update response readme

READMEresponses.md

@@ -93,6 +93,10 @@ At present, our application focuses on **structured comparisons** between protoc
 While the current system is **not agentic**, we may explore **agent-based reasoning** in future versions to dynamically adjust retrieval and processing strategies based on protocol complexity.  At this time all the information we need is in the provided local documents.
 
 
+### Agentic reasoning POC
+There is a separate poc in the app_working_on_agentic.py file.  That utilizes the agentic approach. 
+It has an agent executor that utilizes agentic reasoning in this application by leveraging OpenAI's function calling to dynamically choose and execute the most appropriate tool based on the user's input. The agent is guided by a predefined system prompt that outlines when to use each tool, allowing it to interpret the user's intent and invoke either the document_query_tool for retrieving document content or the document_comparison_tool for comparing documents. This process enables the AI to act autonomously, selecting and executing the correct function without needing explicit conditional logic.
+
 
 ## Task 4: Building a Quick End-to-End Prototype
 
@@ -160,10 +164,12 @@ The base model is the best performing model for the current use case.
 
 In the second half of the course I would like to explore having the application look for external standards to compare the protocols to further help with the comparison process.  I would also like it to evaluate the file type and use reasoning to determine the best approach to extracting the information and potentially accept more than 2 files.
 
-There is a partially working agentic version of the application in app_working_on_agentic.py.  It has issues with the download links provided but works well with the agentic approach. Unfortunately the download link it provies is not workign even though the file get created properly.
+There is a partially working agentic version of the application in app_working_on_agentic.py.  It has issues with the download links provided but works well with the agentic approach. Unfortunately the download link it provides is not working even though the file gets created properly. (https://github.com/drewgenai/midterm_poc/blob/main/app_working_on_agentic.py)  
 
 ## Final Submission
 
-1. GitHub: https://github.com/drewgenai/midterm_poc/blob/main/app.py
-2. Public App link: https://huggingface.co/spaces/drewgenai/midterm_poc
-3. Public Fine-tuned embeddings: https://huggingface.co/drewgenai/midterm-compare-arctic-embed-m-ft
+1. GitHub: https://github.com/drewgenai/midterm_poc/blob/main/app.py  
+2. GitHub: agentic poc https://github.com/drewgenai/midterm_poc/blob/main/app_working_on_agentic.py  
+2. Public App link: https://huggingface.co/spaces/drewgenai/midterm_poc  
+3. Public Fine-tuned embeddings: https://huggingface.co/drewgenai/midterm-compare-arctic-embed-m-ft  
+4. loom link: https://www.loom.com/share/084d6c165917486097bcaea7deb12e88?sid=a5cc196f-76f1-4e18-bb92-ee61018f0b7e  

app.py

@@ -126,7 +126,7 @@ def document_query_tool(question: str) -> str:
 
 @tool
 def document_comparison_tool(question: str) -> str:
-    """Compares the two uploaded documents, identifies matched elements, exports them as JSON, formats into CSV, and provides a download link."""
+    """Compares documents, identifies matched elements, exports them as JSON, formats into CSV, and provides a download link."""
 
     # Retrieve the vector database retriever
     retriever = cl.user_session.get("qdrant_retriever")

app_working_on_agentic.py

@@ -154,10 +154,7 @@ def document_comparison_tool(question: str) -> str:
         df = pd.DataFrame(structured_data, columns=["Derived Description", "Protocol_1", "Protocol_2"])
         df.to_csv(file_path, index=False)
 
-        # Store the file path in the user session for later retrieval
-        cl.user_session.set("comparison_file_path", file_path)
-        
-        return "Comparison complete! CSV file has been generated."
+        return file_path  # Return path to the CSV file
 
     except json.JSONDecodeError:
         return "Error: Response is not valid JSON."
@@ -262,43 +259,22 @@ async def handle_message(message: cl.Message):
         )
     
     # Handle the response based on the tool that was called
-    output = response["output"]
-    
-    if isinstance(output, dict) and "answer" in output:
+    if isinstance(response["output"], dict) and "answer" in response["output"]:
         # This is from document_query_tool
-        await cl.Message(output["answer"]).send()
-        
-    elif "Comparison complete!" in str(output):
-        # This is from document_comparison_tool
-        file_path = cl.user_session.get("comparison_file_path")
-        
-        if file_path and os.path.exists(file_path):
-            # Read the file content
-            with open(file_path, "rb") as f:
-                file_content = f.read()
-            
-            # Create a File element with the content
-            file_element = cl.File(
-                name="comparison_results.csv",
-                content=file_content,
-                display="inline"
-            )
-            
-            # Send the message with the file element
-            await cl.Message(
-                content="Comparison complete! Download the CSV below:",
-                elements=[file_element],
-            ).send()
-        else:
-            await cl.Message(content=str(output)).send()
-        
+        await cl.Message(response["output"]["answer"]).send()
+    elif isinstance(response["output"], str) and response["output"].endswith(".csv"):
+        # This is from document_comparison_tool with a CSV file
+        await cl.Message(
+            content="Comparison complete! Download the CSV below:",
+            elements=[cl.File(name="comparison_results.csv", path=response["output"], display="inline")],
+        ).send()
     else:
         # Generic response
-        await cl.Message(content=str(output)).send()
+        await cl.Message(content=str(response["output"])).send()
     
     # Update chat history with the new exchange
     chat_history.extend([
         HumanMessage(content=message.content),
-        HumanMessage(content=str(output))
+        HumanMessage(content=str(response["output"]))
     ])
     cl.user_session.set("chat_history", chat_history)