fix: clean junk text from Antaf + use int8 TTS model for speed

- Filter "完成资料引用" and other status text from Antaf responses
- Use int8 quantized model for faster TTS inference
- Add configurable num_threads for sherpa-onnx

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/main/xiaozhi-server/core/providers/llm/antaf/antaf.py

@@ -37,6 +37,17 @@ class LLMProvider(LLMProviderBase):
                 return True
         return False
 
+    @staticmethod
+    def _clean_text(text):
+        """清理阿福返回文本中的脏数据"""
+        # 去掉阿福内部状态文本
+        junk = [
+            "完成资料引用", "内容生成", "正在思考", "正在搜索",
+        ]
+        for j in junk:
+            text = text.replace(j, "")
+        return text.strip()
+
     @staticmethod
     def _is_system_injected(content):
         """检测是否为系统注入的消息（非用户真实输入）"""
@@ -108,6 +119,10 @@ class LLMProvider(LLMProviderBase):
                     if self._is_thinking(data):
                         logger.bind(tag=TAG).debug(f"过滤思考内容: {data[:50]}...")
                         continue
+                    # 清理脏数据
+                    data = self._clean_text(data)
+                    if not data:
+                        continue
                     yield data
 
         except requests.exceptions.ConnectionError:

backend/main/xiaozhi-server/core/providers/tts/sherpa_tts.py

@@ -17,14 +17,23 @@ class TTSProvider(TTSProviderBase):
         self.speed = float(speed) if speed else 1.0
         self.sid = int(config.get("sid", 0))
 
+        # 优先使用 int8 量化模型（更快）
+        model_file = f"{model_dir}/model.int8.onnx"
+        import os
+        if not os.path.exists(model_file):
+            model_file = f"{model_dir}/model.onnx"
+
+        num_threads = int(config.get("num_threads", 4))
+
         tts_config = sherpa_onnx.OfflineTtsConfig(
             model=sherpa_onnx.OfflineTtsModelConfig(
                 vits=sherpa_onnx.OfflineTtsVitsModelConfig(
-                    model=f"{model_dir}/model.onnx",
+                    model=model_file,
                     lexicon=f"{model_dir}/lexicon.txt",
                     tokens=f"{model_dir}/tokens.txt",
                     dict_dir=f"{model_dir}/dict",
                 ),
+                num_threads=num_threads,
             ),
             rule_fsts=f"{model_dir}/date.fst,{model_dir}/phone.fst,{model_dir}/number.fst,{model_dir}/new_heteronym.fst",
             max_num_sentences=1,