hey guys,
i am making this thread today ahead of the release of LAM playground at 10am PT tomorrow, since we will be busy with the announcement.
get ready to share the things you try with it in this thread!
hey guys,
i am making this thread today ahead of the release of LAM playground at 10am PT tomorrow, since we will be busy with the announcement.
get ready to share the things you try with it in this thread!
Ready for it! R1 action only or do we need the rabbithole aswell?
You have two choices:
LAM playground (web) - fully operated and prompted on web
LAM playground (r1) - prompt from r1, watch on rabbithole
Hey @simon ,
this will be very very great.
All rabbits will come out of the rabbit hole.
Really looking forward to this and for my IT job, it could be very helpful.
What do you think about a workshop area on the rabbit website or something like else to share and maybe download user creations?
This is really good news!
From all of the bits and pieces I’ve read so far, the LAM Playground is a Generic Website Agent where you can watch rabbit fulfill your request in real-time.
Requests can be submitted either through a text prompt within Rabbit Hole or making a request using natural language to your r1.
One of the big benefits that I see for this playground is that it will help to better understand “how” the LAM actually executes requests. I expect that this understanding will help me to debug, optimize, improve the effectiveness of my requests.
I love the detailed explanation provided on the LAM Playground page: https://www.rabbit.tech/lam-playground
First impression is amazing! I now can watch what rabbit is doing, saw a strange action, adjusted my prompt, and then the r1 nailed it!
Doesn’t seem to be working for me.
I mean it is, but I cannot see what it is doing in the rabbit hole.
And it is slow…
I completely understand this is at an alpha stage.
I asked for it to read out the first article posted on a Christian music website called https://www.christiandance.eu/. Then I asked it to look up song lyrics on the same site. It nicely opened up the search bar and looked for the song lyrics.
It is slow, but I saw exactly what it was doing in the browser version.
I didn’t get the r1 version to work (yet) as described in the manual. There I only saw the final page, and not the steps it took.
So far I am pleasantly surprised at it’s capabilities. It managed to navigate a Dutch site for an Italian restaurant to find the cheapest vegetarian pizza option (spoiler: it was a margarita).
For the Dutch local eBay alternative (marktplaats.nl) it could regrettably not handle the cookie dialog.
Practical tip for the non-english users: you can initiate the request in terminal mode so that you can type instead of say the name of the website and then switch back to normal mode to hear how the command executes to completion.
Just a simple question - the session of the playground is a “Remote Desktop”? I mean what is the point to ask to add a product to your amazon cart if your not logged in in amazon?
I may not be understanding the capabilities. I went to classicalking.org and asked it to listen live. I did it through the R1. It says it is and to enjoy the music, after some navigation, but I hear nothing. Same with KDFC.com. This is through the R1 and through the web interface on the rabbit hole.
When I ask it to go to youtube.com and play nbc news live, it does on the LAM online, but I hear nothing.
I can’t start a LAM session with R1 right now. Not sure what’s wrong. But I did notice some cool things. It looks like you can click on the LAM window and log-in to a website. Does that mean it will remember your login info for that current site? Or do you need to log-in everytime. If you do have to every time, can we somehow store the login info in a Journal note and have it find it to log in for me? If we could figure out this login thing, there are so many possibilities.
I believe you need to login every time. Still useful though. I could login to google and have it summarize the top 3 emails and even send an email.
Hi I am noticing that the LAM on the WEB is sometimes disconnecting half way through . Does it mean my R1 is disconnecting. Is the Browser session running on my r1 as i suspect?
Is there a way to teach the Agent when it makes a mistake. I saw the agent not able to use a slider control on a website can i show the agent how to use it. Will it learn if I do it
you can login and it will save your credential securely.
I built an entire app using Playground and Sutro!
I use ntfy.sh for various things for work / home notifications of automations that I run. LAM Playground is able to navigate specified “topics” and determine the latest notification of each, as well as you can use LAM Playground to send notifications through ntfy.sh by having it use the web app. This is useful for if I want it to gather some information and then send that information to my phone.