ory / lumen / 24913223801

74%

push

github

feat(cmd): select configured server via --model/--backend (#137)

Previously `--model X` just set LUMEN_EMBED_MODEL, which `applyEnvOverrides`
merged into servers[0] — so passing a model configured on a non-primary
server silently ran it against the wrong backend. Unknown model names were
also rejected outright even when present in the user's config.yaml.

`--model` and a new `--backend` flag now filter the YAML server list via
a new `config.WithServerSelection` option. Failover is preserved within
the filtered subset. For users without YAML, the legacy WithModelOverride
path still applies when the model is in KnownModels and --backend is unset.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

63 of 69 new or added lines in 3 files covered. (91.3%)

3523 of 4777 relevant lines covered (73.75%)

764.98 hits per line