Update app.py

app.py

@@ -7,7 +7,8 @@ from google.genai import types
 from mcp import ClientSession, StdioServerParameters
 from mcp.client.stdio import stdio_client
 
-# Konfiguration des MCP-Servers für die DB-Fahrplanauskunft [1, 3]
+# 1. Konfiguration des MCP-Servers via STDIO-Bridge
+# Dies löst das SSE-Problem, indem npx mcp-remote die Kommunikation übernimmt [3].
 server_params = StdioServerParameters(
     command="npx",
     args=[
@@ -22,27 +23,27 @@ async def generate(input_text):
     try:
         client = genai.Client(api_key=os.environ.get("GEMINI_API_KEY"))
     except Exception as e:
-        return f"Error initializing client: {e}", ""
+        return f"Fehler bei der Initialisierung: {e}", ""
 
-    model = "gemini-2.0-flash" # Empfohlen für Tool-Calling [6]
-    
-    # Aufbau der Verbindung zum MCP-Server [9, 10]
+    model = "gemini-2.0-flash" 
+
+    # 2. Aufbau der MCP-Session [4, 5]
     async with stdio_client(server_params) as (read, write):
         async with ClientSession(read, write) as session:
             await session.initialize()
             
-            # MCP-Tools abrufen und für Gemini konvertieren [9, 11]
-            mcp_tools_list = await session.list_tools()
+            # MCP-Tools abrufen und in Gemini-Format konvertieren [5, 6]
+            mcp_tools_data = await session.list_tools()
             mcp_declarations = [
                 {
                     "name": tool.name,
-                    "description": tool.description or "Get train connections between stations",
+                    "description": tool.description or "Ruft Zugverbindungen ab.",
                     "parameters": tool.inputSchema,
                 }
-                for tool in mcp_tools_list.tools
+                for tool in mcp_tools_data.tools
             ]
 
-            # Kombinierte Tools: Google Search + MCP Tools [2, 7]
+            # 3. Kombination der Tools: Google Search + DB-Timetable [5, 7]
             tools = [
                 types.Tool(google_search=types.GoogleSearch()),
                 types.Tool(function_declarations=mcp_declarations)
@@ -50,31 +51,29 @@ async def generate(input_text):
 
             contents = [types.Content(role="user", parts=[types.Part.from_text(text=input_text)])]
             
-            # Initialer Aufruf
+            # Erster Aufruf an das Modell
             response = await client.aio.models.generate_content(
                 model=model,
                 contents=contents,
-                config=types.GenerateContentConfig(
-                    tools=tools,
-                    temperature=0.4
-                )
+                config=types.GenerateContentConfig(tools=tools, temperature=0.4)
             )
             
-            contents.append(response.candidates.content)
-            
-            # Tool Calling Loop (für Fahrplandaten oder Google Search) [7, 8]
+            # 4. Agentic Loop: Bearbeitung von Tool-Calls (z.B. db_timetable_api_ui_wrapper) [8, 9]
             turn_count = 0
             while response.function_calls and turn_count < 5:
                 turn_count += 1
+                contents.append(response.candidates.content)
                 tool_responses = []
                 
                 for fc in response.function_calls:
-                    # Ausführung der MCP-Tools (z.B. db_timetable_api_ui_wrapper) [8, 12]
                     try:
+                        # Ausführung des MCP-Tools [9]
                         tool_result = await session.call_tool(fc.name, fc.args)
-                        result_text = tool_result.content.text if not tool_result.isError else tool_result.content.text
+                        
+                        # Ergebnis formatieren (Erfolg oder Fehler) [9]
+                        result_data = tool_result.content.text if not tool_result.isError else tool_result.content.text
                         tool_responses.append(types.Part.from_function_response(
-                            name=fc.name, response={"result": result_text}
+                            name=fc.name, response={"result": result_data}
                         ))
                     except Exception as e:
                         tool_responses.append(types.Part.from_function_response(
@@ -82,22 +81,23 @@ async def generate(input_text):
                         ))
 
                 contents.append(types.Content(role="user", parts=tool_responses))
+                
+                # Nächster Modell-Aufruf mit den Tool-Ergebnissen [10, 11]
                 response = await client.aio.models.generate_content(
                     model=model, contents=contents, config=types.GenerateContentConfig(tools=tools)
                 )
-                contents.append(response.candidates.content)
 
             return response.text, ""
 
-# Gradio UI (Original Funktionalität und Design beibehalten) [13]
+# 5. Gradio UI Integration [12]
 def ui_wrapper(input_text):
     return asyncio.run(generate(input_text))
 
 if __name__ == '__main__':
     with gr.Blocks() as demo:
-        gr.Markdown("# Gemini 2.0 Flash + Websearch + DB Timetable (MCP)")
+        gr.Markdown("# Gemini 2.0 Flash + Search + DB Timetable")
         output_textbox = gr.Markdown()
-        input_textbox = gr.Textbox(lines=3, label="", placeholder="Frage nach einer Zugverbindung (z.B. Berlin nach München)...")
+        input_textbox = gr.Textbox(lines=3, label="Anfrage", placeholder="z.B. Wie komme ich von Berlin nach Hamburg?")
         submit_button = gr.Button("Senden")
         
         submit_button.click(
@@ -109,7 +109,6 @@ if __name__ == '__main__':
 
 
 
-
 """
 import base64
 import gradio as gr