RAG1

app.py

@@ -26,6 +26,7 @@ with gr.Blocks() as demo:
     )
     clear = gr.ClearButton([chat_input, chatbot])
 
+    ################################ AGENTS ###################################
     # Agent1 - RAG Decision Agent (whether RAG is needed for the user's query)
     def agent1_rag_decision(query):
         decision = rag_decision(query)
@@ -36,6 +37,15 @@ with gr.Blocks() as demo:
         results = get_top_k(query, k=k)
         return results
     
+    # Agent3 - LLM Agent (get query response from LLM)
+    def agent3_llm_agent(messages):
+        response = client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=messages,
+            temperature=0.7
+        )
+        return response.choices[0].message.content.strip()
+    ###########################################################################
     def encode_image(image_path):
         with open(image_path, "rb") as f:
             return base64.b64encode(f.read()).decode("utf-8")
@@ -46,9 +56,6 @@ with gr.Blocks() as demo:
         if message["text"] is not None:
             history.append({"role": "user", "content": message["text"]})
         return history, gr.MultimodalTextbox(value=None, interactive=False)
-        
-    # def user(user_message, history: list):
-    #     return "", history + [{"role": "user", "content": user_message}]
     
     def respond(history):
         # print("history", history)
@@ -80,12 +87,27 @@ with gr.Blocks() as demo:
             }
             clean_messages.append(clean_msg)            
         # print("First messages", messages[0])
-        response = client.chat.completions.create(
-                            model="gpt-4o-mini",
-                            messages=clean_messages,
-                            temperature=0.7
-                        )
-        response = response.choices[0].message.content.strip()
+
+        ########################### AGENTIC WORKFLOW ##########################
+        # Call Agent1- the RAG Decision Agent
+        rag_decision = agent1_rag_decision(messages[-1]["content"])
+
+        if rag_decision == True:
+            #Call Agent2 - the RAG Retrieval Agent
+            top_k_results = agent2_use_rag(messages[-1]["content"][0]["text"], k=3)
+            # Append the top k results to the messages
+            for i, result in enumerate(top_k_results):
+                clean_messages.append({
+                    "role": "system",
+                    "content": f"RAG Retrieved Result-{i+1}: " + result["content"]
+                })
+            # Call Agent3 - the LLM Agent to get query response
+            response = agent3_llm_agent(clean_messages)
+        else:
+            # Call Agent3 - the LLM Agent to get query response
+            response = agent3_llm_agent(messages)
+        #######################################################################
+
         # print("response:", response)
         # history.append({"role": "assistant", "content": response})
         history.append({"role": "assistant", "content": ""})