Feat/agent thinking switch (#15446)

### What problem does this PR solve?

This PR adds an Agent LLM setting to control thinking mode for official
providers that expose a thinking switch.

Related to #12842.  
Closes #15445.

Some providers expose thinking controls through provider-specific
request fields, but Agent LLM settings did not have a unified option for
users to enable or disable thinking mode.

This PR adds a `Thinking` selector with:

- System default
- Enabled
- Disabled
<img width="452" height="278" alt="8566b0b4-0546-4c8a-913d-f9bbd38319f6"
src="https://github.com/user-attachments/assets/25b497f7-1ba0-4bfe-940d-6fe79287d6ab"
/>
<img width="471" height="971" alt="8a0a6bee-f45f-48d5-bd83-17af260de3db"
src="https://github.com/user-attachments/assets/41ad43c1-5087-48f1-bf37-f2ca14c2be2f"
/>
Initial support is limited to the verified official providers:

- Qwen / DashScope: `enable_thinking`
- Kimi / Moonshot: `thinking.type`
- GLM / ZHIPU-AI: `thinking.type`

For LiteLLM-based providers, provider-specific fields are forwarded
through `extra_body` before `drop_params` filtering so the request
parameters are preserved.



### Type of change

- [x] New Feature (non-breaking change which adds functionality)

---------

Co-authored-by: jiashi <jiashi19@outlook.com>
Co-authored-by: Zhichang Yu <yuzhichang@gmail.com>

web/src/components/llm-setting-items/next.tsx

@@ -38,6 +38,7 @@ interface LlmSettingFieldItemsProps {
     | 'presence_penalty'
     | 'frequency_penalty'
     | 'max_tokens'
+    | 'thinking'
   >;
   showCollapse?: boolean;
 }
@@ -61,6 +62,7 @@ export const LlmSettingFieldSchema = {
   frequency_penalty: z.coerce.number().optional(),
   max_tokens: z.number().optional(),
   parameter: z.string().optional(),
+  thinking: z.enum(['default', 'enabled', 'disabled']).optional(),
 };
 
 export const LlmSettingSchema = {
@@ -80,6 +82,7 @@ export function LlmSettingFieldItems({
     'presence_penalty',
     'frequency_penalty',
     'max_tokens',
+    'thinking',
   ],
   llmId,
   showCollapse = false,
@@ -249,6 +252,41 @@ export function LlmSettingFieldItems({
               }}
             ></SliderInputSwitchFormField>
           )}
+          {showFields.some((item) => item === 'thinking') && (
+            <FormField
+              control={form.control}
+              name={getFieldWithPrefix('thinking')}
+              render={({ field }) => (
+                <FormItem className="flex justify-between items-center">
+                  <FormLabel className="flex-1" tooltip={t('thinkingTip')}>
+                    {t('thinking')}
+                  </FormLabel>
+                  <FormControl>
+                    <Select
+                      value={field.value ?? 'default'}
+                      onValueChange={field.onChange}
+                    >
+                      <SelectTrigger className="flex-1 !m-0">
+                        <SelectValue />
+                      </SelectTrigger>
+                      <SelectContent>
+                        <SelectItem value="default">
+                          {t('thinkingDefault')}
+                        </SelectItem>
+                        <SelectItem value="enabled">
+                          {t('thinkingEnabled')}
+                        </SelectItem>
+                        <SelectItem value="disabled">
+                          {t('thinkingDisabled')}
+                        </SelectItem>
+                      </SelectContent>
+                    </Select>
+                  </FormControl>
+                  <FormMessage />
+                </FormItem>
+              )}
+            />
+          )}
         </section>
       </CollapseComponent>
     </div>

web/src/locales/en.ts

@@ -1056,6 +1056,12 @@ This auto-tagging feature enhances retrieval by adding another layer of domain-s
       maxTokensTip: `The maximum context size of the model; an invalid or incorrect value will cause an error. Defaults to 512.`,
       maxTokensInvalidMessage: 'Please enter a valid number for Max tokens.',
       maxTokensMinMessage: 'Max tokens cannot be less than 0.',
+      thinking: 'Thinking',
+      thinkingDefault: 'System default',
+      thinkingEnabled: 'Enabled',
+      thinkingDisabled: 'Disabled',
+      thinkingTip:
+        'Only controls thinking mode for official Qwen, Kimi, and GLM model providers. System default disables Qwen thinking to avoid long-running tasks.',
       quote: 'Show quote',
       quoteTip: 'Whether to display the original text as a reference.',
       selfRag: 'Self-RAG',

web/src/locales/zh-traditional.ts

@@ -506,6 +506,11 @@ export default {
       maxTokensTip: `模型的最大上下文大小；無效或不正確的值會導致錯誤。預設為 512。`,
       maxTokensInvalidMessage: '請輸入有效的最大標記數。',
       maxTokensMinMessage: '最大標記數不能小於 0。',
+      thinking: '思考',
+      thinkingDefault: '系統預設',
+      thinkingEnabled: '開啟',
+      thinkingDisabled: '關閉',
+      thinkingTip: '僅控制官方模型提供商中的 Qwen、Kimi 和 GLM 模型思考模式。系統預設會關閉 Qwen 思考，以避免任務長時間執行。',
       quote: '顯示引文',
       quoteTip: '是否應該顯示原文出處？',
       selfRag: 'Self-RAG',

web/src/locales/zh.ts

@@ -957,6 +957,11 @@ NER：使用 spaCy NER 和基于规则的关键词提取来抽取实体和关系
       maxTokensTip: `模型的最大上下文大小；无效或不正确的值会导致错误。默认值为 512。`,
       maxTokensInvalidMessage: '请输入有效的最大令牌数。',
       maxTokensMinMessage: '最大令牌数不能小于 0。',
+      thinking: '思考',
+      thinkingDefault: '系统默认',
+      thinkingEnabled: '开启',
+      thinkingDisabled: '关闭',
+      thinkingTip: '仅控制官方模型提供商中的 Qwen、Kimi 和 GLM 模型思考模式。系统默认会关闭 Qwen 思考，以避免任务长时间运行。',
       quote: '显示引文',
       quoteTip: '是否应该显示原文出处？',
       selfRag: 'Self-RAG',