Pieces app
Background
Last week, I attended a AI meetup where the creators of pieces presented a demo of their app. I’m in the process of testing it out, so I’ll update this post with use my use cases.
The most appealing feature is that it supports a local mode (with a local llm so that your files don’t get uploaded to the cloud).
The app is current available to the public for free! I hope the momentum will keep the local llms public and make the cloud/openai components as part of the paid version.
For those unaware, Pieces is a app that periodically screenshots your screen, taking new screenshots if there are changes from past screenshots. Using the screenshots stored in RAM, the app will crop for text so that the ocr engine can extract the text. With the screenshot text extracted, the app will use a LLM (local or openai) to enrich the text further and build a local knowledge database that can be queried in a chat interface.
Linux support?
Currently the app is distributed via snap. The pieces developers are always looking for feedback. If you would like for them to make a flatpak release, let them know here!
FYI
note: i’m not getting paid from pieces to promote it. I thought this would be a cool app to share. Through the L1 forums, I found LM studio.
I figure the same audience would also be interested. This might be interesting for @wendell because he started this thread.
I currently use foam suggested by wendell in vscodium and some other markdown extensions to write daily notes and build my own personal knowledge wiki.
When I am in flow state solving issues, I don’t always keep track of the past solutions that worked. If I had to note every solution and the result my flow state is broken. If I encounter the same issue, I won’t remember how to solve it. With pieces it will automatically build a local knowledge repo I can refer back to it giving a relative time window.