Update app.py

app.py

@@ -6,7 +6,7 @@ from langchain.agents import initialize_agent, AgentType
 from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain.tools import Tool
 
-# Optional: Tavily client (only if installed)
+# Optional: Tavily client
 try:
     from tavily import TavilyClient
     tavily_available = True
@@ -26,8 +26,7 @@ def get_current_temperature(query: str) -> str:
 
 def search_wikipedia(query: str) -> str:
     try:
-        summary = wikipedia.summary(query, sentences=2)
-        return summary
+        return wikipedia.summary(query, sentences=2)
     except wikipedia.exceptions.DisambiguationError as e:
         return f"Multiple results found: {', '.join(e.options[:5])}"
     except wikipedia.exceptions.PageError:
@@ -51,7 +50,7 @@ class cbfs:
         # Initialize OpenAI model
         self.model = ChatOpenAI(temperature=0, openai_api_key=openai_key)
 
-        # Initialize Tavily if available
+        # Tavily init (optional)
         self.tavily = None
         if tavily_available and tavily_key:
             self.tavily = TavilyClient(api_key=tavily_key)
@@ -69,7 +68,7 @@ class cbfs:
             MessagesPlaceholder(variable_name="agent_scratchpad")
         ])
 
-        # Initialize agent properly with enum
+        # Agent
         self.chain = initialize_agent(
             tools=tools,
             llm=self.model,
@@ -82,25 +81,33 @@ class cbfs:
     def convchain(self, query: str) -> str:
         if not query:
             return "Please enter a query."
+
         try:
             result = self.chain.invoke({"input": query})
+            print("🔍 Raw agent result:", result)  # log to HuggingFace
 
-            # Debugging: show raw result in logs
-            print("🔍 Raw agent result:", result)
-
-            # Try different output keys
+            # Extract possible outputs
             if isinstance(result, dict):
                 response = (
-                    result.get("output")
-                    or result.get("output_text")
-                    or str(result)
+                    result.get("output") or
+                    result.get("output_text") or
+                    str(result)
                 )
             else:
                 response = str(result)
 
+            # Fallback: if still empty, just ask the model directly
+            if not response.strip():
+                print("⚠️ No structured output, falling back to direct LLM call")
+                resp = self.model.invoke(query)
+                response = getattr(resp, "content", str(resp))
+
+            # Save to memory
             self.memory.save_context({"input": query}, {"output": response})
             return response
+
         except Exception as e:
+            print("❌ Execution Error:", str(e))
             return f"❌ Error: {str(e)}"