Hey @SachinHatikankan, happy to help! So as this widget is culmination of building your RAG system to help supplement the BBC News dataset, which has a knowledge cutoff of 2023, you can this query field to demonstrate how having this RAG system can give you more current information. From the screenshot you shared, you can already see that without RAG, the LLM’s answer is citing its knowledge base cut off, whereas the answer with RAG provided more current news about AI. I’d suggest playing around with other questions about topics you’re interested in (from 2024 or later) and noting the difference between the two answers generated. Does that make sense?
Thanks and appreciate your response!
I am clear about the with RAG and without RAG responses which are appearing.
While I was testing these I found that keeping the Augmented prompt layout textfield empty or adding it with some text is not making any difference. I wanted to know does keeping Augmented prompt layout textfield empty or adding it with some text should make any difference in the output which appears in text boxes “With RAG” and “Without RAG”
Gotcha! Thanks for clarifying. So the default for the widget’s augmented prompt, which gets fed to the LLM, is defined by the generate_final_prompt function earlier in the notebook, but you can definitely play around and change the instructions within that widget box - just make sure you keep both the {query} and {documents} placeholders.
For example, you could go with something like… You're a top-tier journalist, and your task is to craft a fact-driven article on the following topic: {query}. Here are some key facts to include: {documents} With top_k set to 2.
Or something a bit more fun… You're a late-night comedy news host. Your task is to craft a humorous, sarcastic broadcast on the following topic: {query}. Here are some key facts you can include: {documents}, but it's preferable that you sacrifice truthfulness if it means getting a laugh. With top_k set to 20.
You’ll see the different in the “With RAG” response as the augmented prompt is part of your RAG pipeline. Try experimenting and let me know how it goes!
Appreciate your detailed response. I understood the concept now in much better sense. The Augmented prompt is just helping us to achieve a particular behaviour which we want to get from the rag application