Miro en

backend/app/api/graph.py

@@ -213,12 +213,19 @@ def generate_ontology():
 
         # 生成本体
         logger.info("调用 LLM 生成本体定义...")
+        # till this point
         generator = OntologyGenerator()
-        ontology = generator.generate(
-            document_texts=document_texts,
-            simulation_requirement=simulation_requirement,
-            additional_context=additional_context if additional_context else None
-        )
+        logger.info("OntologyGenerator....................")
+        try:
+            ontology = generator.generate(
+                document_texts=document_texts,
+                simulation_requirement=simulation_requirement,
+                additional_context=additional_context if additional_context else None
+            )
+            logger.info("OntologyGenerator  try....................")
+        except Exception as e:
+            logger.error(f"本体生成失败: {traceback.format_exc()}")  # ADD
+            raise 
 
         # 保存本体到项目
         entity_count = len(ontology.get("entity_types", []))

backend/app/config.py

@@ -13,7 +13,6 @@
 if os.path.exists(project_root_env):
     load_dotenv(project_root_env, override=True)
 else:
-    # 如果根目录没有 .env，尝试加载环境变量（用于生产环境）
     load_dotenv(override=True)
 
 
@@ -24,14 +23,27 @@ class Config:
     SECRET_KEY = os.environ.get('SECRET_KEY', 'mirofish-secret-key')
     DEBUG = os.environ.get('FLASK_DEBUG', 'True').lower() == 'true'
 
-    # JSON配置 - 禁用ASCII转义，让中文直接显示（而不是 \uXXXX 格式）
+    # JSON配置
     JSON_AS_ASCII = False
 
-    # LLM配置（统一使用OpenAI格式）
-    LLM_API_KEY = os.environ.get('LLM_API_KEY')
-    LLM_BASE_URL = os.environ.get('LLM_BASE_URL', 'https://api.openai.com/v1')
-    LLM_MODEL_NAME = os.environ.get('LLM_MODEL_NAME', 'gpt-4o-mini')
-
+    # ===== Primary LLM (Groq) - 3 keys, auto-rotates on rate limit =====
+    LLM_API_KEY_1 = os.environ.get('LLM_API_KEY_1')
+    LLM_API_KEY_2 = os.environ.get('LLM_API_KEY_2')
+    LLM_API_KEY_3 = os.environ.get('LLM_API_KEY_3')
+    LLM_API_KEY_4 = os.environ.get('LLM_API_KEY_4')
+    LLM_BASE_URL = os.environ.get('LLM_BASE_URL', 'https://api.groq.com/openai/v1')
+    LLM_MODEL_NAME = os.environ.get('LLM_MODEL_NAME', 'meta-llama/llama-4-scout-17b-16e-instruct')
+
+    # Backward compatibility alias
+    LLM_API_KEY = os.environ.get('LLM_API_KEY_1') or os.environ.get('LLM_API_KEY')
+
+    # ===== Boost LLM (Gemini) - 3 keys, auto-rotates on rate limit =====
+    LLM_BOOST_API_KEY_1 = os.environ.get('LLM_BOOST_API_KEY_1')
+    LLM_BOOST_API_KEY_2 = os.environ.get('LLM_BOOST_API_KEY_2')
+    LLM_BOOST_API_KEY_3 = os.environ.get('LLM_BOOST_API_KEY_3')
+    LLM_BOOST_BASE_URL = os.environ.get('LLM_BOOST_BASE_URL')
+    LLM_BOOST_MODEL_NAME = os.environ.get('LLM_BOOST_MODEL_NAME')
+
     # Zep配置
     ZEP_API_KEY = os.environ.get('ZEP_API_KEY')
 
@@ -41,11 +53,11 @@ class Config:
     ALLOWED_EXTENSIONS = {'pdf', 'md', 'txt', 'markdown'}
 
     # 文本处理配置
-    DEFAULT_CHUNK_SIZE = 500  # 默认切块大小
-    DEFAULT_CHUNK_OVERLAP = 50  # 默认重叠大小
+    DEFAULT_CHUNK_SIZE = 500
+    DEFAULT_CHUNK_OVERLAP = 50
 
     # OASIS模拟配置
-    OASIS_DEFAULT_MAX_ROUNDS = int(os.environ.get('OASIS_DEFAULT_MAX_ROUNDS', '10'))
+    OASIS_DEFAULT_MAX_ROUNDS = int(os.environ.get('OASIS_DEFAULT_MAX_ROUNDS', '3'))
     OASIS_SIMULATION_DATA_DIR = os.path.join(os.path.dirname(__file__), '../uploads/simulations')
 
     # OASIS平台可用动作配置
@@ -59,17 +71,38 @@ class Config:
     ]
 
     # Report Agent配置
-    REPORT_AGENT_MAX_TOOL_CALLS = int(os.environ.get('REPORT_AGENT_MAX_TOOL_CALLS', '5'))
-    REPORT_AGENT_MAX_REFLECTION_ROUNDS = int(os.environ.get('REPORT_AGENT_MAX_REFLECTION_ROUNDS', '2'))
-    REPORT_AGENT_TEMPERATURE = float(os.environ.get('REPORT_AGENT_TEMPERATURE', '0.5'))
-
+    REPORT_AGENT_MAX_TOOL_CALLS = int(os.environ.get('REPORT_AGENT_MAX_TOOL_CALLS', '1'))
+    REPORT_AGENT_MAX_REFLECTION_ROUNDS = int(os.environ.get('REPORT_AGENT_MAX_REFLECTION_ROUNDS', '1'))
+    REPORT_AGENT_TEMPERATURE = float(os.environ.get('REPORT_AGENT_TEMPERATURE', '0.3'))
+
+    @classmethod
+    def get_primary_keys(cls) -> list:
+        """Return all configured primary (Groq) API keys"""
+        return [k for k in [cls.LLM_API_KEY_1, cls.LLM_API_KEY_2, cls.LLM_API_KEY_3, cls.LLM_API_KEY_4] if k]
+
+    @classmethod
+    def get_boost_keys(cls) -> list:
+        """Return all configured boost (Gemini) API keys"""
+        return [k for k in [cls.LLM_BOOST_API_KEY_1, cls.LLM_BOOST_API_KEY_2, cls.LLM_BOOST_API_KEY_3] if k]
+
+    @classmethod
+    def has_boost(cls) -> bool:
+        """Check if boost LLM is configured"""
+        return bool(cls.LLM_BOOST_API_KEY_1 and cls.LLM_BOOST_BASE_URL and cls.LLM_BOOST_MODEL_NAME)
+
     @classmethod
     def validate(cls):
         """验证必要配置"""
         errors = []
-        if not cls.LLM_API_KEY:
-            errors.append("LLM_API_KEY 未配置")
+        primary_keys = cls.get_primary_keys()
+        if not primary_keys:
+            errors.append("LLM_API_KEY_1 未配置")
+        else:
+            print(f"✅ Primary LLM: {len(primary_keys)} key(s) configured")
         if not cls.ZEP_API_KEY:
             errors.append("ZEP_API_KEY 未配置")
-        return errors
-
+        if cls.has_boost():
+            print(f"✅ Boost LLM: {len(cls.get_boost_keys())} key(s) configured")
+        else:
+            print("⚠️ LLM_BOOST not configured - all tasks will use primary LLM")
+        return errors