Having trouble developing with the Assistant API? The API documentation is unclear, please help

[victory_1128]

Since OpenAI developed the API, I have started developing internally using GPT. Because it is not necessary to develop multiple projects at the same time, it is not possible to develop GPT all the time.

Currently, we are developing knowledge retrieval Assistants (such as the retrieval assistant for all system administrators in the group). When I am reading carefully the Assistants API on the OpenAI official website, I feel that some places are not very clear.

1. In the Create assistant API, there is the file_ids parameter (the id of the jsonl type file),
   There is also a file_ids parameter in the Create message API (used for retrieval or code_interpreter, file type list)
   a. What role do these two documents play in the entire operating principle of assistant?
   b. What is the priority?
   c. If it is knowledge retrieval, can only use assistant file_ids?
   No need to use message file_ids?
   

   

2. Create assistant API\Create message API Is there an error when the reference links at file in the two file_ids point to the same page?

3. Create assistant in the playground and find that no content is returned?
   

   

[h.gurgenci]

The answers to your questions are not documented in the OpenAI resources. The question on file IDs with Assistant and/or Message is tricky because the Assistant sometimes does not work depending on where you define the files.

You will see several posts in this forum about that topic but no answers. Getting frustrated, I recently spent a couple of days in doing a deep dive into Assistants API and documented my findings here: https://halimgur.substack.com/p/deep-dive-into-openai-assistants. I think you will find it useful.

[victory_1128]

The official document states that there is a problem but no one responds, which is frustrating.
These logics or processes should be clearly expressed, and the official should have this obligation or responsibility.