feat: aumenta max_tokens padrão para 8192 (evita cortes no compound-groq)

app.py

@@ -285,7 +285,7 @@ class ChatRequest(BaseModel):
     web: Optional[bool] = Field(default=False, description="Usar busca web")
     free_mode: Optional[bool] = Field(default=False, description="Modo livre - pula RAG e persona")
     messages: List[Message] = Field(min_length=1, description="Lista de mensagens")
-    max_tokens: Optional[int] = Field(default=1024, ge=1, le=16384, description="Máximo de tokens")
+    max_tokens: Optional[int] = Field(default=8192, ge=1, le=16384, description="Máximo de tokens")
     temperature: Optional[float] = Field(default=0.7, ge=0.0, le=2.0, description="Temperatura de geração")
 
     @field_validator("model")

index.html

@@ -426,8 +426,8 @@
     </div>
 
     <div class="slider-group">
-      <label>Max Tokens <span id="tokens-val">1024</span></label>
-      <input type="range" id="max-tokens" min="256" max="16384" step="256" value="1024"
+      <label>Max Tokens <span id="tokens-val">8192</span></label>
+      <input type="range" id="max-tokens" min="256" max="16384" step="256" value="8192"
              oninput="document.getElementById('tokens-val').textContent=this.value">
     </div>