Solve Issue 30 adapted to the STAGING branch

radicalxdev / kai-ai-backend

This is the Kai Teaching Assistant ai repo.

MIT License

12 stars 46 forks source link

Solve Issue 30 adapted to the STAGING branch #40

Closed AaronSosaRamos closed 2 weeks ago

AaronSosaRamos commented 3 weeks ago

This PR includes the same content as PR #31 for solving Issue #30, but it is adapted to the new STAGING branch and uses a more robust approach for few-shot inclusion for a more deterministic format of the provided questions by the LLM. Additionally, I tested it with gemini-1.5-flash and gemini-1.5-pro and it only shows the 429 Quota Exceeded error. So, I think that the most stable model for now is gemini-1-0-pro.

mikhailocampo commented 2 weeks ago

Thanks @AaronSosaRamos I tested this and things look to be in order for the Quizzify endpoint. However, the Dynamo endpoint does not implement the fix/ratelimits PR? Essentially, the workflow there means that we use Gemini 1.5 Pro to summarize and output flashcards

I merged that branch. Can you git pull and confirm Dynamo is using the new workflow before going into the staging branch

AaronSosaRamos commented 2 weeks ago

@mikhailocampo Dynamo is using the new workflow and it works appropriately: