Name: Screenpipe
Uploaded: 2025-02-15T07:35:54.802Z
Duration: 2 min 33 s
Description: An open-source platform enabling 24/7 local recording of screen and audio activities for AI application development.

screenpipe

An open-source platform enabling 24/7 local recording of screen and audio activities for AI application development.

Screenpipe

When Microsoft came out with Recall I had like an allergic reaction to it. The idea of Microsoft recording your screen at all times just didn't sit right with me. Something that's that invasive and all knowing I feel should be open source. There's something about wanting to make sure you know the innards of how it's working and what happens with the data where it goes is very important to me. 

So even though I didn't like the implementation I really like the idea because having access to semantically search what you go through the entire day could be really helpful because I know I'll be having a

Conversation like oh I saw I saw this one GIF or I saw this one document I really wanted to share. No idea where it was. I have too many tabs open. It gets lost in the clutter. 

So having the ability to ask my computer or ask an AI to pull something up or to reference something sounds very valuable. So I was pretty excited when I heard about Screen Pipe. 

Screen Pipe is an open source project also MIT licensed where it will record your screen in audio whether it's your microphone or system audio as long as you have it running. And then it's cable. 

Now the project is only 4 months old still relatively new so.

The API isn't very developed yet but it has a lot of promise. So I just wanted to quickly show um. So you get a lot of options. You can set the monitors that you're going to use. You can set the audio devices. So it's getting there. Like it's definitely showing a lot of promise and I just wanted to quickly demonstrate how earlier in the day. 

Also sorry when you search for something like okay where where have I mentioned open adap today or when does it come up on my screen you get this whole blob of information. It's very overwhelming. I pipe it to JQ to kind of structure it a bit and you can see actually.

This this little quote was when I was watching Richard's demo video um or you have like a a OCR over my screen when open adap a was there. So it's just it's a lot of information right now but it's helpful to have this be a component in the system. 

So rather be a standalone thing you use if you use it with something that can consume all this information and present it in a nicer way it becomes very powerful. 

So I actually set up a profile an open interpreter called screen pipe and what this will do is it'll allow you to use your natural language queries get converted into a green pipe database query and be.

Able to return the results and then act upon it. So let me just really quickly run this. When was I talking about open adapt today? So I'm using Gro, that's why you get that snappy speed. 

And I set up this search screen pipe function as a custom tool open interpreter. So it'll pass an open adapt and query it. Let's approve that. 

And again you get a whole bunch of information, but open interpreter takes us all in, processes it, and then it'll actually tell you a more consumable way of what you're dealing with here. So there we go. I was looking at the open adapt AI GitHub earlier today, which is true I was.

Checking it out you know wanted to check out the code. Can you please open up that website for me? And this is where the open interpreter aspect comes in because all of a sudden you're able to take this data and action it. Oh it cut it off. I blame that on grock. I blame that on grock. 

This is just llama 70b so it's nowhere near as powerful as GPT 40. But the premise is there where you're able to use a tool like screen bip or like you showed REM something that is able to consolidate this data and make it queriable. All of a sudden unleashes a whole lot of extra context that you can use in your AI system. So I'm

Trying to encourage people to think less about an individual tool as the solution versus a pipeline solution or a system set in place similar to how Richard showed us the architecture diagram. There's so many moving components that were so early in the process that you want to start piecing things together. 

Being able to have information flow smoothly between the system to really achieve the results you want. So I really encourage people if they're looking to automate a process to explore automating or creating code to satisfy a single version of it and then open source it because if we can take.

These plug andplay components. These little Lego blocks. We'll be able to start gradually building up more and more systems to achieve this level of automation that we're really striving for. 

So uh screen pipe still early days but a potentially valuable component to a larger system. So check it out that's super cool.

Note, that there may be slight differences in text and video. Watch video for the original content.

Screenpipe Video