Error: can not read conda_channels.txt


I started to get the error in the title. My app was working before. I am installing Rdkit from conda as explained here by @randyzwitch : GitHub - randyzwitch/test-conda

My repo: GitHub - mcsorkun/test-streamlit
App link:

any suggestions?


The problem looks like it doesn’t read the channels from conda_channels.txt
I have a new error now like that:

[manager] Python dependencies were installed from /app/test-streamlit-2/requirements.txt using pip.

Collecting package metadata (current_repodata.json): ...working... [2021-04-19 12:40:55.078079] done

Solving environment: ...working... [2021-04-19 12:40:55.078382] failed with initial frozen solve. Retrying with flexible solve.

Collecting package metadata (repodata.json): ...working... [2021-04-19 12:40:59.689719] done

Solving environment: ...working... [2021-04-19 12:40:59.690214] failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:
  - libgcc==5.2.0
  - rdkit

Current channels:

To search for alternate channels that may provide the conda package you're
looking for, navigate to
and use the search bar at the top of the page.

[manager] could not verify installer exit code: open /app/install_status: no such file or directory

Any help?

Hi @mcsorkun, welcome to the Streamlit community!

I’ve sent a note to our engineering team to try and understand if something has changed. Once I hear back, I’ll let you know.


I see the issue…you are using both requirements.txt and the conda install process. Streamlit sharing will pick one of them, but doesn’t guarantee which one it reads.

If you move your requirements.txt packages into the conda.txt, then delete the requirements.txt file, I suspect it will work.


Hello @randyzwitch and thanks for your reply!

I just tried your suggestion but it did not work. What I see is Streamlit sharing reads both conda.txt and requirements.txt but not conda_channels.txt.

Yeah, looks like my dlib example no longer works either. Hopefully our engineering team will be able to shed some light on this quickly.

Thanks, looking forward to it.

I have carried all the requirements and dependencies into the environment.yml file as suggested at the following topic: Managing your Streamlit dependencies using conda


name: aqsolpred
- rdkit
- conda-forge
- defaults
    - libgcc=5.2.0
    - rdkit
    - pip
    - pip:
        - pandas==1.0.1
        - numpy==1.18.1
        - scikit-learn==0.22.2.post1
        - mordred
        - xgboost==1.0.2
        - streamlit==0.69.2
1 Like

Thanks for testing that out, and confirming it works! I’ll have to update my GitHub examples to ensure the new syntax.