Hi Do you know why my pinecone serverless automaticlly delete my vector?

Iniially it shows I have 1004 in my dashboard, but after a few minutes, it is down to only 994 vectors. I do not have any delete request in my code.

Hi @vanthanh12195, and welcome to the Pinecone forums!

Thanks for your question.

The behavior you’re describing would definitely be unexpected, but it’s difficult to diagnose what might be going wrong without seeing all of your relevant code.

Could you please share the code to show:

  • How you’re creating your index
  • How you’re processing and upserting your data
  • How you’re querying your data

This will help us better diagnose what might be going wrong.

Thanks, and looking forward to hearing from you.


Hi Zack,

Thank you for the response. I’m figuring out the issue for auto delete. However, I have another issue: when I upsert a vector, it takes around 2-3 minutes for the vector to be available for querying. This delay is problematic since my use case requires immediate querying after upsertion. I’m using the serverless starter plan.

Looking forward to your advice on this.


It sounds like your use case may need two functions. One function handles an upsert while another handles direct use with your LLM. You should be able to use the same embeddings you upsert for a separate use to handle them immediately. It may be redundant to send embeddings to then immediately call them. Without code snippets it’s hard to tell.

Of course I don’t know if that makes sense for your use case or not.