Reported by @sour aurora
I tested with upload a couple of files named 1.py 2.py and ask some questions.
And immediately replace the 1 and 2 with 3.py and 4.py and ask a follow up.
The answer seems still referencing content in 1 and 2.
Then try the prompt to ask a hard fetch immediately. The memory/answer seems starts to read new updated files 3 and 4 but still having conflicts with 1 and 2. Since the content in 1,2 are 80% identical with 3,4.
And the timestamps in the answer are delayed upto 2 hours back.
Should immediate starts to read new content and drop old memories
Causing confuse and Mixed results, repeated answer/coding if it hasn’t be able to read your new files and old files has been corrected according to previous answer. Causing very bad project development experience. Has to manually wait 2 hours for it to forget the old file’s memory.
All environments, 5 pro chat mode