Sagemaker notebook configuration for streamlit

I am using streamlit==1.25.0 in a Sagemaker notebook instance (ml.m5.xlarge) with the following code :

import streamlit as st 

msg = st.chat_message("assistant")
msg.write("Hi")

However, with the link https:///proxy/8501, I have a blank page like this topic. I wanted to contact the author who seems to solve the issue with AWS permissions but I was not able to contact the person (except the tags @Zen but the previous topic was closed)

I tried also the things from this topic but it does not work.

I know the streamlit application works in a Sagemaker notebook (it succeeds in an other instance) but I struggle to see the difference in the aws configuration between both instance.

Can you tell me please the main points, I need to check in the configuration of the sagemaker notebook instance ?

Hi @FFrederic

There’s a post that provides a detailed explanation and starter code for deploying to AWS that may answer your question

Hi @dataprofessor

Thank for your reply. I looked in detail the medium article then I have seen the command used in the Dockerfile but unfortunately it does not work.

I am looking now the CloudFormation stack but I am not confident on this part, I will keep you updated on this.

Hi

I was able to run it by creating a python environment with pyenv and poetry. In this python environment, I only put streamlit=1.25 and he downloaded all the required dependencies. At first, it runs with the command :

streamlit run chat.py args --server.address=0.0.0.0 --server.port=8501 --server.baseUrlPath="/proxy/absolute/8501" --server.fileWatcherType=none --server.enableXsrfProtection=false --server.enableCORS=false --logger.level=debug --server.enableWebsocketCompression=false

But I reduced the command to this one :

`streamlit run chat.py --server.address=0.0.0.0 --server.baseUrlPath=“/proxy/absolute/8501”

The key seems about the rewriting of the url or maybe it was only with our specific Sagemaker notebook instance but It might help others.

By the way, it works on Safari and Firefox, but not on Chrome (blocked at the Please wait stage)

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.