Simulating ChatGPT's 'Canvas' Experience with Lindy

Hey all! I'm wondering - what is the best method of simulating the 'canvas' experience that you get with ChatGPT? For instance, let's say I want my AI to iterate on a document together with me - on chatgpt, it's as simple as just setting up the custom instructions, then we can work together on the canvas. How can I achieve similar with Lindy? Thanks so much! CC Marvin A. and Alon J. 🤞

👀1

8 comments

Marvin A.
·
·
Hey Ben! Thanks 🙏🏾 yeah we don’t have that feature yet — I like to collab with my agents inside of slack threads tbh you could also have it update a google doc for you 👀 I’ll take that as a feature request though!
❤️1
Ben S.
·
·
Thanks! Yeah the only problem with the Google Doc approach is that it only has the option to append or prepend, but I guess I could just get it to rewrite the whole document, but then we start using tokens and also it might not rewrite it exactly the same which would mean rechecking. What about notion? Have you tried experimenting with that? I'm wondering if maybe that might work because then we could get it to update specific blocks rather than having to redo the whole document
Ben S.
·
·
Also one thing I was wondering that you might be able to help me with - How does the default memory/context work? For instance, it seems to default remember what was said in the previous messages. So if I get it to update a Google doc and then get it to update the same Google doc with more information, it seems to remember the context of the previous update, and that informs the next update. But presumably, there's a limit to that context? If I wanted to get the user to manually change the document and tell the AI that it made changes, would I then get it to reread the document to get the newest context? But would it then get confused between the previous context or will this context overwrite it? Not sure if my question is clear. Sorry. I can elaborate if it's not 😅
Ben S.
·
·
Hey Marvin A. no pressure/urgency, but would love to hear your thoughts on this if you have a moment 🙏
📌1
Marvin A.
·
·
Hey Ben S. sorry for the late reply!! On memory/context: You’re spot on about how it works. Lindy does maintain context across messages in a conversation, but there are limits (usually 100k-1m tokens depending on the model).
Marvin A.
·
·
yeah not sure if you’ll get close enough to the canvas experience but I’d likely use google docs and have the agent rewrite the whole doc if needed. If you plan on having many iterations within one task you can experiment with Clear Task Context to only keep relevant information after a certain point 🙂
Marvin A.
·
·
hope this is at least somewhat helpful! 🙏🏾
💯1
👍1
🙏1
Ben S.
·
·
Absolutely helpful, thank you!!
🙏1

Simulating ChatGPT's 'Canvas' Experience with Lindy | Lindy Community