fix max token error

app/gemini_client.py

@@ -2,8 +2,8 @@ from google.generativeai.embedding import embed_content
 from google.generativeai.client import configure
 from google.generativeai.generative_models import GenerativeModel
 from loguru import logger
-from typing import List, Optional
-from google.generativeai.types import GenerationConfig
+from typing import Dict, List, Optional
+from google.generativeai.types import GenerationConfig, HarmCategory, HarmBlockThreshold
 
 from .request_limit_manager import RequestLimitManager
 from .utils import (
@@ -27,6 +27,13 @@ class GeminiClient:
         self._cached_model = None
         self._cached_key = None
         self._cached_model_instance = None
+        # Thêm cấu hình safety_settings để bỏ chặn các phản hồi bị coi là không an toàn
+        self.safety_settings: Dict[HarmCategory, HarmBlockThreshold] = {
+            HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
+            HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE,
+            HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
+            HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
+        }
 
     def _get_model_instance(self, key: str, model: str):
         """
@@ -68,7 +75,11 @@ class GeminiClient:
                 # Sử dụng cached model instance
                 _model = self._get_model_instance(key, model)
                 
-                response = _model.generate_content(prompt, **kwargs)
+                response = _model.generate_content(
+                    prompt,
+                    safety_settings=self.safety_settings,
+                    **kwargs
+                )
                 
                 # Log toàn bộ nội dung response ở mức INFO để tiện gỡ lỗi
                 logger.info(f"[GEMINI][RAW_RESPONSE] {response}")