Long Term memory
- To search in the long term memory store we need indexes which in turn require embeddings
- Embedding: is convert text into vector (numerical value with dimensions)
-
Long Term memory stores require vector db support, postgres has pgvector
-
Start the container
docker run -d `
--name pgvector-db `
-e POSTGRES_USER=postgres `
-e POSTGRES_PASSWORD=password `
-e POSTGRES_DB=experiments `
-p 5442:5432 `
pgvector/pgvector:pg16
-
We will be using embeddings from gcp
-
Refer Here for code added to store the data
-
To make this memory management simple langgraph has built a library called langmem.
