Update app.py

app.py

@@ -18,13 +18,73 @@ class BasicAgent:
         print("BasicAgent initialized.")
         self.graph = simple_graph()
         
-    def __call__(self, question: str) -> str:
-        time.sleep(30) # To keep using the free gemini key
-        print(f"Agent received question (first 200 chars): {question[:200]}...")
-        messages = [HumanMessage(content=question)]
-        messages = self.graph.invoke({"messages": messages})
-        answer = messages['messages'][-1].content
-        return answer.split("FINAL ANSWER:")[-1].strip()
+    def __call__(self, question: str, file_bytes: bytes = None, file_type: str = None) -> str:
+        print(f"Agent received question (first 50 chars): {question[:50]}...")
+        
+        file_info = ""
+        if file_bytes and file_type:
+            file_info = self.process_file(file_bytes, file_type)
+
+        full_prompt = question
+        if file_info:
+            full_prompt += f"\n\n[File Content Below]\n{file_info}"
+            
+        if file_type == "png":
+            try:
+                encoded_image = base64.b64encode(file_bytes).decode("utf-8")
+
+                messages = [HumanMessage(content=[
+                    {"type": "text", "text": question},
+                    {"type": "image_url", "image_url": f"data:image/png;base64,{encoded_image}"}
+                ])]
+
+                response = self.final_agent.invoke({"messages": messages})
+                answer = response['messages'][-1].content
+                return answer.split("FINAL ANSWER:")[-1].strip()
+
+            except Exception as e:
+                print(f"Error processing image: {e}")
+                return f"Error: {e}"
+
+        elif file_type == "mp3":
+            try:
+                encoded_audio = base64.b64encode(file_bytes).decode("utf-8")
+                audio_mime_type = "audio/mpeg"
+
+                messages = [HumanMessage(content=[
+                    {"type": "text", "text": question},
+                    {
+                        "type": "media",
+                        "data": encoded_audio,
+                        "mime_type": audio_mime_type,
+                    }
+                ])]
+
+                response = self.final_agent.invoke({"messages": messages})
+                answer = response['messages'][-1].content
+                return answer.split("FINAL ANSWER:")[-1].strip()
+
+            except Exception as e:
+                print(f"Error processing audio: {e}")
+                return f"Error: {e}"
+
+        else:
+            messages = [HumanMessage(content=full_prompt)]
+            messages = self.final_agent.invoke({"messages": messages})
+            answer = messages['messages'][-1].content
+            return answer.split("FINAL ANSWER:")[-1].strip()
+    
+    def process_file(self, file_bytes: bytes, file_type: str) -> str:
+        try:
+            if file_type == "py":
+                return file_bytes.decode("utf-8")
+
+            elif file_type == "xlsx":
+                df = pd.read_excel(BytesIO(file_bytes))
+                return df.to_string()
+
+        except Exception as e:
+            return f"Error processing file: {e}"
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
@@ -86,8 +146,29 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
+        file_bytes = None
+        file_type = None
+        try:
+            file_url = f"{api_url}/files/{task_id}"
+            file_response = requests.get(file_url, timeout=10)
+            if file_response.status_code == 200:
+                content_type = file_response.headers.get("Content-Type", "")
+                file_bytes = file_response.content
+                if "image/png" in content_type:
+                    file_type = "png"
+                elif "audio/mpeg" in content_type:
+                    file_type = "mp3"
+                elif "application/octet-stream" in content_type:
+                    file_type = "xlsx"
+                elif "text/x-python" in content_type or "text/plain" in content_type:
+                    file_type = "py"
+        except Exception as e:
+            print(f"Could not fetch or interpret file for task {task_id}: {e}")
+            
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = agent(question_text, file_bytes=file_bytes, file_type=file_type)
+            if submitted_answer is None:
+                submitted_answer = "No answer returned."
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e: