File Search/RAG Not Prioritizing Document Retrieval as Expected

API
1

The file search or RAG does not seem to be working as expected. I have an assistant that uses a vector store with an updated file.

When I ask a question that is covered in the file, the assistant responds based on its own knowledge. Then, when I ask if it can answer based on retrieved documents, it does so correctly.

Any idea why it is not priorizing document retrieval as expected?

1 Like
2

There could be a few issues, some are known, and some just needs a few updates in how you prompt.

Known issues:

  • Assistants some times do not always recognize that there was a file attached, and you have to add your file ID to the prompt

Others:

  • Shift to v2 of Assistants from v1
  • maybe try a few different prompts
3

Thanks for answering!

maybe try a few different prompts

Not using any prompt. In the assistant instructions recommending the usage of retrieval whenever possible. Isn’t this enough?

Shift to v2 of Assistants from v1

already in v2

Assistants some times do not always recognize that there was a file attached, and you have to add your file ID to the prompt

This do not really scale because I have to add more files with different contents and I may need it to change them in the future. I am not passing files ids to run/thread only defining them in the assistant otherwise would be redundant, right?

4

Im also exactly facing same issue. for smaller document pdf like 10 pages I’m able to get answer correctly .but when I try pdf with 40-50 pages its not working returning any random answer.