Evaluate AI Response Relevance with OpenAI and Cosine Similarity

The workflow is designed for Q&A agents and leverages OpenAI models to generate questions from AI responses. These generated questions are then compared to the original user questions using cosine similarity to determine relevance. A high similarity score indicates a relevant response, while a low score suggests potential issues such as irrelevant information or hallucinations. This process is based on the RAGAS evaluation metric for answer relevance.