RAG Chat Docs

"Retrieval-Augmented Generation (RAG) is an advanced AI framework that combines information retrieval with text generation models like GPT to produce more accurate and up-to-date responses. Instead of relying only on pre-trained data like traditional language models, RAG fetches relevant documents from an external knowledge source before generating an answer."

geeksforgeeks.org

What can RAG do for You

For this MVP there are 4 seperate document libraries that are individually queried depending on the user's selection. If the document libraries doesn't supply an adequate answer, it will not hallucinate a response. View the library documents to understand the scope of information LLM has access to.

Library # of Docs Desc.
dog_breeds 3 docs Dog breed characteristics and attributes. Shows how focused a library can be
classic_games 10 docs Collection of popular game reviews. Show off how opinionated explinations can influence LLM output
dog_rescue 20 docs Assortment of rescue dog personalities and traits. Show off a collection of one-of-a-kind documents and connecting them to the right user
employee_wiki 20 docs List of tutorials and walkthroughs. For scope sake, they only focus on Zoom and Outlook problems

Example Prompts

Try out these example prompts. Some of the prompts purposfully ask wrong questions to test hallucination

employee_wiki

"how to reset employee password?"
"how to create a calendar group?"
"Help me my computer doesn't work?"

classic_games

"open world games that are good even in modern times?"
"What is a good game for new gamers with low skill?"

dog_breeds

"dog with the best sense of smell"
"what are some cute orange cats?"

RAG App MVP

RAG AI processing goes through 2 processes. The Left Column is what happens behind the scenes, semantically searching and returning relevant documents from the library. The Right Column is the end user experience. It runs the same "Search Documents", then process those documents with an LLM to return a response in natural language.

Search Documents

The words submitted are semantically proccessed against a limited number of documents. Returned are the top 3 closest matches.

Search Results:

    Chat With Docs

    RAG retreives the 3 closest matches. Those matches are then read over by the LLM. The prompt is tuned with extra instructions. The response will read in natural language.

    Chat Response:

    Documents Referenced