Prompt Evaluations Hack

Hello, Recently, I've been working on creating contract documents, scoring, and conversational apps using generative AI.

I'm not only interested in RAG but also want to improve the accuracy of prompts. I'm currently using around 100-200 prompts and would be happy if I could do component-based prompt engineering.

I am comprehensively using:

1. OpenAI Eval 2. Azure Prompt flow 3. Promptfoo 4. LangSmith 5. Wandb

If there are any better hacks, I would like to know.