🐛 fix: fix claude 3.5+ models context max output (#6984)

src/libs/agent-runtime/anthropic/index.test.ts

@@ -123,21 +123,21 @@ describe('LobeAnthropicAI', () => {
           { content: 'You are an awesome greeter', role: 'system' },
           { content: 'Hello', role: 'user' },
         ],
-        model: 'claude-3-haiku-20240307',
+        model: 'claude-3-7-sonnet-20250219',
         temperature: 0,
       });
 
       // Assert
       expect(instance['client'].messages.create).toHaveBeenCalledWith(
         {
-          max_tokens: 4096,
+          max_tokens: 8192,
           messages: [
             {
               content: [{ cache_control: { type: 'ephemeral' }, text: 'Hello', type: 'text' }],
               role: 'user',
             },
           ],
-          model: 'claude-3-haiku-20240307',
+          model: 'claude-3-7-sonnet-20250219',
           stream: true,
           system: [
             {

src/libs/agent-runtime/anthropic/index.ts

@@ -25,6 +25,8 @@ export interface AnthropicModelCard {
   id: string;
 }
 
+const modelsWithSmallContextWindow = new Set(['claude-3-opus-20240229', 'claude-3-haiku-20240307']);
+
 const DEFAULT_BASE_URL = 'https://api.anthropic.com';
 
 interface AnthropicAIParams extends ClientOptions {
@@ -132,7 +134,7 @@ export class LobeAnthropicAI implements LobeRuntimeAI {
     return {
       // claude 3 series model hax max output token of 4096, 3.x series has 8192
       // https://docs.anthropic.com/en/docs/about-claude/models/all-models#:~:text=200K-,Max%20output,-Normal%3A
-      max_tokens: max_tokens ?? (model.startsWith('claude-3-') ? 4096 : 8192),
+      max_tokens: max_tokens ?? (modelsWithSmallContextWindow.has(model) ? 4096 : 8192),
       messages: postMessages,
       model,
       system: systemPrompts,