LangFriend: a Journal with Long-Term Memory

One of the concepts we are most interested in at LangChain is memory. Whenever we are interested in a concept, we like to build an example app showing off that concept. For memory, we decided to build a journaling app! We're hosting a version of it that anyone can try out. We're also starting to work with a few alpha users on a developer facing API. If you are interested in this, please sign up below.
Key Links:
Blog: blog.langchain.dev/langfriend/
Journal App: journal.langchain.com/
Developer API Access: forms.gle/j3Aaa2ibNpg5pC4q7

[@lexandery]

Why the API, why it's not in the library?

[@rajibdeb4059]

Love this. Have been writing about long term memory for a while. This one looks closer. Addition of episodic(raw) memory will further help to find more contextual and personalized relations

Very nice! The only other framework able to do this was autogen teachable agent. Integrating this long term memory with selective capacities into such a wide and open source framework such as LangChain opens lots of new AGI perspectves. Thanks HC.

[@Cloudways-AI]

Another nice one Harrison.

[@benjamellamo]

Where could i learn how this works?

[@sa-fj2dm]

Superb

[@curiouslycory]

Very nice! This solves very similar challenges as I've been facing recently on some personal projects. How would you say this differs from something like vectordb + RAG? Hope we can get access to the APIs to play around with soon!

[@rajibdeb4059]

I felt memory architecture and vector DB has a impedance mismatch. memories are like multiple connected patterns and probably vector db may not be the right place to store those patterns. Not sure what storage journaling is using. For us humans, our memories are all connected. We can time travel back and forward in our memory

[@leleutd]

This would be something I'm very interested in if it can be self-hosted... Are there any plans around doing that with the journal app?

[@timothywcrane]

I have been exploring various noting apps like obsidian and logsec and even mediawiki and wikibase to layer conceptual and action oriented pieces of data into a "concept traversal mapping" for inference use. This looks very promising as an "in-memory" encapsulation. Are there plans for persistence and retrieval or programmatic "feeding" of concept notes into the context memory session ( based on user selection or dynamic input inference )?

[@timothywcrane]

Last 3 seconds answered... Thanks.

[@mofa6423]

Have you checked Google's lmnotebook?

[@diallo9149]

nicee

[@MirGlobalAcademy]

😊

[@jorgefelipegaviriafierro705]

Nice video :), Forza Inter :v