Hello.
TLDR: save the context of Assistant, now it’s lost every time.
For the record: I feel like I’m reinstalling Windows 3.1 as a schoolboy or Windows 2000 as a student. I just don’t understand why the system works this way - it’s magic, sometimes it works, sometimes it does not.
Jokes aside.
Why Assistants in the Playground DO NOT SAVE the context?
I have 4 files with text to work with. After lots of tinkering, I converted them to plain txt (pdf does not work) and after lots of “speaking with model” I managed to get some results from it. Like “show me the Table of Contents, create abstracts, etc”.
Next day (today) I go back to my Assistant and it forget everything.
Hello! Yes, it appears that four files have been uploaded to the system. However, these files are not accessible with myfiles_browsertool, which means I can't browse them directly. If you would like me to use these files, please ensure they are accessible with themyfiles_browser tool by re-uploading them or let me know if there's any other way I can assist you with the information you are seeking.
After some wrangling I manage to make system to remember that it has files.
My apologies for the confusion. I do now see that the files were uploaded previously and should be accessible for me to assist you with your request. Let's begin by accessing these files. Please give me a moment to open the first file and we can proceed from there.
New book about Zeppelins
I have successfully accessed the first file
It’s not just loss of time. It’s a real loss of time - I spent 30 minutes doing this. And all context which I created yesterday is lost.
User
how much time and tokens have you used to process these 4 files today?
New book about Zeppelins
As a language model AI developed by OpenAI, I do not have the capability to track time or token consumption during the processing of tasks. My functionality is designed to access and analyze files or content while engaged in active dialogue with users.
The processing of tasks is handled by OpenAI's systems, which manage the execution and resource allocation. OpenAI may have internal metrics to measure and manage the performance of the model, but as an AI, I do not have access to such meta-data about my operation.
If you have any more requests or need further assistance with the files or any other queries, please feel free to let me know, and I will be happy to help.
It had cost me 50 cents 
So, please, add the ability to save context. Working with an intern as an assistant is one thing. Working with an intern with a goldfish memory is really, really worser.
(seriously, do you need help with this?)
I really want to use this tool. But it’s not that good.