lingo pull
Retrieve the latest locally recorded run without overwriting divergent target edits accidentally.
After a push without --wait, run lingo pull from the same project root and machine. It reads the latest-run pointer for that absolute path and collects the recorded run. There is no arbitrary run-ID argument or automatic search through project history.
lingo pull --dry-run
lingo pullDry run reports write decisions without changing files or the lockfile. It can still contact the server and poll unfinished work. --force (-f) permits overwriting divergent local targets; preserve those edits before using it. Both options default to false.
How local edits affect collection #
Pull compares each target's current bytes with the server result, the target captured at push, and the lockfile:
| Current target | Decision |
|---|---|
| Absent | Create it. |
| Matches completed server artifact | Already synchronized. |
| Matches target captured at push | Safe to replace with the result. |
| Matches recorded lock hash | Safe to replace. |
| Matches none of those | Conflict. |
Without force or dry run, any conflict aborts all target writes. Committing an edit does not change its bytes or resolve this comparison. Preserve and reconcile the edit, or restore a known version before collecting.
Successful collection records completed-target hashes. It advances an existing source baseline only when the completed document covers every configured target locale and has no key scope. Collection for fewer locales keeps the existing source hash, or initializes it if absent. A key refresh records target hashes only, so pending ordinary source edits still reach the next plain push. Direct push --wait records the same hashes but does not run this fresh conflict check.
Read unfinished and failed results #
Pull polls pending work until every unfinished target is done or waiting for human review, then collects per document: a document is written once all of its locales are done. Documents still waiting for review are left alone, and running lingo pull again after the reviewers finish writes the rest.
A failed run still writes the documents that completed, then reports the failure and exits nonzero. A document with a failed locale writes none of its locales and keeps its source baseline, so the next push of that file requests it again. Read the failure and verify expected files. A missing translation or required review needs its own workflow action, not another download attempt.
When the run cannot be found #
The pointer lives under ~/.lingo/runs/ and depends on the absolute project root. Moving a checkout, using another runner, or submitting another push can change which run is available locally. Check that state and the project binding before concluding remote output is missing.
A run submitted by an older CLI version may be unreadable to the installed one; pull then says so, and another lingo push submits the work again. Lockfile and cache explains the portable baseline and machine-local state.