Update actions/actions.py

actions/actions.py

@@ -32,13 +32,14 @@ openai.api_key = secret_value_0
 # Provide your OpenAI API key
 
 #model_engine="text-davinci-002"
-def generate_openai_response(conversation_data, model_engine="gpt-3.5-turbo", max_tokens=256, temperature=0.5):
+def generate_openai_response(user_queries, model_engine="gpt-3.5-turbo", max_tokens=256, temperature=0.5):
     """Generate a response using the OpenAI API."""
     
-    # Run the main function from search_content.py and store the results in a variable
+
     
+    # Run the main function from search_content.py and store the results in a variable
     #results = main_search(query)
-    results = main_search(conversation_data["current_user_query"])
+    results = main_search(user_queries[-1])
 
     # Create context from the results
     context = "".join([f"#{str(i)}" for i in results])[:2014] # Trim the context to 2014 characters - Modify as necessory
@@ -52,8 +53,8 @@ def generate_openai_response(conversation_data, model_engine="gpt-3.5-turbo", ma
     #prompt_template = f"Using Relevant context:{context}\n\n and Previous User Query: {previous_user_query}\n\n Answer the next question in detail:{current_user_query}"
     messages=[
     {"role": "system", "content": f"You are Omdi, a helpful assistant answers Omdena questions Using Relevant context:{context}"},
-    {"role": "user", "content": conversation_data["previous_user_query"]},
-    {"role": "user", "content": conversation_data["current_user_query"]}
+    {"role": "user", "content": user_queries[-2]},
+    {"role": "user", "content": user_queries[-1]}
   ]
     # Generate a response using the OpenAI API
     response = openai.ChatCompletion.create(
@@ -78,11 +79,21 @@ class GetOpenAIResponse(Action):
             tracker: Tracker,
             domain: Dict[Text, Any]) -> List[Dict[Text, Any]]:
 
+        # Extract conversation data
+        conversation_history = tracker.events
+
+        user_queries = []
+        bot_responses = []
+
+        for event in conversation_history:
+            if event.get("event") == "user":
+                user_queries.append(event.get("text"))
+            elif event.get("event") == "bot":
+                bot_responses.append(event.get("text"))
+
         # Use OpenAI API to generate a response
         #query = tracker.latest_message.get('text')
-        conversation_data = extract_conversation_history()
-        #response = generate_openai_response(conversation_data[0])
-        response = conversation_data
+        response = generate_openai_response(user_queries)
                 
         # Output the generated response to user
         dispatcher.utter_message(text=str(response))
@@ -228,46 +239,3 @@ class SayHelloWorld(Action):
         # Output the generated response to user
         generated_text = response.choices[0].text
         dispatcher.utter_message(text=generated_text)
-
-def extract_conversation_history(dispatcher: CollectingDispatcher, tracker: Tracker, domain: Dict[Text, Any]) -> Dict[Text, Any]:
-    conversation_history = tracker.events
-
-    user_queries = []
-    bot_responses = []
-    current_user_query = ""
-    previous_user_query = None
-    previous_bot_response = None
-
-    for event in conversation_history:
-        if event.get("event") == "user":
-            user_queries.append(event.get("text"))
-        elif event.get("event") == "bot":
-            bot_responses.append(event.get("text"))
-
-    if user_queries:
-        if len(user_queries) >= 2:
-            previous_user_query = user_queries[-2]
-        else:
-            pass
-
-        try:
-            current_user_query = user_queries[-1]
-        except:
-            pass
-
-    if bot_responses:
-        if len(bot_responses) >= 2:
-            previous_bot_response = bot_responses[-2]
-        else:
-            pass
-    else:
-        pass
-        
-    conversation_data = {
-        "previous_user_query": previous_user_query,
-        "previous_bot_response": previous_bot_response,
-        "current_user_query": current_user_query
-    }
-    
-    # Now you can use the conversation_data dictionary as needed.
-    return conversation_data