[ ] Should we generate 'artifact_upload' events? What for? Mail notifications? Watches?
[ ] I should have two namespaces for artifacts: user and repo.
	Think about pushing into a repo to update the user wiki.
	Shouldn't we just set the wiki to point to the repo artifacts dir?
	Think about.
[ ] We should not allow an artifacts_upload_done without an active file.
[ ] I need to set the temporary name, save the state and only after that start
	writing to it. Else, I will pollute the fs.
[ ] On success/error, allow the calling of another webhook?
	This way we can build a complex pipeline.
	What about labels and evaluating sequentialy all hooks?
[ ] Add an optional timeout for the commands.
[ ] Look at some pipeline presentation to be inspired.
[ ] If ANN/anwer is not received in a resonable time, close the connection.
[ ] If a worker goes down, after a timeout we should remove the jobs from it
	if there are other valid workers. And we will not accept anything from
	that worker? If we start the job in other place, yes, we will not accept.
	Maybe in the future, we can abort the second worker if the first comes
	online and it is done.
[ ] Show in the web interface what worker built a job? Already done?
[ ] sent_jobs: shouldn't we save this information to be able to recover?
[ ] We need to store in meta indo also the job_id, to check all details about that build.
[ ]