You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have spent a few days working out the kinks deploying Zep to GCP Kubernetes Engine and I found two major issues that I think I have solutions for. I was going to submit a pull request, but wanted to check here first in case I was missing something.
First, there is seemingly a Typo in the deployment.yaml file. It should be ZEP_STORE_POSTGRES_DSN not ZEP_MEMORY_STORE_POSTGRES_DSN
Second, I had all kinds of trouble with the healthcheck and the solution was to remove host: 127.0.0.1 entirely from the healthcheck and that solved the issue for me.
And not a bug, but in case someone else is looking for this, I was able to use a managed SQL Postgres instance instead of the docker/k8s based one and that seems to be working just fine. I did enable the pgvector extension, but it supports it with no apparent issues.
Thanks all,
-Corey
The text was updated successfully, but these errors were encountered:
Thanks, @CoreyH. Yes - the k8 template was inadvertently not updated when the DSN key changed. A PR would be appreciated!
It's fine to remove the health check. We can leave that as an exercise to the implementor.
Zep definitely works with managed Postgres instances. The only caveat is that not all providers have upgraded to pgvector v0.5 yet, so HNSW indexing may not be available.
I have spent a few days working out the kinks deploying Zep to GCP Kubernetes Engine and I found two major issues that I think I have solutions for. I was going to submit a pull request, but wanted to check here first in case I was missing something.
First, there is seemingly a Typo in the deployment.yaml file. It should be
ZEP_STORE_POSTGRES_DSN
notZEP_MEMORY_STORE_POSTGRES_DSN
Second, I had all kinds of trouble with the healthcheck and the solution was to remove
host: 127.0.0.1
entirely from the healthcheck and that solved the issue for me.And not a bug, but in case someone else is looking for this, I was able to use a managed SQL Postgres instance instead of the docker/k8s based one and that seems to be working just fine. I did enable the pgvector extension, but it supports it with no apparent issues.
Thanks all,
-Corey
The text was updated successfully, but these errors were encountered: