math_helper

LLM based math helper that supports the student with math tasks.

Note:

- ollama uses a local setup and models need to be pulled before use 

functionality:

• provide the user with math tasks and check the answer
• process user input - if user asks for help, create a helpful answer
• uses current student tasks and previous answers as part of its context
• help is provided stepwise (no giving up the answer at first try)
• offers different LLM-setups

example usecase:

student has a problem in their math quiz app. --> they answer couple of times and get wrong answers --> they ask help from the math helper: -I'm getting wrong answers, can you help me with this? --> math creates a hint based on the current task and previous answers --> students tries again, with the helpful message and gets the correct answer

Parts of the program:

math_app

• provide tasks
• check answer correctness
• take input from the user and print outputs
• communicate with assistant
• operate task database

assistant

• create responses for queries
• LLM setup
• store chat history
• store pedagogic prompt
• communicate with calculator if needed

task_database

• store question - answer pairs
• hold statistics?

calculator

• solve math problems stepwise and return steps

RAG-database

• store educational and informational math documents
• initialize vecotor store
• provide documents based on queries