What gets synced
Which files a commit takes, which stay on the machine that wrote them, and what happens to a file over the size limit.
zeon commit (and zeon sync, which commits for you) takes every file in the project folder except the ones below.
Never scanned
Folders with these names, anywhere in the project: .zeon, .git, .venv, __pycache__, node_modules, .ruff_cache, .pytest_cache, .mypy_cache.
Run folders: the log travels, the run state stays
Runs write into the project's executions/<run id>/ folder on the machine that runs them. From a run folder, a commit takes only:
- the run log:
run_log_<run id>.jsonl(orrun_log.jsonl); metadata.json.
Everything else in a run folder (checkpoints, resume state, world snapshots, subfolders) is the run's state. It stays on the machine that ran it: it is never committed or pushed, and a clone, pull or checkout never writes it into a working tree. A resume on another machine therefore cannot pick up a run's state by accident.
If an older CLI pushed run state, zeon status lists those files under Leaving the project and the next commit drops them from the project. The files stay on your disk.
Credentials: .env files
.env filesA file named .env or .env.*, at any depth, is never committed or pushed, and a clone, pull or checkout never writes or deletes one. The match is by name only.
Other local-only folders
| Path | Synced |
|---|---|
.runtime/ | Never. |
data/runs/ | Never. |
the rest of data/ | Yes, in both directions. |
data/ is the project's shared data channel: files a skill writes there (captures, logs, results of external API calls) and files you put there for skills to read. When a run started from the web app ends, the app pushes the project's data/ folder (except data/runs/) to the cloud project, so zeon pull brings the run's output to your computer.
Leaving out more: .zeonignore
.zeonignoreA .zeonignore file in the project root lists more files for a commit to leave out, in .gitignore syntax. As with git, it affects only new files: a file the project already has stays tracked until you delete it. It cannot bring back what the rules on this page leave out. .zeonignore itself is synced with the project.
Plan caches
Skill folders can contain a motion plan cache file, .motion_plan_cache.pkl, written when skills run. It is kept: it syncs like any other file, within the size limit.
Kept on checkout
zeon checkout and zeon reset --hard write a commit's files into your working tree and delete the files the commit does not have, except anything under executions/, data/ and .runtime/, plan cache files, and files over the size limit.
The size limit: 64 MB per file
The cloud stores files up to 64 MB each. A larger file is left out of the commit and named:
left out data/scan.ply (80.5 MB, over the 64 MB sync limit): not in the project
If the parent commit already has a version of the file, the project keeps that version and the message ends in the project keeps its last committed version instead. zeon status lists such files under Left out (over the 64 MB sync limit). A pull never overwrites or deletes your local copy of such a file; it prints kept your local <path>.
If the cloud refuses a file during a push, nothing is pushed and the error names the file. To push without it: zeon reset <remote head> (moves HEAD back and keeps your files), move the file out of the project, then commit and push again.
Hidden files are synced like any other, except
.envand.env.*files. Keep other credentials out of the project folder.zeon loginstores the token in~/.zeon/.env.
Updated 1 day ago