My father is suffering from increasingly bad short term memory. This has been a slow process that has taken many decades. My mom was worried about his memory 20 years ago. Now he lives alone and has nobody to help remind him of things.
He is highly functional, able to multiply two digit numbers in his head and run heavy equipment on his farm. But he'll ask you the same question, five minutes after he asked it the last time. He'll forget what neighbor just visited him the day before, or whether he called the gas company to refill his propane barrel.
I believe that a voice interactive LLM with memory could be extremely helpful to him and to other elderly people. He could use it as a daily log, telling it things that have occurred, and then interrogate it later. In truth, who among us could not benefit from such a system, even in our younger years?
This framework looks interesting to use for such a project. Are there any other people working on similar things?
Our team has been working really hard on a low-latency voice integration that enables (open source) "advanced voice mode", but with long-term memory!
If you're interested definitely hop in our Discord to keep track of dev progress (you can also monitor our docs). It's all OSS of course, so you can use it with any model you want and run it all locally (your agent data stays on device).
Once available you can definitely use Letta for this usecase - basically provision a stateful Letta agent for your dad, with instructions / prompting that your dad has memory loss issues, and the agent should be proactive in recording memories (and potentially jogging your dad's memory too). All the memory/state will be stored in postgres, so you can expose it to other tools as well, eg could even collate the memories into a nextjs/mobile app where you dad can browse them.
> This word takes its root from the word "nisyan", which literally means “forgetfulness”.
There are two linguistic theories: the one you mention, and the other which is from “uns” (أنس) which means “companionship” or “affection.”
While both explanations have historical significance, the stronger linguistic evidence points to the latter explanation (companionship or sociability) by Classical Arabic linguists. The former (forget) is a more recent interpretation linked to philosophical and theological discussions rather than libguistic roots.
In Letta that's referring to an LLM agent that has persistent memory (eg keep running the agent far beyond its context window limit, but maintain some persistent state to avoid derailment / enable indefinite use / learning over long horizons).
If you implement this persistent memory via tool use, then it's related to inserting/querying databases, because some agent has to read/write to a data store (potentially backed by a database) to maintain memories.
We're working on it, I'm hoping we can get an integration in sometime next week! If you want to track progress on it would recommend joining our Discord.
To clarify - by "it", I mean making Letta agents an MCP client, so that you can easily connect to MCP servers as tools (not the other way around, where Letta itself is an MCP server - if you want that, someone in the community already made one: https://github.com/oculairmedia/Letta-MCP-server)
Totally serious question: is MCP basically an in-joke right now? I mean, I know what it is. I just don’t know why I’m seeing it everywhere in the last 3 days?
I'm also not really sure what's going on with the hype wave atm (people like having things to hype?) since MCP has also been around for a bit, seems like a bit of a meme.
IMO, the main value (as an agent / AI developer) that you get from MCP is another style of tool repository. Similar to how it's nice to have pre-made tools (eg Composio), MCP servers are another repo of tools to pull from (you just need to add MCP client functionality to your agent loop). So if a ton of energy is going into building MCP servers, might as well add MCP client support so that you can take advantage of the additional tools.
Probably people are excited about the prospect of having this problem be solved so that we can focus more on getting the agents to work rather than all be reimplementing the same data fetchers/parsers etc.
https://en.wikipedia.org/wiki/List_of_Tron_characters#Master_Control_Program
The Master Control Program (MCP) ... is the main digital antagonist of the first Tron film.
My father is suffering from increasingly bad short term memory. This has been a slow process that has taken many decades. My mom was worried about his memory 20 years ago. Now he lives alone and has nobody to help remind him of things.
He is highly functional, able to multiply two digit numbers in his head and run heavy equipment on his farm. But he'll ask you the same question, five minutes after he asked it the last time. He'll forget what neighbor just visited him the day before, or whether he called the gas company to refill his propane barrel.
I believe that a voice interactive LLM with memory could be extremely helpful to him and to other elderly people. He could use it as a daily log, telling it things that have occurred, and then interrogate it later. In truth, who among us could not benefit from such a system, even in our younger years?
This framework looks interesting to use for such a project. Are there any other people working on similar things?
> "voice interactive LLM with memory"
Our team has been working really hard on a low-latency voice integration that enables (open source) "advanced voice mode", but with long-term memory!
If you're interested definitely hop in our Discord to keep track of dev progress (you can also monitor our docs). It's all OSS of course, so you can use it with any model you want and run it all locally (your agent data stays on device).
Once available you can definitely use Letta for this usecase - basically provision a stateful Letta agent for your dad, with instructions / prompting that your dad has memory loss issues, and the agent should be proactive in recording memories (and potentially jogging your dad's memory too). All the memory/state will be stored in postgres, so you can expose it to other tools as well, eg could even collate the memories into a nextjs/mobile app where you dad can browse them.
-charles from letta
I find it somewhat amusing that Lethe was the river of forgetfulness, and that the new name for the project is so close to that.
Thanks for the info, never heard of Lethe before [1].
Fun facts in Arabic, "insan" meaning human being or mankind. This word takes its root from the word "nisyan", which literally means “forgetfulness”.
[1] Lethe:
https://en.wikipedia.org/wiki/Lethe
[2] Arabic for forgetfulness:
https://arabic.fi/words/3766
> This word takes its root from the word "nisyan", which literally means “forgetfulness”.
There are two linguistic theories: the one you mention, and the other which is from “uns” (أنس) which means “companionship” or “affection.”
While both explanations have historical significance, the stronger linguistic evidence points to the latter explanation (companionship or sociability) by Classical Arabic linguists. The former (forget) is a more recent interpretation linked to philosophical and theological discussions rather than libguistic roots.
Man, mind, to remain, to remind: they all originate from the same intertwined proto-indo-european roots "men" (thinking) and "mon/man" (human).
Edit: just found this etymological trail for insan:
>From Proto-Semitic *ʔināš- (“man, person”). Has the form of a verbal noun , as if from أَنِسَ (ʔanisa, “to sense (visually or auditorily)”).
What's an "LLM with memory"? Is it an agent that knows how to insert into, and query, a database?
https://docs.letta.com/letta-platform
In Letta that's referring to an LLM agent that has persistent memory (eg keep running the agent far beyond its context window limit, but maintain some persistent state to avoid derailment / enable indefinite use / learning over long horizons).
If you implement this persistent memory via tool use, then it's related to inserting/querying databases, because some agent has to read/write to a data store (potentially backed by a database) to maintain memories.
-charles from letta
I didn't find any thing on Letta and MCP. Any thoughts on integrating MCP into the platform?
We're working on it, I'm hoping we can get an integration in sometime next week! If you want to track progress on it would recommend joining our Discord.
To clarify - by "it", I mean making Letta agents an MCP client, so that you can easily connect to MCP servers as tools (not the other way around, where Letta itself is an MCP server - if you want that, someone in the community already made one: https://github.com/oculairmedia/Letta-MCP-server)
-charles from letta
Totally serious question: is MCP basically an in-joke right now? I mean, I know what it is. I just don’t know why I’m seeing it everywhere in the last 3 days?
I'm also not really sure what's going on with the hype wave atm (people like having things to hype?) since MCP has also been around for a bit, seems like a bit of a meme.
IMO, the main value (as an agent / AI developer) that you get from MCP is another style of tool repository. Similar to how it's nice to have pre-made tools (eg Composio), MCP servers are another repo of tools to pull from (you just need to add MCP client functionality to your agent loop). So if a ton of energy is going into building MCP servers, might as well add MCP client support so that you can take advantage of the additional tools.
-charles from letta
Probably people are excited about the prospect of having this problem be solved so that we can focus more on getting the agents to work rather than all be reimplementing the same data fetchers/parsers etc.
Sure, but I mean why so suddenly? This has been out a while...
I think it’s just reached critical mass (and has been picked up by the hype peddlers). It’s been gaining momentum for months.
Tangential cultural reference ...
Interesting question, I personally first saw it because jetbrains now has a plugin for it on all their IDEs.
Maybe this will help you figure it out:
https://github.com/modelcontextprotocol/servers
Thank you for the mind-blowing 2 hr rabbit hole. I'm going to use this in my personal helper bot.
This was the berkeley team that previously released MemGPT
indeed - hello there