Smarter Memory with Semantic Search in LangGraph

[@MatthewSanders-l7k]

Thanks for breaking down semantic search in LangGraph, this tutorial is super helpful for building a chatbot with long-term memory.

[@CharlotteLopez-n3i]

Great tutorial! I'm excited to try building a chatbot with long-term memory using LangGraph. Your explanations of semantic search and index configuration were super helpful, thanks!

[@AlexJohnson-g4n]

Thanks for the detailed tutorial on building a chatbot with long-term memory using LangGraph. I'm excited to try it out and see how semantic search can improve my chatbot's conversations.

[@WinonaNagy]

Thanks for the detailed tutorial! Excited to try out semantic search in LangGraph, been looking for a way to improve my chatbot's memory.

[@brunomesquitazamberlan8876]

The Perfect vídeo! I’m creating a ReAct agent and i was finding some alternatives to use the memory saver. Thanks for sharing

[@DmytroDragan]

Thanks for sharing! Use of the RAG approach for the long-term memory solution makes perfect sense. I guess the next step will be to extend it to some Hybrid RAG to store the metadata relationship to provide some sort of association memory as well.

[@gr8tbigtreehugger]

Many thanks for this video! Exactly what I'm working on!

[@andydataguy]

Stoked to try this out! Wonder how to get it to work with qdrant

[@banglaLang]

Found it, when needed it

[@aifarmerokay]

Can we use redis for memory? Helps us with that

[@i2c_jason]

Could you create a list or dictionary of embeddings to hunt down manifolds in the latent space, perhaps using some clever numpy algorithms? So, use higher level code to track down solution spaces in the low level ML realm from a sparse set of starting semantic terms?

[@i2c_jason]

Does this mean we wouldn't need to use Neo4J for graphRAG? Just use semantic search in whatever storage mechanism we prefer?

[@willfu-hinthorn]

Most of the Graph RAG implementations I've seen don't use cyphertext and so don't require neo4j. Neo4J is nice if you want to be running complex multi-hop queries, which are a minority of lookups you typically do

[@CiaoKizomba]

How is this different from Rag?

[@willfu-hinthorn]

This uses rag, but memory also deals with the write-path and the synthesis path.

[@swyxTV]

you guys can afford to get Will a better mic!

[@VoyceAtlas]

so dont use command?

[@LangChain]

You can use whatever you'd like

[@Cenot4ph]

langraph studio only being macos is a big fail, when is this getting released for linux, windows?

[@LangChain]

It has worked on Windows and Linux for over a month. ``` pip install "langgraph-cli[inmem]" langgraph dev ```