i updated the conversation history and fixed some api errors

app.py

@@ -17,6 +17,7 @@ from app.vectorstore import vectorstore
 from app.tts import tts
 from app.utils import remove_stars
 
+
 # Set a custom user agent.
 os.environ["USER_AGENT"] = "my-app/1.0"
 
@@ -30,6 +31,7 @@ class CancerApp:
         # Run an initial search to load vectorstore contents.
         self.vectorstore.search_vectorstore("cancer", 1)
         self.tts = tts()
+        self.conversation_history = []  # Initialize an empty list for history
 
     def text_to_speech(self, message, output_filename, Saved_response=""):
         search_results = self.vectorstore.search_vectorstore(message, 5)
@@ -100,21 +102,53 @@ class CancerApp:
         self.audio_processor.log_conversation(transcription, bot_text=gemini_response)
         return gemini_response, Saved_response
 
-    def text_to_text(self, message, Saved_response=""):
+    # def text_to_text(self, message, Saved_response=""):
+    #     search_results = self.vectorstore.search_vectorstore(message, 5)
+    #     query = Saved_response + message 
+    #     for i, result in enumerate(search_results, 1):
+    #         query += f"\n{i}. {result}"
+        
+    #     try:
+    #         gemini_response = self.llm_processor.call_gemini_llm("gemini-2.5-flash-preview-04-17", query)
+    #         gemini_response = remove_stars(gemini_response)
+    #     except Exception as e:
+    #         print("Gemini error:", e)
+    #         gemini_response = f"Error: {str(e)}"
+    #     Saved_response += "message: " + message + "\n" + "gemini_response: " + gemini_response + "\n"
+    #     self.audio_processor.log_conversation(message, bot_text=gemini_response)
+    #     return gemini_response, Saved_response
+     # 1. Add user's current message to history
+    def text_to_text(self, message): # Removed Saved_response from args, use self.conversation_history
+        # 1. Add user's current message to history
+        self.conversation_history.append({"role": "user", "parts": [message]})
+
+        # 2. Prepare RAG context
         search_results = self.vectorstore.search_vectorstore(message, 5)
-        query = Saved_response + message 
+        rag_context_for_query = ""
         for i, result in enumerate(search_results, 1):
-            query += f"\n{i}. {result}"
-        
+            rag_context_for_query += f"\n{i}. {result}"
+
+        # 3. Call LLM with the full history and current RAG context
         try:
-            gemini_response = self.llm_processor.call_gemini_llm("gemini-2.5-flash-preview-04-17", query)
-            gemini_response = remove_stars(gemini_response)
+            # Pass self.conversation_history to the LLM call
+            # The LLM method will need to be adapted to take history
+            gemini_response_text = self.llm_processor.call_gemini_llm_with_history(
+                model_name="gemini-2.5-flash-preview-04-17", # Or your specific model
+                history=self.conversation_history,
+                rag_context=rag_context_for_query
+            )
+            gemini_response_text = remove_stars(gemini_response_text)
         except Exception as e:
             print("Gemini error:", e)
-            gemini_response = f"Error: {str(e)}"
-        Saved_response += "message: " + message + "\n" + "gemini_response: " + gemini_response + "\n"
-        self.audio_processor.log_conversation(message, bot_text=gemini_response)
-        return gemini_response, Saved_response
+            gemini_response_text = f"Error processing your request: {str(e)}"
+
+        # 4. Add bot's response to history
+        self.conversation_history.append({"role": "model", "parts": [gemini_response_text]})
+
+        # 5. Log (your existing log_conversation is fine for separate logging)
+        self.audio_processor.log_conversation(message, bot_text=gemini_response_text) # Logs current turn
+
+        return gemini_response_text # Return only the latest response
 
 # Create Flask API instance.
 flask_app = Flask(__name__)
@@ -149,13 +183,33 @@ cancer_app_instance = CancerApp(vectordb_path)
 
 
 # Endpoint for text-to-text processing.
+# @flask_app.route('/text_to_text', methods=['POST'])
+# def api_text_to_text():
+#     data = request.get_json()
+#     message = data.get("message", "")
+#     saved_response = data.get("Saved_response", "")
+#     response, saved_response = cancer_app_instance.text_to_text(message, Saved_response=saved_response)
+#     return jsonify({"gemini_response": response, "Saved_response": saved_response})
+
 @flask_app.route('/text_to_text', methods=['POST'])
 def api_text_to_text():
     data = request.get_json()
     message = data.get("message", "")
-    saved_response = data.get("Saved_response", "")
-    response, saved_response = cancer_app_instance.text_to_text(message, Saved_response=saved_response)
-    return jsonify({"gemini_response": response, "Saved_response": saved_response})
+    # Client would send the history, server appends and sends back updated history
+    # For simplicity, let's assume history is managed by cancer_app_instance for now
+    # If you want stateless, client sends history, method uses it, returns new history + response
+
+    if not message:
+        return jsonify({"error": "No message provided"}), 400
+
+    # The text_to_text method now manages its own history via self.conversation_history
+    response_text = cancer_app_instance.text_to_text(message)
+
+    return jsonify({
+        "gemini_response": response_text,
+        "conversation_history": cancer_app_instance.conversation_history # Send back updated history
+    })
+
 
 # Endpoint for text-to-speech processing.
 @flask_app.route('/text_to_speech', methods=['POST'])