Problem with retrieval in Assistant API

API
, ,
1

Dear Developers,

I am reaching out for assistance with a few issues I have encountered with the OpenAI assistant:

  1. Retrieval Issues: Occasionally, when I ask a question, the assistant fails to retrieve the correct response from the files, even though I know the information is present. For example, if I have a text file stating “Jon likes orange color,” sometimes the chatbot provides the correct answer, but other times it does not. Could you help me understand why this inconsistency occurs?

  2. Vector Store Charges: The documentation mentions that charges apply if the data exceeds 1 GB. However, I have not been charged despite surpassing this limit. Can you clarify why this is the case?

  3. Red Point in Vector Database: There is a red point in my assistant’s vector database. Could you explain what this indicates?

001

  1. Token Processing Limit: I am trying to understand the token processing limits. For instance, with the default value of Max num results set to 20, increasing it to 50, will it exceed the token limit? If a user’s question is 4000 tokens and the retrieved result is 20 (20800=16000), the total tokens will be 4000 (question) + 16000 (retrieved text) + 4000 (response) = 24000 tokens. If I increase the Max num results to 50, the total consumption would be 4000 (question) + (50800 retrieved text) + 4000 (response) = 48000 tokens. Can the assistant process this?

Thank you for your support and assistance with these matters.

Best regards,

2 Likes
2

Hello :slight_smile:

Have you discovered what is this red “flag” is?

regards

3

yes, one of the files uploaded to the vector store has problem and you have to remove the file form the uploaded files not from the vector store. it will be solved.