmix/DeepSeek/Dashscope/GLM/Moonshot/SiliconFlow/OpenRouter/Gemini/Custom)
- Expand config_registry with ~50 new fields for web settings coverage
- Rewrite .env.example AI section with clear quick-start guide (Scenario A vs B)
- Add litellm_config.example.yaml template
- Add PyYAML dependency for YAML config support
- Full backward compatibility: existing single-key configs work unchanged

* chore: replace placeholder key values with empty defaults in .env.example

* fix: resolve ESLint errors in LLMChannelEditor and HomePage

* refactor: use native litellm deepseek/ provider and extract shared LLM helpers

- Replace openai/deepseek-* + manual api_base with deepseek/ prefix (litellm
  natively resolves DEEPSEEK_API_KEY and base_url)
- Extract duplicated _get_api_keys_for_model and _extra_litellm_params from
  analyzer.py and llm_adapter.py into shared functions in config.py
- Update litellm_config.example.yaml to use deepseek/ prefix
- Thinking mode (deepseek-chat opt-in, deepseek-reasoner auto) unaffected:
  get_thinking_extra_body uses model short name stripped of provider prefix

.env.example

@@ -10,7 +10,7 @@ STOCK_LIST=600519,300750,002594
 
 # 数据源配置
 # Tushare Pro Token（可选，从 https://tushare.pro/weborder/#/login?reg=834638 获取）
-TUSHARE_TOKEN=your_tushare_token_here
+TUSHARE_TOKEN=
 
 # ===================================
 # 定时任务配置（本地/Docker运行）
@@ -23,68 +23,92 @@ TUSHARE_TOKEN=your_tushare_token_here
 # TRADING_DAY_CHECK_ENABLED=true
 
 # ===================================
-# AI 模型配置（统一通过 LiteLLM，至少配置一个 API Key）
+# AI 模型配置
+# ===================================
 #
-# 核心配置（二选一或组合）：
-#   LITELLM_MODEL         - 主模型，仅填一个，格式 provider/model-name；使用第三方模型提供商或 OpenAI 兼容 API 时须加 openai 前缀
-#   LITELLM_FALLBACK_MODELS - 备选模型，逗号分隔，主模型全部失败时按序尝试
-# 若未配置 LITELLM_MODEL，系统将根据已有 API Key 自动推断（推断结果打印在日志中）
+# 【快速上手 — 根据使用场景选一种即可】
 #
-# 模型格式示例：gemini/gemini-2.5-flash、anthropic/claude-3-5-sonnet-20241022、openai/gpt-4o
+#   场景 A：只用一个模型（最简单）
+#     → 填对应 API Key 即可，系统自动识别模型。
+#       例：只用 Gemini → 填 GEMINI_API_KEY
+#       例：只用 DeepSeek → 填 DEEPSEEK_API_KEY
+#       例：想用 AIHubmix 聚合 → 填 AIHUBMIX_KEY
+#
+#   场景 B：同时使用多个模型/平台（推荐渠道模式）
+#     → 配置 LLM_CHANNELS，每个渠道独立填 base_url / api_key / models。
+#       也可在 Web 设置页 → AI 模型 → 渠道编辑器中可视化配置。
+#       详见下方「多渠道配置」区域。
+#
+# ⚠️ 两种方式不要混用：配了渠道后，传统 API Key 区域的配置会被忽略。
+#
+# 高级选项（通常无需手动设置，可自动推断）：
+#   LITELLM_MODEL           主模型，格式 provider/model-name
+#   LITELLM_FALLBACK_MODELS 备选模型，逗号分隔
 # ===================================
 
 # LITELLM_MODEL=gemini/gemini-3-flash-preview
 # LITELLM_FALLBACK_MODELS=anthropic/claude-3-5-sonnet-20241022,openai/gpt-4o-mini
 
-# 温度参数
+# -----------------------------------
+# API Key 配置（场景 A：只用一个模型）
+# -----------------------------------
+
+# 【Gemini】免费额度（https://aistudio.google.com）
+GEMINI_API_KEY=
+# 多 Key 负载均衡：GEMINI_API_KEYS=key1,key2,key3
 GEMINI_TEMPERATURE=0.7
 
-# ===================================
-# API Key 配置（支持多个 Key，逗号分隔，自动负载均衡）
-# ===================================
+# 【DeepSeek】https://platform.deepseek.com
+# DEEPSEEK_API_KEY=sk-xxxxxxxxxxxxxxxx
 
-# 【方案一】Gemini API（有免费额度，从 https://aistudio.google.com 获取）
-# 单 Key：
-GEMINI_API_KEY=
-# 多 Key：GEMINI_API_KEYS=key1,key2,key3
-# 示例：LITELLM_MODEL=gemini/gemini-3-flash-preview
+# 【推荐】AIHubmix 聚合（https://aihubmix.com/?aff=CfMq）
+# 一个 Key 用 GPT/Claude/Gemini/GLM/Qwen 等模型，无需科学上网
+# AIHUBMIX_KEY=
 
-# 【方案二】Anthropic Claude API（从 https://console.anthropic.com 获取）
+# 【Anthropic Claude】https://console.anthropic.com
 # ANTHROPIC_API_KEY=sk-ant-xxxxxxxxxxxxxxxx
-# 多 Key：ANTHROPIC_API_KEYS=key1,key2
-# 示例：LITELLM_MODEL=anthropic/claude-3-5-sonnet-20241022
 # ANTHROPIC_TEMPERATURE=0.7
 
-# 【方案三】OpenAI / DeepSeek / OpenRouter 等兼容 API
-# 单 Key：OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx
-# 多 Key：OPENAI_API_KEYS=key1,key2
-# OPENAI_BASE_URL=第三方 API 地址（官方 OpenAI 可不填；DeepSeek: https://api.deepseek.com/v1；OpenRouter: https://openrouter.ai/api/v1）
-# OPENAI_TEMPERATURE=0.7
-# OPENAI_VISION_MODEL=gpt-4o   # 图片识别专用模型（可选）
-#
-# --- 【推荐】AIHubmix 一站式 ---
-# AIHubmix 支持一站式使用全球主流 AI 模型，一个 Key 可在本项目中切换使用任何模型，
-# 无需科学上网，含免费模型（glm-5、gpt-4o-free 等顶级模型），
-# 付费模型拥有极高的稳定性和无限并发能力，适合大规模生产级应用使用。
-# 使用 AIHUBMIX_KEY 时无需配置 OPENAI_BASE_URL，系统自动使用 aihubmix.com/v1。
-# 获取 Key：https://aihubmix.com/?aff=CfMq
-# AIHUBMIX_KEY=your_aihubmix_key_here
-# 示例：LITELLM_MODEL=openai/gemini-3.1-pro-preview
-#
-# --- DeepSeek ---
+# 【OpenAI 兼容】适用于 OpenAI / 任意兼容 API
 # OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx
-# OPENAI_BASE_URL=https://api.deepseek.com/v1
-# 示例：LITELLM_MODEL=openai/deepseek-chat 或 openai/deepseek-reasoner
-# 思考模式：deepseek-reasoner/deepseek-r1/qwq 自动识别，deepseek-chat 需 extra_body 启用
+# OPENAI_BASE_URL=（官方可不填；第三方填对应地址）
+# OPENAI_TEMPERATURE=0.7
+# OPENAI_VISION_MODEL=gpt-4o
+
+# 【其他】https://docs.litellm.ai/docs/providers
+# 设好环境变量 + provider/model 格式即可（如 COHERE_API_KEY + cohere/command-r-plus）
+
+# -----------------------------------
+# 多渠道配置（场景 B：同时使用多个模型/平台）
+# -----------------------------------
+# 每个渠道独立配置 base_url / api_key / models，互不冲突。
+# 任何 OpenAI 兼容 API（DeepSeek、Qwen、GLM、Moonshot 等）都可以直接作为渠道添加。
+#
+# 示例：AIHubmix + DeepSeek + Gemini 三渠道共存
+# LLM_CHANNELS=aihubmix,deepseek,gemini
+#
+# LLM_AIHUBMIX_BASE_URL=https://aihubmix.com/v1
+# LLM_AIHUBMIX_API_KEY=
+# LLM_AIHUBMIX_MODELS=gpt-4o-mini,claude-3-5-sonnet,qwen-plus
+#
+# LLM_DEEPSEEK_BASE_URL=https://api.deepseek.com/v1
+# LLM_DEEPSEEK_API_KEY=sk-xxxxxxxxxxxxxxxx
+# LLM_DEEPSEEK_MODELS=deepseek-chat,deepseek-reasoner
+#
+# LLM_GEMINI_API_KEYS=key1,key2,key3
+# LLM_GEMINI_MODELS=gemini/gemini-2.5-flash
+#
+# 高级：YAML 配置（可选，标准 LiteLLM 格式，参考 litellm_config.example.yaml）
+# LITELLM_CONFIG=./litellm_config.yaml
 
 # 搜索引擎配置（用于获取股票新闻）
 # Tavily API Keys（支持多个，逗号分隔）
-TAVILY_API_KEYS=your_tavily_key_here
+TAVILY_API_KEYS=
 # SerpAPI Keys（支持多个，逗号分隔）
-SERPAPI_API_KEYS=your_serpapi_key_here
+SERPAPI_API_KEYS=
 # Brave Search API Keys（支持多个，逗号分隔）
 # 获取: https://brave.com/search/api/
-BRAVE_API_KEYS=your_brave_key_here
+BRAVE_API_KEYS=
 
 # ===================================
 # 新闻时效与分析筛选配置
@@ -154,8 +178,8 @@ AGENT_SKILLS=bull_trend,ma_golden_cross,volume_breakout,shrink_pullback
 # 1. 获取授权码（以QQ邮箱为例）：设置 -> 账户 -> POP3/SMTP服务 -> 开启 -> 获取授权码
 # 2. 填写下面两项即可：
 #
-# EMAIL_SENDER=your_email@qq.com
-# EMAIL_PASSWORD=your_email_auth_code
+# EMAIL_SENDER=
+# EMAIL_PASSWORD=
 # EMAIL_RECEIVERS=receiver@example.com  # 可选，留空则发给自己
 #
 # 【方式四扩展】股票分组发往不同邮箱（Issue #268，可选）
@@ -170,19 +194,19 @@ AGENT_SKILLS=bull_trend,ma_golden_cross,volume_breakout,shrink_pullback
 # 系统会自动识别常见服务并使用对应格式
 #
 # CUSTOM_WEBHOOK_URLS=https://oapi.dingtalk.com/robot/send?access_token=xxx,https://hooks.slack.com/services/xxx
-# CUSTOM_WEBHOOK_BEARER_TOKEN=your_bearer_token  # 可选，用于需要认证的 Webhook (Header Authorization: Bearer <token>)
+# CUSTOM_WEBHOOK_BEARER_TOKEN=  # 可选，用于需要认证的 Webhook (Header Authorization: Bearer <token>)
 # WEBHOOK_VERIFY_SSL=true   # 默认校验。设为 false 可支持自签名证书。警告：禁用后存在 MITM 劫持风险，仅限可信内网
 #
 # 【方式六】Pushover 配置
 # 注册Pushover账号，并创建应用Token https://pushover.net/apps/build
-# PUSHOVER_USER_KEY=your_user_key
-# PUSHOVER_API_TOKEN=your_api_token
+# PUSHOVER_USER_KEY=
+# PUSHOVER_API_TOKEN=
 #
 # 【方式七】PushPlus 配置（国内推送服务，推荐）
 # 注册PushPlus账号并获取Token https://www.pushplus.plus
-# PUSHPLUS_TOKEN=your_pushplus_token
+# PUSHPLUS_TOKEN=
 # 群组推送：填写群组编码后，消息推送给群组所有订阅用户（一对多）
-# PUSHPLUS_TOPIC=your_group_topic_code
+# PUSHPLUS_TOPIC=
 #
 # 【方式八】Discord 配置
 # 支持两种方式：Webhook（推荐，配置简单）和 Bot API（权限高）
@@ -195,12 +219,12 @@ AGENT_SKILLS=bull_trend,ma_golden_cross,volume_breakout,shrink_pullback
 # 1. 创建 Bot：https://discord.com/developers/applications -> 新建应用 -> Bot -> 创建 Bot
 # 2. 获取 Bot Token：Bot 页面 -> 重置 Token
 # 3. 获取频道 ID：Discord 开启开发者模式 -> 右键频道 -> 复制 ID
-# DISCORD_BOT_TOKEN=your_bot_token_here
-# DISCORD_MAIN_CHANNEL_ID=your_channel_id_here
+# DISCORD_BOT_TOKEN=
+# DISCORD_MAIN_CHANNEL_ID=
 # 
 # 【方式九】Server酱3 配置（国内推送服务，支持微信推送）
 # 注册Server酱3账号并获取SendKey https://sc3.ft07.com/
-# SERVERCHAN3_SENDKEY=your_serverchan3_sendkey
+# SERVERCHAN3_SENDKEY=
 #
 # 【高级配置】消息长度限制（字节）
 # 超过限制会自动分批发送，一般无需修改

apps/dsa-web/src/components/settings/LLMChannelEditor.tsx

@@ -0,0 +1,399 @@
+import { useState, useMemo, useCallback } from 'react';
+import type React from 'react';
+import { EyeToggleIcon } from '../common';
+import { systemConfigApi } from '../../api/systemConfig';
+
+/** Well-known channel presets for quick-add dropdown. */
+const CHANNEL_PRESETS: Record<string, { label: string; baseUrl: string; placeholder: string }> = {
+  aihubmix: {
+    label: 'AIHubmix（聚合平台）',
+    baseUrl: 'https://aihubmix.com/v1',
+    placeholder: 'gpt-4o-mini,claude-3-5-sonnet,qwen-plus',
+  },
+  deepseek: {
+    label: 'DeepSeek 官方',
+    baseUrl: 'https://api.deepseek.com/v1',
+    placeholder: 'deepseek-chat,deepseek-reasoner',
+  },
+  dashscope: {
+    label: '通义千问（Dashscope）',
+    baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1',
+    placeholder: 'qwen-plus,qwen-turbo',
+  },
+  zhipu: {
+    label: '智谱 GLM',
+    baseUrl: 'https://open.bigmodel.cn/api/paas/v4',
+    placeholder: 'glm-4-flash,glm-4-plus',
+  },
+  moonshot: {
+    label: 'Moonshot（月之暗面）',
+    baseUrl: 'https://api.moonshot.cn/v1',
+    placeholder: 'moonshot-v1-8k',
+  },
+  siliconflow: {
+    label: '硅基流动（SiliconFlow）',
+    baseUrl: 'https://api.siliconflow.cn/v1',
+    placeholder: 'deepseek-ai/DeepSeek-V3',
+  },
+  openrouter: {
+    label: 'OpenRouter',
+    baseUrl: 'https://openrouter.ai/api/v1',
+    placeholder: 'gpt-4o,claude-3.5-sonnet',
+  },
+  gemini: {
+    label: 'Gemini（原生，无需 base_url）',
+    baseUrl: '',
+    placeholder: 'gemini/gemini-2.5-flash',
+  },
+  custom: {
+    label: '自定义渠道',
+    baseUrl: '',
+    placeholder: 'model-name-1,model-name-2',
+  },
+};
+
+interface ChannelConfig {
+  /** Channel identifier (used in env var prefix). */
+  name: string;
+  baseUrl: string;
+  apiKey: string;
+  models: string;
+}
+
+interface LLMChannelEditorProps {
+  /** All config items from the server (to read existing channel vars). */
+  items: Array<{ key: string; value: string }>;
+  /** Current config version for API calls. */
+  configVersion: string;
+  /** Mask token for secrets. */
+  maskToken: string;
+  /** Called after successful save to reload config. */
+  onSaved: () => void;
+  /** Disable interactions while parent is busy. */
+  disabled?: boolean;
+}
+
+/** Extract `LLM_{NAME}_*` env vars from items and group them by channel. */
+function parseChannelsFromItems(items: Array<{ key: string; value: string }>): ChannelConfig[] {
+  const itemMap = new Map(items.map((i) => [i.key, i.value]));
+  const channelNames = (itemMap.get('LLM_CHANNELS') || '')
+    .split(',')
+    .map((s) => s.trim().toUpperCase())
+    .filter(Boolean);
+
+  if (channelNames.length === 0) {
+    return [];
+  }
+
+  return channelNames.map((name) => ({
+    name: name.toLowerCase(),
+    baseUrl: itemMap.get(`LLM_${name}_BASE_URL`) || '',
+    apiKey: itemMap.get(`LLM_${name}_API_KEY`) || itemMap.get(`LLM_${name}_API_KEYS`) || '',
+    models: itemMap.get(`LLM_${name}_MODELS`) || '',
+  }));
+}
+
+/** Build env var update items from channel list. */
+function channelsToUpdateItems(
+  channels: ChannelConfig[],
+  previousChannelNames: string[],
+): Array<{ key: string; value: string }> {
+  const updates: Array<{ key: string; value: string }> = [];
+  const activeNames = channels.map((c) => c.name.toUpperCase());
+
+  // LLM_CHANNELS
+  updates.push({ key: 'LLM_CHANNELS', value: channels.map((c) => c.name).join(',') });
+
+  // Per-channel vars
+  for (const ch of channels) {
+    const prefix = `LLM_${ch.name.toUpperCase()}`;
+    updates.push({ key: `${prefix}_BASE_URL`, value: ch.baseUrl });
+    // Use API_KEY for single key, API_KEYS for comma-separated multi-key
+    const isMultiKey = ch.apiKey.includes(',');
+    updates.push({ key: `${prefix}_API_KEY${isMultiKey ? 'S' : ''}`, value: ch.apiKey });
+    // Clear the other key variant
+    updates.push({ key: `${prefix}_API_KEY${isMultiKey ? '' : 'S'}`, value: '' });
+    updates.push({ key: `${prefix}_MODELS`, value: ch.models });
+  }
+
+  // Clear removed channel vars
+  for (const oldName of previousChannelNames) {
+    const upper = oldName.toUpperCase();
+    if (!activeNames.includes(upper)) {
+      const prefix = `LLM_${upper}`;
+      updates.push({ key: `${prefix}_BASE_URL`, value: '' });
+      updates.push({ key: `${prefix}_API_KEY`, value: '' });
+      updates.push({ key: `${prefix}_API_KEYS`, value: '' });
+      updates.push({ key: `${prefix}_MODELS`, value: '' });
+    }
+  }
+
+  return updates;
+}
+
+export const LLMChannelEditor: React.FC<LLMChannelEditorProps> = ({
+  items,
+  configVersion,
+  maskToken,
+  onSaved,
+  disabled = false,
+}) => {
+  const initialChannels = useMemo(() => parseChannelsFromItems(items), [items]);
+  const initialNames = useMemo(
+    () => initialChannels.map((c) => c.name),
+    [initialChannels],
+  );
+
+  const [channels, setChannels] = useState<ChannelConfig[]>(initialChannels);
+  const [isSaving, setIsSaving] = useState(false);
+  const [saveMessage, setSaveMessage] = useState<{ type: 'success' | 'error'; text: string } | null>(null);
+  const [visibleKeys, setVisibleKeys] = useState<Record<number, boolean>>({});
+  const [isCollapsed, setIsCollapsed] = useState(initialChannels.length === 0);
+  const [addPreset, setAddPreset] = useState('aihubmix');
+
+  // Detect if user has unsaved channel changes
+  const hasChanges = useMemo(() => {
+    if (channels.length !== initialChannels.length) return true;
+    return channels.some((ch, idx) => {
+      const init = initialChannels[idx];
+      if (!init) return true;
+      return (
+        ch.name !== init.name ||
+        ch.baseUrl !== init.baseUrl ||
+        ch.apiKey !== init.apiKey ||
+        ch.models !== init.models
+      );
+    });
+  }, [channels, initialChannels]);
+
+  const updateChannel = useCallback((index: number, field: keyof ChannelConfig, value: string) => {
+    setChannels((prev) => {
+      const next = [...prev];
+      next[index] = { ...next[index], [field]: value };
+      return next;
+    });
+  }, []);
+
+  const removeChannel = useCallback((index: number) => {
+    setChannels((prev) => prev.filter((_, i) => i !== index));
+    setVisibleKeys((prev) => {
+      const next = { ...prev };
+      delete next[index];
+      return next;
+    });
+  }, []);
+
+  const addChannel = useCallback(() => {
+    const preset = CHANNEL_PRESETS[addPreset] || CHANNEL_PRESETS.custom;
+    // Determine a unique name
+    const baseName = addPreset === 'custom' ? 'custom' : addPreset;
+    const existingNames = new Set(channels.map((c) => c.name));
+    let name = baseName;
+    let counter = 2;
+    while (existingNames.has(name)) {
+      name = `${baseName}${counter}`;
+      counter++;
+    }
+
+    setChannels((prev) => [
+      ...prev,
+      { name, baseUrl: preset.baseUrl, apiKey: '', models: '' },
+    ]);
+    setIsCollapsed(false);
+  }, [addPreset, channels]);
+
+  const handleSave = useCallback(async () => {
+    setIsSaving(true);
+    setSaveMessage(null);
+
+    try {
+      const updateItems = channelsToUpdateItems(channels, initialNames);
+      await systemConfigApi.update({
+        configVersion,
+        maskToken,
+        reloadNow: true,
+        items: updateItems,
+      });
+      setSaveMessage({ type: 'success', text: '渠道配置已保存' });
+      onSaved();
+    } catch (error: unknown) {
+      const msg = error instanceof Error ? error.message : '保存失败';
+      setSaveMessage({ type: 'error', text: msg });
+    } finally {
+      setIsSaving(false);
+    }
+  }, [channels, configVersion, initialNames, maskToken, onSaved]);
+
+  const toggleKeyVisibility = useCallback((index: number) => {
+    setVisibleKeys((prev) => ({ ...prev, [index]: !prev[index] }));
+  }, []);
+
+  const busy = disabled || isSaving;
+
+  return (
+    <div className="rounded-xl border border-cyan/20 bg-elevated/50 p-4">
+      <button
+        type="button"
+        className="flex w-full items-center justify-between text-left"
+        onClick={() => setIsCollapsed((prev) => !prev)}
+      >
+        <div>
+          <h3 className="text-sm font-semibold text-white">LLM 渠道配置</h3>
+          <p className="mt-0.5 text-xs text-muted">
+            {channels.length > 0
+              ? `已配置 ${channels.length} 个渠道：${channels.map((c) => c.name).join('、')}`
+              : '同时使用多个模型平台时启用；只用单个模型可跳过此项'}
+          </p>
+        </div>
+        <span className="text-xs text-muted">{isCollapsed ? '▶ 展开' : '▼ 收起'}</span>
+      </button>
+
+      {!isCollapsed && (
+        <div className="mt-4 space-y-3">
+          {channels.map((channel, index) => (
+            <div
+              key={`${channel.name}-${index}`}
+              className="rounded-lg border border-white/8 bg-card/40 p-3 space-y-2"
+            >
+              <div className="flex items-center justify-between">
+                <div className="flex items-center gap-2">
+                  <span className="text-xs font-medium text-accent">
+                    {CHANNEL_PRESETS[channel.name]?.label || channel.name}
+                  </span>
+                </div>
+                <button
+                  type="button"
+                  className="text-xs text-red-400 hover:text-red-300 disabled:opacity-40"
+                  disabled={busy}
+                  onClick={() => removeChannel(index)}
+                >
+                  删除
+                </button>
+              </div>
+
+              {/* Channel name */}
+              <div>
+                <label className="mb-1 block text-xs text-secondary">渠道名称</label>
+                <input
+                  type="text"
+                  className="input-terminal w-full"
+                  value={channel.name}
+                  disabled={busy}
+                  onChange={(e) => updateChannel(index, 'name', e.target.value.replace(/[^a-zA-Z0-9_]/g, '').toLowerCase())}
+                  placeholder="如 aihubmix、deepseek"
+                />
+              </div>
+
+              {/* Base URL */}
+              <div>
+                <label className="mb-1 block text-xs text-secondary">API 地址（Base URL）</label>
+                <input
+                  type="text"
+                  className="input-terminal w-full"
+                  value={channel.baseUrl}
+                  disabled={busy}
+                  onChange={(e) => updateChannel(index, 'baseUrl', e.target.value)}
+                  placeholder="https://api.example.com/v1（Gemini 原生可留空）"
+                />
+              </div>
+
+              {/* API Key */}
+              <div>
+                <label className="mb-1 block text-xs text-secondary">API Key（多个用逗号分隔）</label>
+                <div className="flex items-center gap-2">
+                  <input
+                    type={visibleKeys[index] ? 'text' : 'password'}
+                    className="input-terminal flex-1"
+                    value={channel.apiKey}
+                    disabled={busy}
+                    onChange={(e) => updateChannel(index, 'apiKey', e.target.value)}
+                    placeholder="sk-xxxxxxxxxxxxxxxx"
+                  />
+                  <button
+                    type="button"
+                    className="btn-secondary !p-2"
+                    onClick={() => toggleKeyVisibility(index)}
+                    title={visibleKeys[index] ? '隐藏' : '显示'}
+                  >
+                    <EyeToggleIcon visible={!!visibleKeys[index]} />
+                  </button>
+                </div>
+              </div>
+
+              {/* Models */}
+              <div>
+                <label className="mb-1 block text-xs text-secondary">模型列表（逗号分隔）</label>
+                <input
+                  type="text"
+                  className="input-terminal w-full"
+                  value={channel.models}
+                  disabled={busy}
+                  onChange={(e) => updateChannel(index, 'models', e.target.value)}
+                  placeholder={CHANNEL_PRESETS[channel.name]?.placeholder || 'model-1,model-2'}
+                />
+                <p className="mt-1 text-[11px] text-muted">
+                  有 Base URL 的渠道无需加 openai/ 前缀，系统自动补全
+                </p>
+              </div>
+            </div>
+          ))}
+
+          {/* Add channel */}
+          <div className="flex flex-wrap items-center gap-2">
+            <select
+              className="input-terminal text-xs"
+              value={addPreset}
+              disabled={busy}
+              onChange={(e) => setAddPreset(e.target.value)}
+            >
+              {Object.entries(CHANNEL_PRESETS).map(([key, preset]) => (
+                <option key={key} value={key}>
+                  {preset.label}
+                </option>
+              ))}
+            </select>
+            <button
+              type="button"
+              className="btn-secondary !px-3 !py-1.5 text-xs"
+              disabled={busy}
+              onClick={addChannel}
+            >
+              + 添加渠道
+            </button>
+          </div>
+
+          {/* Save */}
+          {hasChanges && (
+            <div className="flex items-center gap-3 border-t border-white/8 pt-3">
+              <button
+                type="button"
+                className="btn-primary !px-4 !py-1.5 text-xs"
+                disabled={busy}
+                onClick={() => void handleSave()}
+              >
+                {isSaving ? '保存中...' : '保存渠道'}
+              </button>
+              <button
+                type="button"
+                className="btn-secondary !px-3 !py-1.5 text-xs"
+                disabled={busy}
+                onClick={() => setChannels(initialChannels)}
+              >
+                撤销
+              </button>
+              <span className="text-[11px] text-muted">渠道配置独立保存，与下方字段互不影响</span>
+            </div>
+          )}
+
+          {saveMessage && (
+            <p
+              className={`text-xs ${saveMessage.type === 'success' ? 'text-green-400' : 'text-red-400'}`}
+            >
+              {saveMessage.text}
+            </p>
+          )}
+        </div>
+      )}
+    </div>
+  );
+};

apps/dsa-web/src/components/settings/index.ts

@@ -1,3 +1,4 @@
+export * from './LLMChannelEditor';
 export * from './SettingsAlert';
 export * from './ChangePasswordCard';
 export * from './ImageStockExtractor';

apps/dsa-web/src/pages/HomePage.tsx

@@ -173,22 +173,21 @@ const HomePage: React.FC = () => {
   }, [fetchHistory, isLoadingMore, hasMore]);
 
   // 初始加载 - 自动选择第一条（仅挂载时执行一次）
-  // eslint-disable-next-line react-hooks/exhaustive-deps
   useEffect(() => {
     fetchHistory(true);
+    // eslint-disable-next-line react-hooks/exhaustive-deps
   }, []);
 
   // Background polling: re-fetch history every 30s for CLI-initiated analyses
-  // eslint-disable-next-line react-hooks/exhaustive-deps
   useEffect(() => {
     const interval = setInterval(() => {
       fetchHistory(false, true, true);
     }, 30_000);
     return () => clearInterval(interval);
+    // eslint-disable-next-line react-hooks/exhaustive-deps
   }, []);
 
   // Refresh when tab regains visibility (e.g. user ran main.py in another terminal)
-  // eslint-disable-next-line react-hooks/exhaustive-deps
   useEffect(() => {
     const handleVisibilityChange = () => {
       if (document.visibilityState === 'visible') {
@@ -197,6 +196,7 @@ const HomePage: React.FC = () => {
     };
     document.addEventListener('visibilitychange', handleVisibilityChange);
     return () => document.removeEventListener('visibilitychange', handleVisibilityChange);
+    // eslint-disable-next-line react-hooks/exhaustive-deps
   }, []);
 
   // 点击历史项加载报告

apps/dsa-web/src/pages/SettingsPage.tsx

@@ -4,6 +4,7 @@ import { useAuth, useSystemConfig } from '../hooks';
 import {
   ChangePasswordCard,
   ImageStockExtractor,
+  LLMChannelEditor,
   SettingsAlert,
   SettingsField,
   SettingsLoading,
@@ -53,7 +54,15 @@ const SettingsPage: React.FC = () => {
     };
   }, [clearToast, toast]);
 
-  const activeItems = itemsByCategory[activeCategory] || [];
+  const rawActiveItems = itemsByCategory[activeCategory] || [];
+
+  // Hide per-channel LLM_*_ env vars from the normal field list;
+  // they are managed by the LLMChannelEditor component instead.
+  const LLM_CHANNEL_KEY_RE = /^LLM_[A-Z0-9]+_(BASE_URL|API_KEY|API_KEYS|MODELS|EXTRA_HEADERS)$/;
+  const activeItems =
+    activeCategory === 'ai_model'
+      ? rawActiveItems.filter((item) => !LLM_CHANNEL_KEY_RE.test(item.key))
+      : rawActiveItems;
 
   return (
     <div className="min-h-screen px-4 pb-6 pt-4 md:px-6">
@@ -151,6 +160,15 @@ const SettingsPage: React.FC = () => {
                 />
               </div>
             ) : null}
+            {activeCategory === 'ai_model' ? (
+              <LLMChannelEditor
+                items={rawActiveItems}
+                configVersion={configVersion}
+                maskToken={maskToken}
+                onSaved={() => void load()}
+                disabled={isSaving || isLoading}
+              />
+            ) : null}
             {activeCategory === 'system' && passwordChangeable ? (
               <div className="space-y-3">
                 <ChangePasswordCard />

apps/dsa-web/src/utils/systemConfigI18n.ts

@@ -30,6 +30,12 @@ const fieldTitleMap: Record<string, string> = {
   BRAVE_API_KEYS: 'Brave API Keys',
   REALTIME_SOURCE_PRIORITY: '实时数据源优先级',
   ENABLE_REALTIME_TECHNICAL_INDICATORS: '盘中实时技术面',
+  LITELLM_MODEL: '主模型',
+  LITELLM_FALLBACK_MODELS: '备选模型',
+  LITELLM_CONFIG: 'LiteLLM 配置文件',
+  LLM_CHANNELS: 'LLM 渠道列表',
+  AIHUBMIX_KEY: 'AIHubmix Key',
+  DEEPSEEK_API_KEY: 'DeepSeek API Key',
   GEMINI_API_KEY: 'Gemini API Key',
   GEMINI_MODEL: 'Gemini 模型',
   GEMINI_TEMPERATURE: 'Gemini 温度参数',
@@ -64,6 +70,12 @@ const fieldDescriptionMap: Record<string, string> = {
   BRAVE_API_KEYS: '用于新闻检索的 Brave Search 密钥，支持逗号分隔多个。',
   REALTIME_SOURCE_PRIORITY: '按逗号分隔填写数据源调用优先级。',
   ENABLE_REALTIME_TECHNICAL_INDICATORS: '盘中分析时用实时价计算 MA5/MA10/MA20 与多头排列（Issue #234）；关闭则用昨日收盘。',
+  LITELLM_MODEL: '主模型，格式 provider/model（如 gemini/gemini-2.5-flash）。配置渠道后自动推断。',
+  LITELLM_FALLBACK_MODELS: '备选模型，逗号分隔，主模型失败时按序尝试。',
+  LITELLM_CONFIG: 'LiteLLM YAML 配置文件路径（高级用法），优先级最高。',
+  LLM_CHANNELS: '渠道名称列表（逗号分隔）。推荐使用上方渠道编辑器管理。',
+  AIHUBMIX_KEY: 'AIHubmix 一站式密钥，自动指向 aihubmix.com/v1。',
+  DEEPSEEK_API_KEY: 'DeepSeek 官方 API 密钥。填写后自动使用 deepseek-chat 模型。',
   GEMINI_API_KEY: '用于 Gemini 服务调用的密钥。',
   GEMINI_MODEL: '设置 Gemini 分析模型名称。',
   GEMINI_TEMPERATURE: '控制模型输出随机性，范围通常为 0.0 到 2.0。',

litellm_config.example.yaml

@@ -0,0 +1,74 @@
+# ===================================
+# LiteLLM Router 配置模板
+# ===================================
+#
+# 用法：
+# 1. 复制此文件为 litellm_config.yaml
+# 2. 在 .env 中设置 LITELLM_CONFIG=./litellm_config.yaml
+# 3. 按需配置下面的 model_list
+#
+# 密钥引用格式：
+#   api_key: "os.environ/ENV_VAR_NAME"  → 从环境变量读取，避免明文写入文件
+#   api_key: "sk-xxxxxxxx"              → 直接写入（不推荐）
+#
+# 更多文档: https://docs.litellm.ai/docs/proxy/configs
+# ===================================
+
+model_list:
+  # --- AIHubmix (OpenAI 兼容，一个 Key 使用多种模型) ---
+  - model_name: openai/gpt-4o-mini
+    litellm_params:
+      model: openai/gpt-4o-mini
+      api_key: "os.environ/AIHUBMIX_KEY"
+      api_base: https://aihubmix.com/v1
+
+  - model_name: openai/claude-3-5-sonnet-20241022
+    litellm_params:
+      model: openai/claude-3-5-sonnet-20241022
+      api_key: "os.environ/AIHUBMIX_KEY"
+      api_base: https://aihubmix.com/v1
+
+  # --- DeepSeek 官方 API (原生 provider，自动解析 base_url) ---
+  - model_name: deepseek/deepseek-chat
+    litellm_params:
+      model: deepseek/deepseek-chat
+      api_key: "os.environ/DEEPSEEK_API_KEY"
+
+  - model_name: deepseek/deepseek-reasoner
+    litellm_params:
+      model: deepseek/deepseek-reasoner
+      api_key: "os.environ/DEEPSEEK_API_KEY"
+
+  # --- Google Gemini (原生，多 Key 负载均衡) ---
+  - model_name: gemini/gemini-2.5-flash
+    litellm_params:
+      model: gemini/gemini-2.5-flash
+      api_key: "os.environ/GEMINI_API_KEY_1"
+
+  - model_name: gemini/gemini-2.5-flash
+    litellm_params:
+      model: gemini/gemini-2.5-flash
+      api_key: "os.environ/GEMINI_API_KEY_2"
+
+  # --- Anthropic Claude (原生) ---
+  # - model_name: anthropic/claude-3-5-sonnet-20241022
+  #   litellm_params:
+  #     model: anthropic/claude-3-5-sonnet-20241022
+  #     api_key: "os.environ/ANTHROPIC_API_KEY"
+
+  # --- OpenRouter (聚合平台) ---
+  # - model_name: openai/meta-llama/llama-3-70b-instruct
+  #   litellm_params:
+  #     model: openai/meta-llama/llama-3-70b-instruct
+  #     api_key: "os.environ/OPENROUTER_API_KEY"
+  #     api_base: https://openrouter.ai/api/v1
+
+# ===================================
+# Router 设置（可选）
+# ===================================
+router_settings:
+  routing_strategy: simple-shuffle    # simple-shuffle / least-busy / latency-based
+  num_retries: 2                      # 单个 deployment 失败后重试次数
+  # timeout: 30                       # 请求超时（秒）
+  # allowed_fails: 3                  # deployment 被冷却前允许的失败次数
+  # cooldown_time: 60                 # deployment 冷却时间（秒）

requirements.txt

@@ -28,6 +28,7 @@ json-repair>=0.55.1         # JSON 修复
 # AI 分析
 litellm>=1.80.10            # Unified LLM client (Gemini/Anthropic/OpenAI/DeepSeek etc.)
 openai>=1.0.0               # OpenAI SDK (transitive dependency of litellm, kept explicit)
+PyYAML>=6.0                 # YAML parser for LITELLM_CONFIG support
 
 # 搜索引擎（用于获取股票新闻）
 tavily-python>=0.3.0        # Tavily 搜索 API（每月 1000 次免费）

src/agent/llm_adapter.py

@@ -15,7 +15,7 @@ from typing import Any, Dict, List, Optional
 import litellm
 from litellm import Router
 
-from src.config import get_config
+from src.config import get_config, get_api_keys_for_model, extra_litellm_params
 
 logger = logging.getLogger(__name__)
 
@@ -110,62 +110,70 @@ class LLMToolAdapter:
         self._litellm_available = False
         self._init_litellm()
 
-    def _get_api_keys_for_model(self, model: str) -> List[str]:
-        """Return API keys for the given litellm model based on provider prefix."""
-        config = self._config
-        if model.startswith("gemini/") or model.startswith("vertex_ai/"):
-            return [k for k in config.gemini_api_keys if k and len(k) >= 8]
-        if model.startswith("anthropic/"):
-            return [k for k in config.anthropic_api_keys if k and len(k) >= 8]
-        # openai/, deepseek/, or any other provider uses openai_api_keys
-        return [k for k in config.openai_api_keys if k and len(k) >= 8]
-
-    def _extra_litellm_params(self, model: str) -> dict:
-        """Build extra litellm params (api_base, custom headers) for a model."""
-        config = self._config
-        params: Dict[str, Any] = {}
-        if not model.startswith("gemini/") and not model.startswith("anthropic/") and not model.startswith("vertex_ai/"):
-            if config.openai_base_url:
-                params["api_base"] = config.openai_base_url
-            if config.openai_base_url and "aihubmix.com" in config.openai_base_url:
-                params["extra_headers"] = {"APP-Code": "GPIJ3886"}
-        return params
+    def _has_channel_config(self) -> bool:
+        """Check if multi-channel config (channels / YAML) is active."""
+        return bool(self._config.llm_model_list) and not all(
+            e.get('model_name', '').startswith('__legacy_') for e in self._config.llm_model_list
+        )
 
     def _init_litellm(self) -> None:
-        """Initialize litellm Router for multi-key, or flag single-key availability."""
+        """Initialize litellm Router from channels / YAML / legacy keys."""
         config = self._config
         litellm_model = config.litellm_model
         if not litellm_model:
             logger.warning("Agent LLM: LITELLM_MODEL not configured")
             return
 
-        keys = self._get_api_keys_for_model(litellm_model)
-        if not keys:
-            logger.warning(f"Agent LLM: No API keys found for model {litellm_model}")
-            return
-
         self._litellm_available = True
 
-        if len(keys) > 1:
-            extra_params = self._extra_litellm_params(litellm_model)
-            model_list = [
-                {
-                    "model_name": litellm_model,
-                    "litellm_params": {
-                        "model": litellm_model,
-                        "api_key": k,
-                        **extra_params,
-                    },
-                }
-                for k in keys
-            ]
+        # --- Channel / YAML path ---
+        if self._has_channel_config():
+            model_list = config.llm_model_list
             self._router = Router(
                 model_list=model_list,
                 routing_strategy="simple-shuffle",
                 num_retries=2,
             )
-            models_in_router = list(dict.fromkeys(m["litellm_params"]["model"] for m in model_list))
-            logger.info(f"Agent LLM: Router initialized with {len(keys)} keys for {litellm_model} (models: {models_in_router})")
+            unique_models = list(dict.fromkeys(
+                e['litellm_params']['model'] for e in model_list
+            ))
+            logger.info(
+                f"Agent LLM: Router initialized from channels/YAML — "
+                f"{len(model_list)} deployment(s), models: {unique_models}"
+            )
+            return
+
+        # --- Legacy path ---
+        keys = get_api_keys_for_model(litellm_model, config)
+        if not keys:
+            logger.info(
+                f"Agent LLM: litellm initialized (model={litellm_model}, "
+                f"API key from environment)"
+            )
+            return
+
+        if len(keys) > 1:
+            ep = extra_litellm_params(litellm_model, config)
+            legacy_model_list = [
+                {
+                    "model_name": litellm_model,
+                    "litellm_params": {
+                        "model": litellm_model,
+                        "api_key": k,
+                        **ep,
+                    },
+                }
+                for k in keys
+            ]
+            self._router = Router(
+                model_list=legacy_model_list,
+                routing_strategy="simple-shuffle",
+                num_retries=2,
+            )
+            logger.info(
+                f"Agent LLM: Legacy Router initialized with {len(keys)} keys "
+                f"for {litellm_model}"
+            )
         else:
             logger.info(f"Agent LLM: litellm initialized (model={litellm_model})")
 
@@ -246,13 +254,19 @@ class LLMToolAdapter:
             call_kwargs["tools"] = tools
 
         # Use Router for primary model (multi-key), direct litellm for others
-        if self._router and model == self._config.litellm_model:
+        use_channel_router = self._has_channel_config()
+        if use_channel_router and self._router:
+            # Channel / YAML path: Router manages all models
+            response = self._router.completion(**call_kwargs)
+        elif self._router and model == self._config.litellm_model:
+            # Legacy path: Router for primary model multi-key
             response = self._router.completion(**call_kwargs)
         else:
-            keys = self._get_api_keys_for_model(model)
+            # Legacy path: direct call for fallback/other models
+            keys = get_api_keys_for_model(model, self._config)
             if keys:
                 call_kwargs["api_key"] = keys[0]
-            call_kwargs.update(self._extra_litellm_params(model))
+            call_kwargs.update(extra_litellm_params(model, self._config))
             response = litellm.completion(**call_kwargs)
 
         return self._parse_litellm_response(response, model)

src/analyzer.py

@@ -21,7 +21,7 @@ from json_repair import repair_json
 from litellm import Router
 
 from src.agent.llm_adapter import get_thinking_extra_body
-from src.config import Config, get_config
+from src.config import Config, get_config, get_api_keys_for_model, extra_litellm_params
 
 logger = logging.getLogger(__name__)
 
@@ -533,44 +533,46 @@ class GeminiAnalyzer:
         if not self._litellm_available:
             logger.warning("No LLM configured (LITELLM_MODEL / API keys), AI analysis will be unavailable")
 
-    @staticmethod
-    def _get_api_keys_for_model(model: str, config: Config) -> List[str]:
-        """Return API keys for a litellm model based on provider prefix."""
-        if model.startswith("gemini/") or model.startswith("vertex_ai/"):
-            return [k for k in config.gemini_api_keys if k and len(k) >= 8]
-        if model.startswith("anthropic/"):
-            return [k for k in config.anthropic_api_keys if k and len(k) >= 8]
-        return [k for k in config.openai_api_keys if k and len(k) >= 8]
-
-    @staticmethod
-    def _extra_litellm_params(model: str, config: Config) -> dict:
-        """Build extra litellm params (api_base, headers) for OpenAI-compatible models."""
-        params: Dict[str, Any] = {}
-        if not model.startswith("gemini/") and not model.startswith("anthropic/") and not model.startswith("vertex_ai/"):
-            if config.openai_base_url:
-                params["api_base"] = config.openai_base_url
-            if config.openai_base_url and "aihubmix.com" in config.openai_base_url:
-                params["extra_headers"] = {"APP-Code": "GPIJ3886"}
-        return params
+    def _has_channel_config(self, config: Config) -> bool:
+        """Check if multi-channel config (channels / YAML / legacy model_list) is active."""
+        return bool(config.llm_model_list) and not all(
+            e.get('model_name', '').startswith('__legacy_') for e in config.llm_model_list
+        )
 
     def _init_litellm(self) -> None:
-        """Initialize litellm Router (multi-key) or flag single-key availability."""
+        """Initialize litellm Router from channels / YAML / legacy keys."""
         config = get_config()
         litellm_model = config.litellm_model
         if not litellm_model:
             logger.warning("Analyzer LLM: LITELLM_MODEL not configured")
             return
 
-        keys = self._get_api_keys_for_model(litellm_model, config)
-        if not keys:
-            logger.warning(f"Analyzer LLM: No API keys found for model {litellm_model}")
-            return
-
         self._litellm_available = True
 
+        # --- Channel / YAML path: build Router from pre-built model_list ---
+        if self._has_channel_config(config):
+            model_list = config.llm_model_list
+            self._router = Router(
+                model_list=model_list,
+                routing_strategy="simple-shuffle",
+                num_retries=2,
+            )
+            unique_models = list(dict.fromkeys(
+                e['litellm_params']['model'] for e in model_list
+            ))
+            logger.info(
+                f"Analyzer LLM: Router initialized from channels/YAML — "
+                f"{len(model_list)} deployment(s), models: {unique_models}"
+            )
+            return
+
+        # --- Legacy path: build Router for multi-key, or use single key ---
+        keys = get_api_keys_for_model(litellm_model, config)
+
         if len(keys) > 1:
-            extra_params = self._extra_litellm_params(litellm_model, config)
-            model_list = [
+            # Build legacy Router for primary model multi-key load-balancing
+            extra_params = extra_litellm_params(litellm_model, config)
+            legacy_model_list = [
                 {
                     "model_name": litellm_model,
                     "litellm_params": {
@@ -582,14 +584,21 @@ class GeminiAnalyzer:
                 for k in keys
             ]
             self._router = Router(
-                model_list=model_list,
+                model_list=legacy_model_list,
                 routing_strategy="simple-shuffle",
                 num_retries=2,
             )
-            models_in_router = list(dict.fromkeys(m["litellm_params"]["model"] for m in model_list))
-            logger.info(f"Analyzer LLM: Router initialized with {len(keys)} keys for {litellm_model} (models: {models_in_router})")
-        else:
+            logger.info(
+                f"Analyzer LLM: Legacy Router initialized with {len(keys)} keys "
+                f"for {litellm_model}"
+            )
+        elif keys:
             logger.info(f"Analyzer LLM: litellm initialized (model={litellm_model})")
+        else:
+            logger.info(
+                f"Analyzer LLM: litellm initialized (model={litellm_model}, "
+                f"API key from environment)"
+            )
 
     def is_available(self) -> bool:
         """Check if LiteLLM is properly configured with at least one API key."""
@@ -598,6 +607,11 @@ class GeminiAnalyzer:
     def _call_litellm(self, prompt: str, generation_config: dict) -> str:
         """Call LLM via litellm with fallback across configured models.
 
+        When channels/YAML are configured, every model goes through the Router
+        (which handles per-model key selection, load balancing, and retries).
+        In legacy mode, the primary model may use the Router while fallback
+        models fall back to direct litellm.completion().
+
         Args:
             prompt: User prompt text.
             generation_config: Dict with optional keys: temperature, max_output_tokens, max_tokens.
@@ -616,12 +630,10 @@ class GeminiAnalyzer:
         models_to_try = [config.litellm_model] + (config.litellm_fallback_models or [])
         models_to_try = [m for m in models_to_try if m]
 
+        use_channel_router = self._has_channel_config(config)
+
         last_error = None
         for model in models_to_try:
-            keys = self._get_api_keys_for_model(model, config)
-            if not keys:
-                logger.debug(f"[LiteLLM] Skipping {model}: no API keys")
-                continue
             try:
                 model_short = model.split("/")[-1] if "/" in model else model
                 call_kwargs: Dict[str, Any] = {
@@ -637,11 +649,18 @@ class GeminiAnalyzer:
                 if extra:
                     call_kwargs["extra_body"] = extra
 
-                if self._router and model == config.litellm_model:
+                if use_channel_router and self._router:
+                    # Channel / YAML path: Router manages key + base_url per model
+                    response = self._router.completion(**call_kwargs)
+                elif self._router and model == config.litellm_model:
+                    # Legacy path: Router only for primary model multi-key
                     response = self._router.completion(**call_kwargs)
                 else:
+                    # Legacy path: direct call for fallback models
+                    keys = get_api_keys_for_model(model, config)
+                    if keys:
                         call_kwargs["api_key"] = keys[0]
-                    call_kwargs.update(self._extra_litellm_params(model, config))
+                    call_kwargs.update(extra_litellm_params(model, config))
                     response = litellm.completion(**call_kwargs)
 
                 if response and response.choices and response.choices[0].message.content:

src/config.py

@@ -10,10 +10,11 @@ A股自选股智能分析系统 - 配置管理模块
 3. 提供类型安全的配置访问接口
 """
 
+import json
 import os
 import re
 from pathlib import Path
-from typing import List, Optional, Tuple
+from typing import Any, Dict, List, Optional, Tuple
 from dotenv import load_dotenv, dotenv_values
 from dataclasses import dataclass, field
 
@@ -64,10 +65,19 @@ class Config:
     litellm_model: str = ""  # Primary model; must include provider prefix when set explicitly
     litellm_fallback_models: List[str] = field(default_factory=list)  # Cross-model fallback list
 
+    # --- Multi-channel LLM config (new) ---
+    # LITELLM_CONFIG: path to a standard litellm_config.yaml file (most powerful)
+    litellm_config_path: Optional[str] = None
+    # LLM_CHANNELS: list of channel dicts, each with name/base_url/api_keys/models
+    llm_channels: List[Dict[str, Any]] = field(default_factory=list)
+    # Pre-built LiteLLM Router model_list (populated from channels, YAML, or legacy keys)
+    llm_model_list: List[Dict[str, Any]] = field(default_factory=list)
+
     # Multi-key support: each list is parsed from *_API_KEYS (comma-separated) with single-key fallback
     gemini_api_keys: List[str] = field(default_factory=list)
     anthropic_api_keys: List[str] = field(default_factory=list)
     openai_api_keys: List[str] = field(default_factory=list)
+    deepseek_api_keys: List[str] = field(default_factory=list)
 
     # Legacy single-key fields (kept for backward compatibility; gemini_api_keys[0] when set)
     gemini_api_key: Optional[str] = None
@@ -394,6 +404,14 @@ class Config:
             if _fallback_key:
                 openai_api_keys = [_fallback_key]
 
+        # DEEPSEEK_API_KEYS > DEEPSEEK_API_KEY (independent from OpenAI-compatible layer)
+        _deepseek_keys_raw = os.getenv('DEEPSEEK_API_KEYS', '')
+        deepseek_api_keys = [k.strip() for k in _deepseek_keys_raw.split(',') if k.strip()]
+        if not deepseek_api_keys:
+            _single_deepseek = os.getenv('DEEPSEEK_API_KEY', '').strip()
+            if _single_deepseek:
+                deepseek_api_keys = [_single_deepseek]
+
         # LITELLM_MODEL: explicit config takes precedence; else infer from available keys
         litellm_model = os.getenv('LITELLM_MODEL', '').strip()
         if not litellm_model:
@@ -404,6 +422,8 @@ class Config:
                 litellm_model = f'gemini/{_gemini_model_name}'
             elif anthropic_api_keys:
                 litellm_model = f'anthropic/{_anthropic_model_name}'
+            elif deepseek_api_keys:
+                litellm_model = 'deepseek/deepseek-chat'
             elif openai_api_keys:
                 # For openai-compatible models, add prefix only if not already prefixed
                 if '/' not in _openai_model_name:
@@ -424,6 +444,50 @@ class Config:
             else:
                 litellm_fallback_models = []
 
+        # === LLM Channels + YAML config ===
+        litellm_config_path = os.getenv('LITELLM_CONFIG', '').strip() or None
+        llm_channels: List[Dict[str, Any]] = []
+        llm_model_list: List[Dict[str, Any]] = []
+
+        # Priority 1: LITELLM_CONFIG (standard LiteLLM YAML config file)
+        if litellm_config_path:
+            llm_model_list = cls._parse_litellm_yaml(litellm_config_path)
+
+        # Priority 2: LLM_CHANNELS (env var based channel config)
+        if not llm_model_list:
+            _channels_str = os.getenv('LLM_CHANNELS', '').strip()
+            if _channels_str:
+                llm_channels = cls._parse_llm_channels(_channels_str)
+                llm_model_list = cls._channels_to_model_list(llm_channels)
+
+        # Priority 3: Legacy env vars → auto-build model_list (backward compatible)
+        if not llm_model_list:
+            llm_model_list = cls._legacy_keys_to_model_list(
+                gemini_api_keys, anthropic_api_keys, openai_api_keys,
+                os.getenv('OPENAI_BASE_URL') or (
+                    'https://aihubmix.com/v1' if os.getenv('AIHUBMIX_KEY') else None
+                ),
+                deepseek_api_keys,
+            )
+
+        # Auto-infer LITELLM_MODEL from channels when not explicitly set
+        if not litellm_model and llm_channels:
+            for _ch in llm_channels:
+                if _ch.get('models'):
+                    litellm_model = _ch['models'][0]
+                    break
+
+        # Auto-infer LITELLM_FALLBACK_MODELS from channels when not explicitly set
+        if not litellm_fallback_models and llm_channels and litellm_model:
+            _all_ch_models: List[str] = []
+            for _ch in llm_channels:
+                _all_ch_models.extend(_ch.get('models', []))
+            _seen = {litellm_model}
+            litellm_fallback_models = [
+                m for m in _all_ch_models
+                if m not in _seen and not _seen.add(m)  # type: ignore[func-returns-value]
+            ]
+
         # 解析搜索引擎 API Keys（支持多个 key，逗号分隔）
         bocha_keys_str = os.getenv('BOCHA_API_KEYS', '')
         bocha_api_keys = [k.strip() for k in bocha_keys_str.split(',') if k.strip()]
@@ -455,9 +519,13 @@ class Config:
             tushare_token=os.getenv('TUSHARE_TOKEN'),
             litellm_model=litellm_model,
             litellm_fallback_models=litellm_fallback_models,
+            litellm_config_path=litellm_config_path,
+            llm_channels=llm_channels,
+            llm_model_list=llm_model_list,
             gemini_api_keys=gemini_api_keys,
             anthropic_api_keys=anthropic_api_keys,
             openai_api_keys=openai_api_keys,
+            deepseek_api_keys=deepseek_api_keys,
             gemini_api_key=os.getenv('GEMINI_API_KEY'),
             gemini_model=os.getenv('GEMINI_MODEL', 'gemini-3-flash-preview'),
             gemini_model_fallback=os.getenv('GEMINI_MODEL_FALLBACK', 'gemini-2.5-flash'),
@@ -596,6 +664,204 @@ class Config:
             circuit_breaker_cooldown=int(os.getenv('CIRCUIT_BREAKER_COOLDOWN', '300'))
         )
     
+    @classmethod
+    def _parse_litellm_yaml(cls, config_path: str) -> List[Dict[str, Any]]:
+        """Parse a standard LiteLLM config YAML file into Router model_list.
+
+        Supports the ``os.environ/VAR_NAME`` syntax for secret references.
+        Returns an empty list on any error (logged, never raises).
+        """
+        import logging
+        _logger = logging.getLogger(__name__)
+        try:
+            import yaml
+        except ImportError:
+            _logger.warning("PyYAML not installed; LITELLM_CONFIG ignored. Install with: pip install pyyaml")
+            return []
+
+        path = Path(config_path)
+        if not path.is_absolute():
+            path = Path(__file__).parent.parent / path
+        if not path.exists():
+            _logger.warning(f"LITELLM_CONFIG file not found: {path}")
+            return []
+
+        try:
+            with open(path, encoding='utf-8') as f:
+                yaml_config = yaml.safe_load(f) or {}
+        except Exception as e:
+            _logger.warning(f"Failed to parse LITELLM_CONFIG: {e}")
+            return []
+
+        model_list = yaml_config.get('model_list', [])
+        if not isinstance(model_list, list):
+            _logger.warning("LITELLM_CONFIG: model_list must be a list")
+            return []
+
+        # Resolve os.environ/ references in string params
+        for entry in model_list:
+            params = entry.get('litellm_params', {})
+            for key in list(params.keys()):
+                val = params.get(key)
+                if isinstance(val, str) and val.startswith('os.environ/'):
+                    env_name = val.split('/', 1)[1]
+                    params[key] = os.getenv(env_name, '')
+
+        _logger.info(f"LITELLM_CONFIG: loaded {len(model_list)} model deployment(s) from {path}")
+        return model_list
+
+    @classmethod
+    def _parse_llm_channels(cls, channels_str: str) -> List[Dict[str, Any]]:
+        """Parse LLM_CHANNELS env var and per-channel env vars.
+
+        Format:
+            LLM_CHANNELS=aihubmix,deepseek,gemini
+            LLM_AIHUBMIX_BASE_URL=https://aihubmix.com/v1
+            LLM_AIHUBMIX_API_KEY=sk-xxx           (or LLM_AIHUBMIX_API_KEYS=k1,k2)
+            LLM_AIHUBMIX_MODELS=openai/gpt-4o-mini,openai/claude-3-5-sonnet
+        """
+        import logging
+        _logger = logging.getLogger(__name__)
+
+        channels: List[Dict[str, Any]] = []
+        for raw_name in channels_str.split(','):
+            ch_name = raw_name.strip()
+            if not ch_name:
+                continue
+            ch_upper = ch_name.upper()
+
+            base_url = os.getenv(f'LLM_{ch_upper}_BASE_URL', '').strip() or None
+
+            # API keys: LLM_{NAME}_API_KEYS (multi) > LLM_{NAME}_API_KEY (single)
+            api_keys_raw = os.getenv(f'LLM_{ch_upper}_API_KEYS', '')
+            api_keys = [k.strip() for k in api_keys_raw.split(',') if k.strip()]
+            if not api_keys:
+                single_key = os.getenv(f'LLM_{ch_upper}_API_KEY', '').strip()
+                if single_key:
+                    api_keys = [single_key]
+
+            # Models
+            models_raw = os.getenv(f'LLM_{ch_upper}_MODELS', '')
+            models = [m.strip() for m in models_raw.split(',') if m.strip()]
+            # Auto-prefix: models without provider prefix in channels with base_url → openai/
+            models = [
+                (f'openai/{m}' if '/' not in m and base_url else m)
+                for m in models
+            ]
+
+            # Extra headers (JSON string, optional)
+            extra_headers_raw = os.getenv(f'LLM_{ch_upper}_EXTRA_HEADERS', '').strip()
+            extra_headers = None
+            if extra_headers_raw:
+                try:
+                    extra_headers = json.loads(extra_headers_raw)
+                except json.JSONDecodeError:
+                    _logger.warning(f"LLM_{ch_upper}_EXTRA_HEADERS: invalid JSON, ignored")
+
+            if not api_keys:
+                _logger.warning(f"LLM channel '{ch_name}': no API key configured, skipped")
+                continue
+            if not models:
+                _logger.warning(f"LLM channel '{ch_name}': no models configured, skipped")
+                continue
+
+            channels.append({
+                'name': ch_name.lower(),
+                'base_url': base_url,
+                'api_keys': api_keys,
+                'models': models,
+                'extra_headers': extra_headers,
+            })
+            _logger.info(f"LLM channel '{ch_name}': {len(models)} model(s), {len(api_keys)} key(s)")
+
+        return channels
+
+    @classmethod
+    def _channels_to_model_list(cls, channels: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+        """Convert parsed LLM channels to LiteLLM Router model_list format."""
+        model_list: List[Dict[str, Any]] = []
+        for ch in channels:
+            for model_name in ch['models']:
+                for api_key in ch['api_keys']:
+                    litellm_params: Dict[str, Any] = {
+                        'model': model_name,
+                        'api_key': api_key,
+                    }
+                    if ch['base_url']:
+                        litellm_params['api_base'] = ch['base_url']
+                    # Auto-inject aihubmix sponsored header
+                    headers = dict(ch.get('extra_headers') or {})
+                    if ch['base_url'] and 'aihubmix.com' in ch['base_url']:
+                        headers.setdefault('APP-Code', 'GPIJ3886')
+                    if headers:
+                        litellm_params['extra_headers'] = headers
+
+                    model_list.append({
+                        'model_name': model_name,
+                        'litellm_params': litellm_params,
+                    })
+        return model_list
+
+    @classmethod
+    def _legacy_keys_to_model_list(
+        cls,
+        gemini_keys: List[str],
+        anthropic_keys: List[str],
+        openai_keys: List[str],
+        openai_base_url: Optional[str],
+        deepseek_keys: Optional[List[str]] = None,
+    ) -> List[Dict[str, Any]]:
+        """Build Router model_list from legacy per-provider keys (backward compat).
+
+        Returns a model_list where each provider's keys are expanded into
+        deployments, keyed by placeholder model_name tokens.  The analyzer
+        resolves actual model_names at call time from LITELLM_MODEL /
+        LITELLM_FALLBACK_MODELS.
+        """
+        model_list: List[Dict[str, Any]] = []
+
+        # Gemini keys
+        for k in gemini_keys:
+            if k and len(k) >= 8:
+                model_list.append({
+                    'model_name': '__legacy_gemini__',
+                    'litellm_params': {'model': '__legacy_gemini__', 'api_key': k},
+                })
+
+        # Anthropic keys
+        for k in anthropic_keys:
+            if k and len(k) >= 8:
+                model_list.append({
+                    'model_name': '__legacy_anthropic__',
+                    'litellm_params': {'model': '__legacy_anthropic__', 'api_key': k},
+                })
+
+        # OpenAI-compatible keys
+        for k in openai_keys:
+            if k and len(k) >= 8:
+                params: Dict[str, Any] = {'model': '__legacy_openai__', 'api_key': k}
+                if openai_base_url:
+                    params['api_base'] = openai_base_url
+                if openai_base_url and 'aihubmix.com' in openai_base_url:
+                    params['extra_headers'] = {'APP-Code': 'GPIJ3886'}
+                model_list.append({
+                    'model_name': '__legacy_openai__',
+                    'litellm_params': params,
+                })
+
+        # DeepSeek keys (native litellm provider — auto-resolves api_base)
+        for k in (deepseek_keys or []):
+            if k and len(k) >= 8:
+                model_list.append({
+                    'model_name': '__legacy_deepseek__',
+                    'litellm_params': {
+                        'model': '__legacy_deepseek__',
+                        'api_key': k,
+                    },
+                })
+
+        return model_list
+
     @classmethod
     def _parse_stock_email_groups(cls) -> List[Tuple[List[str], List[str]]]:
         """
@@ -765,6 +1031,47 @@ def get_config() -> Config:
     return Config.get_instance()
 
 
+# ============================================================
+# Shared LLM helpers (used by both analyzer and agent/llm_adapter)
+# ============================================================
+
+def get_api_keys_for_model(model: str, config: Config) -> List[str]:
+    """Return explicitly managed API keys for a litellm model (legacy path only).
+
+    When llm_model_list is populated (channels / YAML), the Router handles key
+    selection, so this function is not needed.  Kept for backward compat when
+    no Router is built and a direct litellm.completion() call is needed.
+    """
+    if model.startswith("gemini/") or model.startswith("vertex_ai/"):
+        return [k for k in config.gemini_api_keys if k and len(k) >= 8]
+    if model.startswith("anthropic/"):
+        return [k for k in config.anthropic_api_keys if k and len(k) >= 8]
+    if model.startswith("deepseek/"):
+        return [k for k in config.deepseek_api_keys if k and len(k) >= 8]
+    if model.startswith("openai/") or "/" not in model:
+        return [k for k in config.openai_api_keys if k and len(k) >= 8]
+    # Other LiteLLM-native providers – API key resolved from env vars
+    return []
+
+
+def extra_litellm_params(model: str, config: Config) -> Dict[str, Any]:
+    """Build extra litellm params for a model (legacy path only).
+
+    When llm_model_list is populated, the Router already carries api_base
+    and headers per-deployment, so this is not called.
+    """
+    params: Dict[str, Any] = {}
+    # deepseek/ provider: litellm auto-resolves api_base, no manual override needed
+    if model.startswith("deepseek/"):
+        return params
+    if model.startswith("openai/") or "/" not in model:
+        if config.openai_base_url:
+            params["api_base"] = config.openai_base_url
+        if config.openai_base_url and "aihubmix.com" in config.openai_base_url:
+            params["extra_headers"] = {"APP-Code": "GPIJ3886"}
+    return params
+
+
 if __name__ == "__main__":
     # 测试配置加载
     config = get_config()

src/core/config_registry.py

@@ -78,6 +78,113 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {"min_items": 1},
         "display_order": 10,
     },
+    # ------------------------------------------------------------------
+    # AI Model – LiteLLM unified config
+    # ------------------------------------------------------------------
+    "LITELLM_MODEL": {
+        "title": "Primary Model (LiteLLM)",
+        "description": "Unified primary model in provider/model format (e.g. gemini/gemini-3-flash-preview, openai/deepseek-chat, anthropic/claude-3-5-sonnet-20241022). If empty, auto-inferred from available API keys.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 1,
+    },
+    "LITELLM_FALLBACK_MODELS": {
+        "title": "Fallback Models (LiteLLM)",
+        "description": "Comma-separated fallback models tried when the primary model fails (e.g. anthropic/claude-3-5-sonnet-20241022,openai/gpt-4o-mini). Enables cross-provider redundancy.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 2,
+    },
+    # ------------------------------------------------------------------
+    # AI Model – Multi-channel LLM configuration
+    # ------------------------------------------------------------------
+    "LITELLM_CONFIG": {
+        "title": "LiteLLM Config File",
+        "description": "Path to litellm_config.yaml (advanced). Takes priority over channels and legacy keys.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 3,
+    },
+    "LLM_CHANNELS": {
+        "title": "LLM Channels",
+        "description": "Channel names (comma-separated). Managed by the channel editor above.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 4,
+    },
+    "AIHUBMIX_KEY": {
+        "title": "AIHubmix Key",
+        "description": "AIHubmix one-stop API key – access all mainstream models with a single key, no VPN required. Auto-sets base URL to aihubmix.com/v1. Get key: https://aihubmix.com/?aff=CfMq",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 5,
+    },
+    # ------------------------------------------------------------------
+    # AI Model – DeepSeek official (independent from OpenAI-compatible)
+    # ------------------------------------------------------------------
+    "DEEPSEEK_API_KEY": {
+        "title": "DeepSeek API Key",
+        "description": "Official DeepSeek API key (from https://platform.deepseek.com). Auto-infers openai/deepseek-chat when set alone. Also works in multi-channel mode.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 6,
+    },
+    "DEEPSEEK_API_KEYS": {
+        "title": "DeepSeek API Keys (Multi)",
+        "description": "Comma-separated DeepSeek API keys for load balancing. Takes priority over DEEPSEEK_API_KEY.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 7,
+    },
     "TUSHARE_TOKEN": {
         "title": "Tushare Token",
         "description": "Token for Tushare Pro API.",
@@ -162,6 +269,76 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {"multi_value": True, "delimiter": ","},
         "display_order": 50,
     },
+    "BOCHA_API_KEYS": {
+        "title": "Bocha API Keys",
+        "description": "Comma-separated Bocha Search API keys.",
+        "category": "data_source",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 51,
+    },
+    "ENABLE_REALTIME_QUOTE": {
+        "title": "Enable Realtime Quote",
+        "description": "Enable realtime market quotes. Disable to only use historical close prices.",
+        "category": "data_source",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "true",
+        "options": [],
+        "validation": {},
+        "display_order": 22,
+    },
+    "ENABLE_CHIP_DISTRIBUTION": {
+        "title": "Enable Chip Distribution",
+        "description": "Enable chip distribution analysis. May be unstable; recommended to disable on cloud deployments.",
+        "category": "data_source",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "true",
+        "options": [],
+        "validation": {},
+        "display_order": 23,
+    },
+    "NEWS_MAX_AGE_DAYS": {
+        "title": "News Max Age (Days)",
+        "description": "Maximum age of news in days. Older articles are excluded from analysis context.",
+        "category": "data_source",
+        "data_type": "integer",
+        "ui_control": "number",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "3",
+        "options": [],
+        "validation": {"min": 1, "max": 30},
+        "display_order": 60,
+    },
+    "BIAS_THRESHOLD": {
+        "title": "Bias Threshold (%)",
+        "description": "Deviation threshold from MA5 (%). Exceeding this triggers 'do not chase' warning. Strong trend stocks auto-widen to 1.5x.",
+        "category": "data_source",
+        "data_type": "number",
+        "ui_control": "number",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "5.0",
+        "options": [],
+        "validation": {"min": 0.0, "max": 50.0},
+        "display_order": 61,
+    },
     "PYTDX_HOST": {
         "title": "Pytdx Host",
         "description": "Tongdaxin data server IP. Used with PYTDX_PORT. Overrides built-in defaults.",
@@ -206,7 +383,7 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
     },
     "GEMINI_API_KEY": {
         "title": "Gemini API Key",
-        "description": "API key for Gemini service.",
+        "description": "Single API key for Gemini service (from https://aistudio.google.com).",
         "category": "ai_model",
         "data_type": "string",
         "ui_control": "password",
@@ -218,6 +395,20 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 10,
     },
+    "GEMINI_API_KEYS": {
+        "title": "Gemini API Keys (Multi)",
+        "description": "Comma-separated Gemini API keys for load balancing. Takes priority over GEMINI_API_KEY.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 11,
+    },
     "GEMINI_MODEL": {
         "title": "Gemini Model",
         "description": "Gemini model name.",
@@ -232,6 +423,20 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 20,
     },
+    "GEMINI_MODEL_FALLBACK": {
+        "title": "Gemini Fallback Model",
+        "description": "Fallback Gemini model name (used when LITELLM_FALLBACK_MODELS is not set and primary is Gemini).",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "gemini-2.5-flash",
+        "options": [],
+        "validation": {},
+        "display_order": 21,
+    },
     "GEMINI_TEMPERATURE": {
         "title": "Gemini Temperature",
         "description": "Temperature in range [0.0, 2.0].",
@@ -260,6 +465,20 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 40,
     },
+    "OPENAI_API_KEYS": {
+        "title": "OpenAI API Keys (Multi)",
+        "description": "Comma-separated OpenAI-compatible API keys for load balancing. Takes priority over AIHUBMIX_KEY and OPENAI_API_KEY.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 41,
+    },
     "OPENAI_BASE_URL": {
         "title": "OpenAI Base URL",
         "description": "Base URL for OpenAI-compatible endpoint.",
@@ -302,9 +521,23 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 61,
     },
+    "OPENAI_TEMPERATURE": {
+        "title": "OpenAI Temperature",
+        "description": "Temperature for OpenAI-compatible models in range [0.0, 2.0].",
+        "category": "ai_model",
+        "data_type": "number",
+        "ui_control": "number",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "0.7",
+        "options": [],
+        "validation": {"min": 0.0, "max": 2.0},
+        "display_order": 62,
+    },
     "ANTHROPIC_API_KEY": {
         "title": "Anthropic API Key",
-        "description": "Anthropic Claude 服务的 API Key。",
+        "description": "Anthropic Claude API key (from https://console.anthropic.com).",
         "category": "ai_model",
         "data_type": "string",
         "ui_control": "password",
@@ -316,6 +549,20 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 35,
     },
+    "ANTHROPIC_API_KEYS": {
+        "title": "Anthropic API Keys (Multi)",
+        "description": "Comma-separated Anthropic API keys for load balancing. Takes priority over ANTHROPIC_API_KEY.",
+        "category": "ai_model",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 35,
+    },
     "ANTHROPIC_MODEL": {
         "title": "Anthropic Model",
         "description": "Claude 模型名称（如 claude-3-5-sonnet-20241022）。",
@@ -470,6 +717,290 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 53,
     },
+    # ------------------------------------------------------------------
+    # Notification – Feishu
+    # ------------------------------------------------------------------
+    "FEISHU_WEBHOOK_URL": {
+        "title": "Feishu Webhook URL",
+        "description": "Webhook URL for Feishu (Lark) bot notifications.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 12,
+    },
+    "FEISHU_APP_ID": {
+        "title": "Feishu App ID",
+        "description": "Feishu app bot App ID (for event-driven bot mode).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 13,
+    },
+    "FEISHU_APP_SECRET": {
+        "title": "Feishu App Secret",
+        "description": "Feishu app bot App Secret.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 14,
+    },
+    # ------------------------------------------------------------------
+    # Notification – Telegram
+    # ------------------------------------------------------------------
+    "TELEGRAM_BOT_TOKEN": {
+        "title": "Telegram Bot Token",
+        "description": "Telegram bot token (from @BotFather).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 15,
+    },
+    "TELEGRAM_CHAT_ID": {
+        "title": "Telegram Chat ID",
+        "description": "Telegram chat/group ID to send messages to.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 16,
+    },
+    "TELEGRAM_MESSAGE_THREAD_ID": {
+        "title": "Telegram Thread ID",
+        "description": "Telegram topic/thread ID for group messages (optional).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 17,
+    },
+    # ------------------------------------------------------------------
+    # Notification – Email
+    # ------------------------------------------------------------------
+    "EMAIL_SENDER": {
+        "title": "Email Sender",
+        "description": "Sender email address (SMTP host auto-detected).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 25,
+    },
+    "EMAIL_PASSWORD": {
+        "title": "Email Password",
+        "description": "Email password or app-specific authorization code.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 26,
+    },
+    "EMAIL_RECEIVERS": {
+        "title": "Email Receivers",
+        "description": "Comma-separated recipient email addresses. Leave empty to send to yourself.",
+        "category": "notification",
+        "data_type": "array",
+        "ui_control": "textarea",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {"multi_value": True, "delimiter": ","},
+        "display_order": 27,
+    },
+    # ------------------------------------------------------------------
+    # Notification – Discord
+    # ------------------------------------------------------------------
+    "DISCORD_WEBHOOK_URL": {
+        "title": "Discord Webhook URL",
+        "description": "Discord webhook URL for channel notifications.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 33,
+    },
+    "DISCORD_BOT_TOKEN": {
+        "title": "Discord Bot Token",
+        "description": "Discord bot token for interactive bot mode.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 34,
+    },
+    "DISCORD_MAIN_CHANNEL_ID": {
+        "title": "Discord Channel ID",
+        "description": "Discord main channel ID for sending messages.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 35,
+    },
+    # ------------------------------------------------------------------
+    # Notification – Pushover
+    # ------------------------------------------------------------------
+    "PUSHOVER_USER_KEY": {
+        "title": "Pushover User Key",
+        "description": "Pushover user key (from https://pushover.net).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 42,
+    },
+    "PUSHOVER_API_TOKEN": {
+        "title": "Pushover API Token",
+        "description": "Pushover application API token.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 43,
+    },
+    "PUSHPLUS_TOPIC": {
+        "title": "PushPlus Topic",
+        "description": "PushPlus group topic code for one-to-many push.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "text",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 41,
+    },
+    # ------------------------------------------------------------------
+    # Notification – Server酱 / misc
+    # ------------------------------------------------------------------
+    "SERVERCHAN3_SENDKEY": {
+        "title": "ServerChan3 SendKey",
+        "description": "Server酱3 SendKey for push notifications.",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "password",
+        "is_sensitive": True,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": None,
+        "options": [],
+        "validation": {},
+        "display_order": 45,
+    },
+    "SINGLE_STOCK_NOTIFY": {
+        "title": "Single Stock Notify",
+        "description": "Push immediately after each single stock analysis instead of batching all results together.",
+        "category": "notification",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "false",
+        "options": [],
+        "validation": {},
+        "display_order": 54,
+    },
+    "REPORT_TYPE": {
+        "title": "Report Type",
+        "description": "Report format: 'simple' (concise) or 'full' (detailed).",
+        "category": "notification",
+        "data_type": "string",
+        "ui_control": "select",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "simple",
+        "options": ["simple", "full"],
+        "validation": {"enum": ["simple", "full"]},
+        "display_order": 55,
+    },
+    "MERGE_EMAIL_NOTIFICATION": {
+        "title": "Merge Email Notification",
+        "description": "Merge stock analysis and market review into a single email notification.",
+        "category": "notification",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "false",
+        "options": [],
+        "validation": {},
+        "display_order": 56,
+    },
     "SCHEDULE_TIME": {
         "title": "Schedule Time",
         "description": "Daily schedule time in HH:MM format.",
@@ -540,6 +1071,118 @@ _FIELD_DEFINITIONS: Dict[str, Dict[str, Any]] = {
         "validation": {},
         "display_order": 45,
     },
+    "SCHEDULE_ENABLED": {
+        "title": "Schedule Enabled",
+        "description": "Enable daily scheduled analysis run.",
+        "category": "system",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "false",
+        "options": [],
+        "validation": {},
+        "display_order": 8,
+    },
+    "SCHEDULE_RUN_IMMEDIATELY": {
+        "title": "Schedule Run Immediately",
+        "description": "Whether to run one analysis immediately on startup in schedule mode.",
+        "category": "system",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "true",
+        "options": [],
+        "validation": {},
+        "display_order": 11,
+    },
+    "TRADING_DAY_CHECK_ENABLED": {
+        "title": "Trading Day Check",
+        "description": "Skip analysis on non-trading days. Set to false or use --force-run to override.",
+        "category": "system",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "true",
+        "options": [],
+        "validation": {},
+        "display_order": 12,
+    },
+    "MARKET_REVIEW_ENABLED": {
+        "title": "Market Review Enabled",
+        "description": "Enable market overview/review in analysis reports.",
+        "category": "system",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "true",
+        "options": [],
+        "validation": {},
+        "display_order": 46,
+    },
+    "MARKET_REVIEW_REGION": {
+        "title": "Market Review Region",
+        "description": "Market region for review: cn (A-shares), us (US stocks), or both.",
+        "category": "system",
+        "data_type": "string",
+        "ui_control": "select",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "cn",
+        "options": ["cn", "us", "both"],
+        "validation": {"enum": ["cn", "us", "both"]},
+        "display_order": 47,
+    },
+    "MAX_WORKERS": {
+        "title": "Max Workers",
+        "description": "Maximum concurrent analysis threads. Keep low to avoid API rate limits.",
+        "category": "system",
+        "data_type": "integer",
+        "ui_control": "number",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "3",
+        "options": [],
+        "validation": {"min": 1, "max": 20},
+        "display_order": 50,
+    },
+    "ANALYSIS_DELAY": {
+        "title": "Analysis Delay",
+        "description": "Delay in seconds between individual stock analyses (for API rate limiting).",
+        "category": "system",
+        "data_type": "number",
+        "ui_control": "number",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "0",
+        "options": [],
+        "validation": {"min": 0, "max": 60},
+        "display_order": 51,
+    },
+    "DEBUG": {
+        "title": "Debug Mode",
+        "description": "Enable debug mode with verbose logging.",
+        "category": "system",
+        "data_type": "boolean",
+        "ui_control": "switch",
+        "is_sensitive": False,
+        "is_required": False,
+        "is_editable": True,
+        "default_value": "false",
+        "options": [],
+        "validation": {},
+        "display_order": 55,
+    },
     "BACKTEST_ENABLED": {
         "title": "Backtest Enabled",
         "description": "Whether backtest is enabled.",
@@ -740,7 +1383,7 @@ def _infer_category(key: str) -> str:
         return "base"
     if key.startswith("BACKTEST_"):
         return "backtest"
-    if key.startswith(("GEMINI_", "OPENAI_", "ANTHROPIC_")):
+    if key.startswith(("GEMINI_", "OPENAI_", "ANTHROPIC_", "LITELLM_", "AIHUBMIX_", "DEEPSEEK_", "LLM_")):
         return "ai_model"
     if key.endswith("_PRIORITY") or key.startswith(
         (
@@ -753,8 +1396,11 @@ def _infer_category(key: str) -> str:
             "TAVILY",
             "SERPAPI",
             "BRAVE",
+            "BOCHA",
+            "NEWS_",
+            "BIAS_",
         )
-    ):
+    ) or key in ("ENABLE_REALTIME_QUOTE", "ENABLE_CHIP_DISTRIBUTION"):
         return "data_source"
     if key.startswith((
         "WECHAT",
@@ -771,7 +1417,7 @@ def _infer_category(key: str) -> str:
         "ASTRBOT",
     )) or "WEBHOOK" in key:
         return "notification"
-    if key.startswith(("LOG_", "SCHEDULE_", "WEBUI_", "HTTP_", "HTTPS_", "MAX_", "DEBUG")):
+    if key.startswith(("LOG_", "SCHEDULE_", "WEBUI_", "HTTP_", "HTTPS_", "MAX_", "DEBUG", "MARKET_REVIEW_", "TRADING_DAY_", "ANALYSIS_DELAY")):
         return "system"
     return "uncategorized"