The repository uses llamaindex to build 3 levels of RAG applications on my personal blog. It also uses TrueEra to evaluate each of the RAG methodologies based on context relevance, answer relevance, and groundedness
IMP : This is a follow-along from Deeplearning.AI's short course. But I am using my own keys and data to expand on the implementation.
PROMPT : Where was Unmesh employed in May 2022?
RESPONSE : Unmesh was employed at CEEW as a Sustainable Mobility Research Assistant in May 2022.
Error : openai.RateLimitError: Error code: 429 - {'error': {'message': 'You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.', 'type': 'insufficient_quota', 'param': None, 'code': 'insufficient_quota'}} (umenv) MACFG33C2C21M:Building and Evaluating Advanced RAG c536898$ pip3 freeze > requirements.txt
Response : Unmesh was employed at CEEW as a Sustainable Mobility Research Assistant in May 2022. My take : This is wrong answer. I don't have any weekly blogs in 2022. However, I was employed at the said organization in May 2021. This RAG application needs to be more advanced (will ask same question through window sentence retrieval) and also setup evaluation later.