Feature Request: Direct AI Access to Google Drive & PC Files
What if AI could securely access your Google Drive or even local PC files — all with your permission — to streamline memory and collaboration?
Right now, AI tools like ChatGPT rely on in-chat memory, but imagine an opt-in system where you could:
Link cloud storage so AI remembers your projects without constant re-uploads.
Access local files (like a connected app) to instantly pull info when needed.
Full control over permissions — only the folders or files you choose, ensuring privacy.
This would be a game-changer for entrepreneurs, creators, and students — no more re-explaining ideas or hunting for documents.
Would love to hear others’ thoughts!
This is implamented and has been for quite a while in Google Ai studio. I do advanced, fringe research relative to teaching llms, nlps and multimodal models EI and EQ. So I use their experimental and unreleased models to conduct my research.
Ai studio allows a setting within the UI to auto save to my Google Drive once it’s setup. I used to be limited to context within the chat window unless I was using my locally hosted models.
This was added many months ago, but like I mentioned, it has to be set up to work properly and don’t blindly rely on it as I’ve had times where it’s not properly saved and lost recent data. Before I had this feature, I would have the LLMs I work with create a distilled Json file that encompassed the important aspects within the chat window each time.
There’s a specific way of prompting them to do this in a highly effective way where you have them assign their own set of ontology, categories and indexing. I always add a request to have it referencable by a human, but created to be intuitive for a LLM. Also, make sure you use markdowns correctly for this if you do it as you want it to be as clear and simple to understand as possible for the next implamentation to an LLM. It does make a difference.
This methodology can be used to remind or retrain within a different chat window. This is less than ideal but can be a way to help mitigate a persistent interaction or training set.
Thanks for the tip on Ai studio. I’ll check it out. Using similar persistence work-around. 