zyratlo opened a new issue, #7931:
URL: https://github.com/apache/texera/issues/7931

   ### Feature Summary
   
   The notebook migration tool converts a Jupyter notebook into a Texera 
workflow. The backend microservice and the workspace UI have both landed (#5258 
and the follow-up frontend work under #4301), but neither is part of any 
deployment. A user who runs Texera through the single-node Docker Compose stack 
cannot use the feature at all: nothing starts the service that serves its 
endpoints, and there is no JupyterLab server for the workspace to embed. Today 
the tool only works if a developer starts the pieces by hand.
   
   ### Proposed Solution or Design
   
   Run the notebook migration service and a JupyterLab server as part of the 
single-node stack, and route the tool's API traffic through the existing 
reverse proxy so the frontend reaches it on the same origin as every other 
service.
   
   Two behaviors this needs to get right:
   
   1. The browser loads the JupyterLab server directly, because the workspace 
embeds it in an iframe. The address handed to the browser is therefore not the 
address the backend itself uses to reach Jupyter inside the deployment network. 
Both need to be correct at the same time.
   2. The conversion is powered by an LLM, so it depends on the same API key 
configuration the Texera agent already uses. With no key configured, the tool 
should still appear and fail with a clear provider error, matching how the 
agent behaves today.
   
   An operator should be able to turn the feature off from deployment 
configuration, and the port exposed for JupyterLab should be configurable like 
the other published ports. The deployment documentation should cover the new 
port, how to enable or disable the tool, and the security note below.
   
   Scope and limitations:
   
   * This covers the single-node Docker Compose deployment only. Kubernetes 
deployment is out of scope.
   * Support in the local development tooling is a separate follow-up, since it 
depends on this work landing first.
   * Every user of a single-node deployment shares one JupyterLab server and 
one access token, so anyone who can reach the deployment sees the same 
notebooks. That is acceptable on a single machine the operator controls, and it 
must not be carried into a multi-user deployment.
   
   ### Affected Area
   
   Deployment / Infrastructure


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to