fix: handle reasoning_content for thinking mode in OpenAI compatibility layer

- Add reasoning_content field when thinking mode is enabled
- Set reasoning_content to empty string when no thinking content available
- Enable model auto-detection for kimi provider
- Add thinking mode support to costrict provider

packages/@ant/model-provider/src/shared/openaiConvertMessages.ts

@@ -211,21 +211,29 @@ function convertInternalAssistantMessage(
   const content = msg.message.content
 
   if (typeof content === 'string') {
-    return [
-      {
-        role: 'assistant',
-        content,
-      } satisfies ChatCompletionAssistantMessageParam,
-    ]
+    const result: ChatCompletionAssistantMessageParam & { reasoning_content?: string } = {
+      role: 'assistant',
+      content,
+    }
+    // 如果开启了thinking模式，必须提供reasoning_content字段
+    if (preserveReasoning) {
+      // content是字符串，没有reasoning内容，设置为空字符串
+      result.reasoning_content = ' '
+    }
+    return [result]
   }
 
   if (!Array.isArray(content)) {
-    return [
-      {
-        role: 'assistant',
-        content: '',
-      } satisfies ChatCompletionAssistantMessageParam,
-    ]
+    const result: ChatCompletionAssistantMessageParam & { reasoning_content?: string } = {
+      role: 'assistant',
+      content: '',
+    }
+    // 如果开启了thinking模式，必须提供reasoning_content字段
+    if (preserveReasoning) {
+      // content不是数组，没有reasoning内容，设置为空字符串
+      result.reasoning_content = ' '
+    }
+    return [result]
   }
 
   const textParts: string[] = []
@@ -258,11 +266,16 @@ function convertInternalAssistantMessage(
     // Skip redacted_thinking, server_tool_use, etc.
   }
 
-  const result: ChatCompletionAssistantMessageParam = {
+  const result: ChatCompletionAssistantMessageParam & { reasoning_content?: string } = {
     role: 'assistant',
     content: textParts.length > 0 ? textParts.join('\n') : null,
     ...(toolCalls.length > 0 && { tool_calls: toolCalls }),
-    ...(reasoningParts.length > 0 && { reasoning_content: reasoningParts.join('\n') }),
+  }
+
+  // 如果开启了thinking模式，必须提供reasoning_content字段
+  if (preserveReasoning) {
+    // 如果有reasoning内容就用这些内容，没有才设置成空字符串
+    result.reasoning_content = reasoningParts.length > 0 ? reasoningParts.join('\n') : ' '
   }
 
   return [result]

src/costrict/provider/index.ts

@@ -31,6 +31,7 @@ import { createCoStrictFetch } from './fetch.js'
 import { resolveCoStrictModel } from './modelMapping.js'
 import { getCoStrictBaseURL } from './auth.js'
 import { loadCoStrictCredentials } from './credentials.js'
+import { isOpenAIThinkingEnabled } from '../../services/api/openai/requestBody.js'
 
 /**
  * CoStrict 查询路径
@@ -80,9 +81,12 @@ export async function* queryModelCoStrict(
     )
 
     // 5. 转换为 OpenAI 格式
+    // 根据模型名称自动检测是否启用thinking模式
+    const enableThinking = isOpenAIThinkingEnabled(costrictModel)
     const openaiMessages = anthropicMessagesToOpenAI(
       messagesForAPI,
-      systemPrompt
+      systemPrompt,
+      { enableThinking }
     )
     const openaiTools = anthropicToolsToOpenAI(standardTools)
     const openaiToolChoice = anthropicToolChoiceToOpenAI(options.toolChoice)

src/services/api/openai/requestBody.ts

@@ -25,9 +25,9 @@ export function isOpenAIThinkingEnabled(model: string): boolean {
   if (isEnvDefinedFalsy(process.env.OPENAI_ENABLE_THINKING)) return false
   // Explicit enable
   if (isEnvTruthy(process.env.OPENAI_ENABLE_THINKING)) return true
-  // Auto-detect from model name (deepseek-reasoner and DeepSeek-V3.2 support thinking mode)
+  // Auto-detect from model name (deepseek-reasoner, DeepSeek-V3.2, kimi support thinking mode)
   const modelLower = model.toLowerCase()
-  return modelLower.includes('deepseek-reasoner') || modelLower.includes('deepseek-v3.2')
+  return modelLower.includes('deepseek-reasoner') || modelLower.includes('deepseek-v3.2') || modelLower.includes('kimi')
 }
 
 /**