glotfile build-context
AI-generates a short context note for each key — what it’s for, where it appears — by reading the code snippets around its usages. Better context produces better translations. Requires a prior glotfile scan so the usage index exists.
glotfile scan
glotfile build-context
By default it only fills keys that have no context yet. The flags combine to narrow the run.
Options
| Option | Description |
|---|---|
--all |
(Re)build context for every key, not just those missing it. |
--key <glob> |
Only keys matching this glob (e.g. "auth.*"). |
--limit <n> |
Process at most n keys. |
--since <date> |
Only keys added or changed since this date. |
--estimate |
Print the batches, token counts and estimated cost without building. |
--batch |
Submit via the provider’s batch API (50% cost, async; anthropic only). |
--file, -f <path> |
Target a different state file. |
glotfile build-context --estimate # batches, tokens and cost — no API calls
glotfile build-context --key "auth.*" # only the auth namespace
It uses the same AI provider as translate (configured per-machine — see AI Providers). --estimate prints the same token/cost preview translate --estimate does — the input count is dominated by the code snippets attached to each key, so it tracks the real run. If no usage index is found it stops and tells you to run scan first.
Related
- scan · translate · Key Context and Metadata · AI Providers