feat: minimize token usage when testing model (#5905)

src/renderer/src/providers/AiProvider/AnthropicProvider.ts

@@ -678,7 +678,7 @@ export default class AnthropicProvider extends BaseProvider {
     const body = {
       model: model.id,
       messages: [{ role: 'user' as const, content: 'hi' }],
-      max_tokens: 100,
+      max_tokens: 2, // api文档写的 x>1
       stream
     }
 

src/renderer/src/providers/AiProvider/GeminiProvider.ts

@@ -916,7 +916,7 @@ export default class GeminiProvider extends BaseProvider {
           model: model.id,
           contents: [{ role: 'user', parts: [{ text: 'hi' }] }],
           config: {
-            maxOutputTokens: 100
+            maxOutputTokens: 1
           }
         })
         if (isEmpty(result.text)) {
@@ -927,7 +927,7 @@ export default class GeminiProvider extends BaseProvider {
           model: model.id,
           contents: [{ role: 'user', parts: [{ text: 'hi' }] }],
           config: {
-            maxOutputTokens: 100
+            maxOutputTokens: 1
           }
         })
         // 等待整个流式响应结束

src/renderer/src/providers/AiProvider/OpenAICompatibleProvider.ts

@@ -1112,6 +1112,9 @@ export default class OpenAICompatibleProvider extends BaseOpenAiProvider {
     const body = {
       model: model.id,
       messages: [{ role: 'user', content: 'hi' }],
+      max_completion_tokens: 1, // openAI
+      max_tokens: 1, // openAI deprecated 但大部分OpenAI兼容的提供商继续用这个头
+      enable_thinking: false, // qwen3
       stream
     }
 

src/renderer/src/providers/AiProvider/OpenAIProvider.ts

@@ -1026,6 +1026,7 @@ export abstract class BaseOpenAiProvider extends BaseProvider {
       const response = await this.sdk.responses.create({
         model: model.id,
         input: [{ role: 'user', content: 'hi' }],
+        max_output_tokens: 1,
         stream: true
       })
       let hasContent = false
@@ -1042,7 +1043,8 @@ export abstract class BaseOpenAiProvider extends BaseProvider {
       const response = await this.sdk.responses.create({
         model: model.id,
         input: [{ role: 'user', content: 'hi' }],
-        stream: false
+        stream: false,
+        max_output_tokens: 1
       })
       if (!response.output_text) {
         throw new Error('Empty response')