Repository navigation
Log Streaming For gh run watch #3484
Description
Activity
Thanks for the issue :) this is definitely in our roadmap; it would have been in the 1.9.0 ship but unfortunately there's no API support yet for streaming logs and as such is blocked for now.
Reacted by Austin Cawley-Edwards, Alessandro Dal Grande, Ben Plunkert, Noah Masur, Mahmoud Hanafy, Karol Baraniecki and Nathan S. MartinsOh, that's interesting – how does the web UI do it?
Reacted by Hamza PavlovičReacted by Rareș CosmaI think the value of this issue is beyond active feedback: I often find myself trying to grep for particular issues, my current workaround is to download the raw logs and then search.
While the web UI is very neat and slick, it breaks down for very large logs.
I agree this would be a very very useful feature.
Maybe this issue could be tagged
actions?Reacted by Carlos IriarteWould very much love that as well.
Reacted by Carlos IriarteWould the maintainers be open to a PR that parsed the web UI instead of using the API?
Actually, we would just use the endpoint github uses to pull the logs?
https://github.com/deepdivr/deepdivr/actions/runs/<run_id>/jobs/<job_id>=/steps?change_id=<change_id>change_idseem to just be an increasing number.@mollerhoj I think that using an endpoint not in the API is too much risk as it can change at any moment without us being informed and there are no guarantees as to what it might return. I would rather wait for an official API endpoint. @vilmibm do you agree?
I'm not comfortable with parsing the web UI or with using unpublished API endpoints.
fair enough :-)
@samcoe @vilmibm @mislav Is it planned (when?) to add the API the CLI needs to do this? Could not see anything relevant at https://github.com/github/roadmap/issues?q=is%3Aissue+logs :/
Would love this <3
Reacted by Hamza Pavlovič, 56k-modem, FireFragment and Nathan S. MartinsReacted by Austin Cawley-EdwardsPlease implement this. Tired of accessing the UI just for this...
Reacted by bryan hunt, 56k-modem, James Cuzella, FireFragment and Nathan S. MartinsReacted by Austin Cawley-Edwards#6741 might be a duplicate.
Any update on this?
Reacted by Taro, Hamza Pavlovič, spezifisch, Armaiz Adenwala, James Cuzella, FireFragment, Phani Rithvij, Tobias Gårdhus and Nathan S. Martins+1 - would love to add real-time logs for a cli command!
+1 from me too. Having to go into
gh run view --log-failedis a step I'd like to eliminate. At the very least, include the recommended command for viewing the failure - in thewatchoutput (as it appears in theviewoutput)
I was looking for a way to do just that, that would be great!
Reacted by Hamza Pavlovič, Roy Mohanan and Nathan S. MartinsI wonder if it be against the ToS to build a tool that sniffs the authenticated WebSockets subscriptions of a mitm-proxied headless browser and start some lightweight ws clients to perform the actual live tailing...
JK - I actually had a stab at it: https://github.com/rarescosma/octotail/
Reacted by Peter Von der Porten and Nathan S. MartinsReacted by Karol Baraniecki and Dolev HadarReacted by Nicolai Weitkemper, Patrik Ragnarsson, James Cuzella, Josh Taylor and Mykola RybakPretty surprised this is actually not implemented yet. Kudos to @rarescosma for his initiative.
Reacted by Rareș Cosma, Hamza Pavlovič, Michael Simmons, John Hamelink, Ivan Akulov and Nik Bis there an issue for the blocking API gap that we can follow?
Reacted by Vlad Kolbaia- added a commit that references this issue
on Jul 31, 2026 Thanks for the issue :) this is definitely in our roadmap; it would have been in the 1.9.0 ship but unfortunately there's no API support yet for streaming logs and as such is blocked for now.
Describe the feature or problem you’d like to solve
The new
gh runactions are awesome, love them! The utility ofgh run watchis very promising, but the slow feedback (when you have long CI runs) makes it more of a passive "put in another window" command, instead of an active "follow along for the status" command for me. (I know, 1.9.0 just came out and already asking for more 🙄 )Proposed solution
I'm not sure if it's already on your roadmap, but streaming the logs from the run would make this feature usable for active feedback for CI runs, instead of having to go to the web UI to check failures, etc.
Additional context]
From the above screenshot, it's a great summary but I can't take action from here.