Build your own ChatGPT with memory using Letta

In this video we'll show you how to build your own ChatGPT-style web application where every agent (or "chatbot") has long-term memory, using the open source Letta framework.

To build our custom ChatGPT with memory, we'll use:
- Letta (open source agents framework) for the backend

To learn more about the Letta ADE:

Useful links:

Timestamps:
0:00 Intro
1:07 Accessing the template
1:50 Initial setup
2:07 Optional app settings
3:42 Default agent settings
4:26 Start the Letta Server (using Docker)
4:55 Access the chat app
5:12 Chat with an agent
6:10 Toggling reasoning visibility
6:41 Viewing the agent's memory
8:00 View the same agent in the ADE
8:48 Triggering an agent memory edit
10:36 Outro

[letta-ai]

If you're confused about what the "ADE" is in 8:00, it's the "Agent Development Environment", which is a (free) frontend meant to help developers quickly build and test their agents.

[paudebatllejuan]

Hi, does it support third party APIs? Like ollama running on a separate server ednpoint

[christopherhartline]

I'm guessing this is a llamaindex product?

[trond]

i have been on the cloud waitlist for almost two months now

[ChuckWooters]

Another super helpful video! Thanks @letta-ai!