Submit. Tag. Forget.
Qoffee is a free and open-source quantum job monitoring tool that notifies you about your job's status, without you ever needing to open a laptop, refresh a dashboard, or check a queue.
It runs as a GitHub Action inside your own forked copy of the repository. There is no server, no account, and no database. You add one line to the script you already use to submit jobs.
d3k9p2QUEUED on ibm_torino
a71fe4DONE on ibm_brisbane
c02b8aERROR on ibm_torino
The line you add
Qoffee tracks a job because the job carries a tag you gave it. Add the tag to whatever script you already use.
sampler = Sampler(backend)
sampler.options.environment.job_tags = ["qoffee"]
job = sampler.run([isa_circuit])
That is the whole integration. You can also tag a job after you submitted it, and you can give it a readable name. Both are on the tagging page.
One day, three jobs
Here is what actually arrives if you submit three jobs in the morning, one of them fails, and you do not look at anything until the evening. The dots on the left mark the runs where Qoffee sent something.
qoffee.Bell test / QUEUED
VQE sweep / QUEUED
Noise cal / QUEUED
Bell test / RUNNING
VQE sweep / QUEUED
Noise cal / QUEUED
Done (1) / Bell test
Failed (1) / VQE sweep
Failed (1) / VQE sweep
Five messages, over seven and a half hours, from thirty scheduled runs. The last one contains the failure and nothing else, which is the job you need to resubmit. After that, Qoffee goes silent until you tag something new.
The rules that produce this are on when Qoffee talks.
What it runs on
GitHub Actions, in a repository you own. Public repositories get unlimited Actions minutes, so the schedule costs nothing. Discord webhooks are free. The optional Cloudflare scheduler is free.
Nothing about you is stored anywhere. There is no database, no state file, and no jobs list. The tracking state lives in IBM's own job tags, which is why the repository can be public without leaking who is using it. That is explained on how Qoffee works.
Everything here is checkable
Qoffee is written in Python and the repository is public.
Every claim on this site corresponds to something in the source, and the pages below link to the specific file where it lives.
You can also run python -m qoffee --dry-run before you let it change anything.
Setup is three steps and takes about five minutes. It ends with you triggering a run by hand and watching the first message arrive.