Update README.md

README.md

@@ -8,14 +8,20 @@ colorTo: yellow
 pinned: true
 ---
 # USC-GPT
-A chatbot that uses a pinecone vector database of USC Spring 2024 classes and OpenAI's GPT to help students find classes! Using RAG and Prompt Engineering
+A simple MVP chatbot that uses a pinecone vector database of USC Spring 2024 classes and OpenAI's GPT to help students find classes! Using RAG and Prompt Engineering
 
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/63cc674af488db9bb3ca3bd5/c5ySQCW9C2rIq1Rol-Pp2.png)
 
 Starting with the USC schedule of classes website for all classes provided, and following that stored and embedded them into a vector database (Pinecone!).
 Then using multiple agents to handle different steps in the process (to abstract away prompt engineering from the user to the backend), we are able to do complex searches to help students find their ideal class:)
 
+## Room for Improvement:
+1. Using fine-tuned BERT models for certain agents rather than relying on a large language model (a bit overkill)
+2. Optimizing RAG searches and incorporating more information about classes in GPT responses
+3. Support for a user to choose what language models to use
+
 ## Agents
+We use agents as a simple and easy way to handle some complex NLP tasks. These could likely be handled in other ways.
 #### Filter Agent:
 An agent to extract filters from the user's query and outputting them as a json string. These can then be applied to the vector DB index query