Reference
CLI for AI
The same engine commands you already know, pointed at checkpoints and datasets.
There is no AI-specific command set. Every command below comes straight from the shared engine and operates on your model artifacts exactly as it would on any other file. This page is the fast lookup; the full engine CLI reference documents every flag.
Everyday commands
init— create a repository in the current directory.add— stage an artifact; chunks and stores it by content hash.commit— record the staged state as a new version.status— show what is staged, modified, or untracked.log— list the commit history.diff— show what changed between versions.checkout— reconstruct a commit's artifacts, byte-exact.branch— create or list branches for parallel runs.tag— name a commit, e.g. a release checkpoint.restore— restore specific paths from a commit.reset— move the current branch or unstage changes.
A typical session
dits init
dits add model.safetensors
dits commit -m "baseline"
dits status
dits log
dits diff
dits branch lora-experiment
dits tag v1.0
dits checkout <commit>
dits restore model.safetensors
dits reset --soft HEAD~1Warning
Networked sync is not here yet.
push, pull, and fetch are on the roadmap — every command above runs locally only.Related
- Full engine CLI reference — every command and flag.
- Quick start and checkpoint workflow.
- Addressing — what
addandcommitdo under the hood.