Missing necessary newline in st.chat_message

[ShaneTian]

Checklist

• I have searched the existing issues for similar issues.
• I added a very descriptive title to this issue.
• I have provided sufficient information below to help reproduce this issue.

Summary

When I use st.chat_message in LLM demo, I found the output message was missing some of the necessary line breaks.

The assistant message below is from GPT-4, so it is common phenomenon.

Reproducible Code Example

import streamlit as st

message = "**Title: Leveraging Large-Scale Language Models for Enhancing Communication Technologies**\n\nI. Introduction\nA. Overview of the evolution of language models in artificial intelligence (AI)"
st.chat_message("assistant").write(message)

Steps To Reproduce

No response

Expected Behavior

Current Behavior

Is this a regression?

• Yes, this used to work in a previous version.

Debug info

• Streamlit version: 1.13.0
• Python version: 3.9.16
• Operating System: Linux
• Browser: Chrome

Additional Information

No response

[github-actions]

If this issue affects you, please react with a 👍 (thumbs up emoji) to the initial post.

Your feedback helps us prioritize which bugs to investigate and address first.

[lukasmasuch]

@ShaneTian Thanks for reporting this issue. I was able to reproduce this here. The problem comes from markdown rendering (st.write with text uses st.markdown), which is already tracked here. Please upvote the linked markdown issue, and potentially add your example to the thread.

[lukasmasuch]

Btw. a workaround is to add spaces before all-new line symbols. E.g. via a string replacement on the message.

[ShaneTian]

Btw. a workaround is to add spaces before all-new line symbols. E.g. via a string replacement on the message.

Yes, it is a solution, but is is not a permanent one.

[principalideal0]

There are 62 likes on a linked issue and multiple posts about this - when you ask for likes, do likes thread through from linked discussions in your analytics? You may be underestimating the importance if not. I think anyone dealing with generated output is going to find this to be a gigantic headache ... not a trivial problem.