How does vectorization work? And how do I find the right vector and documentation?

[7wafer7]

Hello. I am currently using pinecone. I have a problem finding the right vector in the documentation. What am I doing:

1. I have documentation on the question and answer, I vectorize the question and my answer is recorded in the metadata.

2. Next, I submit this documentation to pinecone.

3. Next, I try to find the correct vector in pinecone, pass the question to him, I took the question from my documents without changing it. I have vectorized the question and am trying to find the correct vector in the database, but I am not getting the correct answer, that is, the question and answer do not match the information from the original documentation. It happens that he gives the correct answer, but gets a score from 40 to 55%. It also happens that I change my question, ask different questions and get the same answer. Tell me, how can I correctly search for information in your service, or how can I provide information to get the correct vector? I’m using llama 2