nexusai 0.2.9

Add your description here

Immediate TODO

• jobs are still being added to blacklisted gpus
• fix screen attaching not showing anything
• merge stdout and stderr
• remove spaces from command combinations
• freeze git states when adding jobs
• if -r in command it bugs out thinking its repeat
• runtime doesnt update if gpu is blacklisted
• Classify failed jobs
• automatically detect wandb runs
• Webhooks for starting, completed, and failed jobs (for discord)
• make webhooks prettier
• if a job doesnt have wandb after 5 minutes, stop pinging it
• clean up git tags that are unused
• make sure the job started webhook waits
• multi user support
• cli: prompt yes or no before removing, killing, or adding
• cli: put wandb url in nexus status at cli
• cli: put runtime and time started on same line
• cli: filter history with command regex
• cli: in history put failed and completed together, then I shuold be able to see which and which completed
• its starting jobs on non available gpus
• actually get gpu processes
• cli: follow logs with -f
• figure out why git tag removal fails some times
• installation walkthrough with the service
• figure out why job failed webhook failes sometimes
• cli: More job details when printing job stuff, especially when removing or killing
• Automatically check for updates
• multi gpu jobs
• dependent jobs (a after b is done)
• cli: ensure cli and api version align
• refactor, move more things away from cli and to the api
• sqlite for state management

Longterm TODO

• systemd service
• multi node
• priority jobs
• vram / flop minimum
• rust rewrite for static binaries that don't require python (or venvs to be activated)
• pretty TUI front end