Note: The Listed platform will permanently shut down December 31, 2026. Your data published on Listed will still be available in your personal Standard Notes account. Learn more

til abt letta

I feel a bit torn about how much robots can remember about me. It adds to the parasocial aspect of using what are tools.

But memory provides contexts and adds to the usefulness of so many agents. so...

I wanted to try out Letta, mostly going through https://docs.letta.com/guides/server/docker

From their AMA (https://www.reddit.com/r/AI_Agents/comments/1grnb8k/ama_with_letta_founders/)

Letta is an open source framework designed for building stateful agents: agents that have long-term memory and the ability to improve over time through self-editing memory.

Letta (fka MemGPT https://www.letta.com/blog/memgpt-and-letta) -- as far as I can tell -- mostly focuses on managing the memories and the context (#2) because then we can shift some of the burden between the layers of these four things -- and by memories I mean having the robot remember things about and for the user, rather than be useless:

  • making and saving memories
  • searching and prioritizing memories
  • applying and contextualizing memories to the current topic

I haven't used any particular robot that extensively to know if Letta is doing this better than say Claude or ChatGPT. But I can say Letta is offering a way to focus on the controls for how a robot does handle memories.

When I followed the docs and did this:

# replace `~/.letta/.persist/pgdata` with wherever you want to store your agent data
docker run \
  -v ~/.letta/.persist/pgdata:/var/lib/postgresql/data \
  -p 8283:8283 \
  -e OPENAI_API_KEY="your_openai_api_key" \
  letta/letta:latest

I got annoyed because it did not tell me that this line -v ~/.letta/.persist/pgdata:/var/lib/postgresql/data was going to create a temporary database, but it is probably implied in some way if you understand the basics of the Docker -- which I do not. But I improved my understanding of the flow in docker for containers, images, and volumes. The -v creates a volume for the container so I eventually switched it to -v letta so that I have a volume called letta that doesn't make my data disappear unless I say so.

I think Letta will put all conversation history and much more into the database on this volume, as well as all the Letta things like agents and identities and whatnot to handle the memories. I think of memories as the heart of a robot, especially when the memories contain the shared experience between you and the robot. (the brain is the model and the system around it, and the body is the various UIs to experience heart/brain outputs).

Letta is not fully functioning - some parts in alpha, some in beta - so the way it connects the heart to the brain and body are not very easy to use at this point.

But I did manage to get things connected across a full stack:

  • Discord.py bot on local machine
  • Letta on docker (cloud is waitlisted)
  • Postgres on docker
  • OpenAI API platform
Summary of flow: 
>> User sends message on Discord to Discord.py bot 
>> Discord.py passes to Letta 
>> Letta handles message and saves to Postgres database 
>> Letta passes message with context to OpenAI API 
>> OpenAI generates response
>> Letta handles response, passes to Discord.py bot 
>> Discord.py bot handles response, sends response message

More from @alexi | today I learned
All posts