Git sync — backup settings and sessions via private repo

[egno]

Git sync

Back up Keystrike settings and typing sessions to a private git repository you control. Sync is opt-in, CLI-only, and runs only when you invoke it — there is no Keystrike cloud, no background daemon, and no automatic upload when a session ends.

What it is (and is not)

Is	A git-backed backup and manual multi-device sync via keystrike sync
Is not	Cloud sync, real-time sync, or a Keystrike-hosted service

Your data stays in your git remote (GitHub, GitLab, self-hosted, etc.). Keystrike never talks to that remote except when you run a sync command.

Prerequisites

• Keystrike with git sync support (see CHANGELOG).
• git on your PATH — sync shells out to the git CLI for clone, pull, and push.
• Git credentials configured for your remote:
  • SSH — e.g. git@github.com:you/keystrike-backup.git (SSH key loaded in your agent).
  • HTTPS — e.g. https://github.com/you/keystrike-backup.git (credential helper or personal access token).
• A private remote repository (strongly recommended). Session logs contain your typing history.

Optional but recommended: set git config user.name and git config user.email on each machine so sync commits have sensible authorship.

One-time setup

1. Create a private backup repo

On GitHub (or elsewhere), create an empty private repository, e.g. keystrike-backup. Do not add a README if you want a truly empty repo — an empty repo works fine; Keystrike initializes the layout on first push.

2. Initialize sync on each machine

keystrike sync init git@github.com:you/keystrike-backup.git
# or
keystrike sync init https://github.com/you/keystrike-backup.git

This command:

1. Writes your remote URL to sync.toml (see File locations).
2. Clones the remote (or runs git init locally if the remote is empty).
3. Merges local and remote data using Keystrike's merge rules (below).
4. Pushes the merged snapshot to the remote.

Run init once per machine. On a second device, use the same remote URL so both machines share one backup repo.

Daily workflow

Command	When to use
keystrike sync push	After practice on this machine — upload new sessions and settings.
keystrike sync pull	Before practice on this machine — download sessions from other devices and rebuild local stats.
keystrike sync status	Check remote URL, session counts, and whether the clone has uncommitted changes.

Typical multi-device habit:

1. Before opening Keystrike: keystrike sync pull
2. Practice as usual (the TUI does not sync by itself).
3. After a session (or end of day): keystrike sync push

Example output:

$ keystrike sync status
remote: git@github.com:you/keystrike-backup.git
sessions: local=42 clone=40 (only local=2, only remote=0)
git: (clean)

$ keystrike sync push
push complete

$ keystrike sync pull
pull complete: 3 session(s) imported

What gets synced

Synced	Not synced
settings.toml	cache/ (derived stats — rebuilt locally)
layouts/*.toml (custom layouts)	Bundled layouts shipped with Keystrike
sessions/index.jsonl	Logs outside the sessions tree
sessions/**/*.jsonl (session keystroke logs)

After a successful pull, Keystrike replays all sessions in the merged index and rebuilds the stats cache for every layout it finds. You do not need to sync cache files.

Merge behavior

Keystrike applies its own merge logic before every push and pull. Git is used only for transport; JSONL session files are not left to git's native merge.

Sessions — union by session_id

• Sessions are identified by session_id in sessions/index.jsonl.
• Missing sessions are imported — if the remote has a session ID your machine lacks, the session file and index entry are copied in.
• Existing IDs are never overwritten — if both sides have the same session_id, the local copy wins and the remote copy is left unchanged on import.

This means you can practice on two machines offline and merge later without losing sessions, as long as each session has a unique ID (Keystrike generates UUIDs).

Settings — last-write-wins

• Keystrike compares updated_at in settings.toml (ISO timestamp, written on every save).
• The newer file is copied to both local and clone before push/pull.
• If updated_at is missing on either side, file modification time is used instead.

Whichever machine saved settings most recently wins globally after the next sync.

Layouts — additive on pull, full copy on push

• Pull: remote layout TOML files that are missing locally are copied in. Existing local files with the same name are not overwritten.
• Push: all local custom layout files are copied into the clone (overwriting same-named files in the clone).

If two machines edit the same layout filename differently, the last push wins on the remote; the other machine won't see those edits until it pulls and may keep its local copy until you resolve the conflict manually.

Multi-device tips

1. Pull before push on each machine to reduce divergent git history.
2. Use a private remote — session JSONL is sensitive.
3. Same remote on every device — one backup repo for all machines.
4. Check status when unsure: only local / only remote counts show what still needs a push or pull.

Known limitations

• Fast-forward pull only — keystrike sync pull runs git pull --ff-only. If you pushed from two machines without pulling in between, the remote history can diverge and pull will fail. Fix on the clone: cd ~/.config/keystrike/sync/repo (paths vary by OS), resolve with git pull --rebase or merge manually, then retry keystrike sync pull.
• No automatic sync — nothing runs when a session ends; you must run push/pull yourself.
• Layout name conflicts — same filename, different content: last push to remote wins; pull does not overwrite existing local layouts.
• Settings conflicts — no three-way merge; newest updated_at (or mtime) wins outright.
• Session duplicates — same session_id on both sides keeps the local version on import; design assumes IDs are unique.

File locations

Paths use platformdirs (same roots as other Keystrike data):

File / directory	Purpose
{config_dir}/sync.toml	Remote URL written by sync init
{config_dir}/sync/repo/	Local git clone used for push/pull
{config_dir}/settings.toml	Synced settings
{config_dir}/layouts/	Synced custom layouts
{data_dir}/sessions/	Synced session logs

Typical {config_dir} / {data_dir}:

OS	Config	Data (sessions)
Linux	~/.config/keystrike/	~/.local/share/keystrike/
macOS	~/Library/Application Support/keystrike/	same tree under Application Support
Windows	%APPDATA%\keystrike\	%LOCALAPPDATA%\keystrike\

sync.toml contains only the remote URL, for example:

remote_url = "git@github.com:you/keystrike-backup.git"

Troubleshooting

sync not configured

Run one-time setup:

keystrike sync init <repo-url>

Authentication failed (SSH)

• Test: ssh -T git@github.com
• Ensure your SSH key is loaded (ssh-add -l) and added to your GitHub account.

Authentication failed (HTTPS)

• Use a credential helper or embed a PAT in the URL (less ideal): https://<token>@github.com/you/keystrike-backup.git
• Prefer gh auth login or Git Credential Manager.

Empty or new remote

sync init handles empty remotes: if git clone fails, Keystrike runs git init, adds origin, merges local data, and pushes.

git pull --ff-only failed / diverged history

Two machines pushed without pulling. On the affected machine:

cd ~/.config/keystrike/sync/repo   # adjust for your OS
git fetch origin
git log --oneline --graph --all -5   # inspect
git pull --rebase origin main        # or master — match your default branch
keystrike sync pull

When in doubt, back up {config_dir} and {data_dir} before fixing git history manually.

git commit failed — user.name / user.email

Configure git on that machine:

git config --global user.name "Your Name"
git config --global user.email "you@example.com"

Then retry keystrike sync push.

Pull succeeded but stats look stale

Pull rebuilds aggregates from the session index. If something still looks wrong, check keystrike sync status session counts and confirm session files exist under {data_dir}/sessions/.

See also

• Keystrike README — Backup sync