Students spend time navigating UT's Course Schedule to test schedule combinations. A UT Registrations MCP + Spark could eliminate that friction as users can describe what they want for their schedule. Spark can explore said possibilities (similar to Degree Audit Plus) with UT Registration MCP.
Using chrome's sidebar, Spark AI could be toggled directly with the UTRP Calendar UI open (or any tab). Since we already constantly update function implementations for UTRP due to various updates, providing a standardized MCP for the chatbot (or say, Cloudflare Worker) to retrieve data from UT's Course Schedule page is plausible.
With the RAG contextual info & users query & MCP call output, surfacing a coherent and genuinely helpful message to the user is possible. The vector database can be refreshed at the start of every semester.
Tried and tested vector database for updated UT Course Catalog (retrieved through OAuth-allowed Course Schedule scraping & other UT online database) is a good start for this. To have said OAuth, user must (upon installing Spark AI / UT Registration MCP along with Spark AI) login to UT Course Schedule.
RAG can be used with the vector space database stored in Supabase (or an equivalent), where data retrieval functions we created for UTRP can be repurposed to provide particular outputs relevant to the user's query.
Embedding, unembedding, and MCP calls made by the Worker.