Setup (Windows) + Memory Retrieval

I installed Gemini CLI on C:\Users\[UserName]

  1. I couldn’t retrieve the suggestions.md initially until I gave it its full github path.

  2. When I instructed to store “My name is [name] and i like [color] on [day], then upon asking “what is my name”, it actually start scanning the different files and dirs in that path instead of first looking it up in its Memory. I had to cancel this look up and specifically instruct it to look up the name in the gemini.md (context file).

    Is there a way of prioritizing resources lookup or is there some kind of error on my part? Thanks.

1 Like

Thank you for reporting! That is an interesting finding with windows setup. Just shared with the team!

No problem, I am a QA by trade, so unfortunately I will pick on everything, haha.

Another potential issue, or perhaps it is a designed, and this pertains to the overall workflow:

  • once signed in into DeepLearning.ai and enrolled in a course, within that course, clicking the “Forum” icon returns a sign up page for community.deeplearning.ai. I would think, perhaps incorrectly, that only one account is needed for both courses and forum.
1 Like

It appears that most of the “Discussion Forum” links on the Learning Platform courses just go to the Forum home page, not to the specific forum for that course.

oh, I see. Anyway, I’m curious about it… have you just followed instructions about adding memory as stated by the instructor? (/memory add …)

and then, was it saved in memory? and you can see in the terminal something like… “memory refreshed…”?

let me know, that will be helpful to support other learners using windows and/or running into similar issues with memory!

Hi Lesly,

Yes, i did follow the instructions for saving to memory/context file. In fact, only after I asked gemini to look the gemini file did it find my name.
I think the issue here is that the github repository containing the data must be cloned?
I originally installed gemini on c:\users\[my username], and upon encountering the problem above, I decided to install in an empty dir I created. That solved that problem above, but for the subsequent lessons in the course, after installing node, there was no webpage page displayed on either port 3000 or port 8571(?). And I believe that’s because the content for the website is not local.

Perhaps,the course info needs to specify cloning that github repo?

1 Like

According to Gemini Convo AI, the problem is/could be a current default settings:


It advised reducing this setting to 1 or 0

Even with “Include Directories” set to false, the CLI’s default behavior is to recursively discover all context and memory files within the active workspace. With a limit of 200, if the gemini command runs from a high-level folder (like the User folder or C:\), the CLI will still attempt to index 200 subfolders beneath that point to find GEMINI.md files or other context..

1 Like