update

docs/class3/class3.rst

@@ -586,6 +586,8 @@ Sample RAG Chatbot conversation
 
 Suggested sample question ask to the RAG chatbot
 
+.. Note:: 
+   AI response is non-deterministric. It means that give the same input, it can produce different output at different times - no gurantee to be consistent and can vary depending on internal factors within the model, like the order of data processing or random initilization. Hence, sometimes, you may need to ask twice for the language model to give an answer.
 
 .. code-block:: bash
 
@@ -606,11 +608,9 @@ Source of inforamtion or "proprietary data" obtained from the text file store on
 ..  image:: ./_static/class3-34.png
 
 .. attention:: 
-   Generic Small Language Model (SLM) may not be as efficient compare to a Large Language Model (LLM) and may constantly encounter hallucination. You may modify the chunking size and chuking overlap to reduce hallucination. For the purpose of a lab, we are not expecting the model to provides an accurate and intelligent answer.
+   Generic Small Language Model (SLM) may not be as efficient compare to a Large Language Model (LLM) and may constantly encounter hallucination. You may modify the chunking size and chunking overlap to reduce hallucination. For the purpose of a lab, we are not expecting the model to provides an accurate and intelligent answer.
+
 
-.. attention:: 
-   You may occasionally see document identifiers, such as *"<doc id='2'>,"* appear in the response output. This issue can arise for several reasons, such as inadequate post-processing where metadata is not properly cleaned or removed, or during pre-processing when data is tagged with metadata that the model interprets as legitimate text. In this particular lab, the issue is likely due to a combination of factors, including the inference and embedding model's behavior and the use of a CPU for processing.
-
 
 **You have successfully build a GenAI RAG Chatbot**
 

docs/class4/class4.rst

@@ -84,6 +84,13 @@ Try to interact with GenAI RAG Chatbot.
 
 ..  image:: ./_static/class4-9.png
 
+
+.. attention:: 
+   You may occasionally see document identifiers, such as *"<doc id='2'>,"* appear in the response output. This issue can arise for several reasons, such as inadequate post-processing where metadata is not properly cleaned or removed, or during pre-processing when data is tagged with metadata that the model interprets as legitimate text. In this particular lab, the issue is likely due to a combination of factors, including the inference and embedding model's behavior and the use of a CPU for processing. **For the purposes of this class, please ignore if any imperfection of the AI responses.**
+
+   ..  image:: ./_static/class4-10-1.png
+
+
 You successfully integrate GenAI RAG chatbot into your modern arcadia trading platform application.
 
 ..  image:: ./_static/class4-10.png

docs/class5/class5.rst

@@ -8,6 +8,8 @@ Class 5: Secure, Deliver and Optimize GenAI ChatBot
 
 AI services and applications are a subset of modern applications. Securing AI apps requires a holistic, end-to-end approach. **You cannot fully protect AI applications without also securing the underlying web applications and APIs.** AI services are powered by APIs, which serve as the backbone of these systems. Securing APIs is critical to maintaining the integrity and reliability of AI services. Below are the 7 key security controls that are essential for ensuring the overall security of modern web applications, API and AI services.
 
+For the purpose of this class, we will only focus on AI Gateway security control - Runtime Security and Traffic Governance. Please reach out for further deep dive session on other controls.
+
 ..  image:: ./_static/class5-1-1.png
 
 1 - Fundamental about F5 AI Gateway