Configuration And CLI
Serving Content From Git Refs URL copied
Vyasa can serve any committed branch or tag of a content repo through a ?ref= query (for example ?ref=origin/feat/my-branch), reading files straight from git objects without a checkout. For those pages to show the latest commit, the local copy of the repo has to be fetched first — otherwise a reload keeps showing whatever commit was last pulled.
A lone vyasa process can run the fetcher itself on a background thread, keeping configured mirrors (git_repos) and clone-backed content roots fresh. Cadence is set by git_fetch_interval (seconds, default 30); 0 disables it.
Run production with
--no-reload.--reload(the default) watches the content tree. Every git fetch writes into.git, the watcher sees it and restarts the worker, the fetcher respawns and refetches — a loop that pins CPU/memory until the machine dies. To prevent that, the in-process fetcher refuses to start under--reload(it logs a line saying so), and ref pages then only update via the branch-menu button. For automatic freshness in production, start with--no-reload; keep--reloadfor local dev only. If you must keep--reload, runvyasa-fetchas a separate process instead and setgit_fetch_interval = 0.
| Setting | Why it exists |
|---|---|
git_repos |
Upstream repos to mirror, as { name = "url" }, exposed as content roots. |
git_fetch_interval |
Seconds between automatic in-process fetches. 0 disables it (use a vyasa-fetch sidecar, and always 0 under --reload). |
Configuration And CLINavigate by section | |
| Slide 2 | |
| Start Here | |
| Put Stable Choices In `.vyasa` | |
| Why These Settings Matter | |
| Serving Content From Git Refs | |
| Keep In Mind | |