Files
npub1mn7jgtj4w2pd0g0zeuhxsa6jy6p0rewxz4kujt98my82ahfmp72sxjexk7andWill Pfleger fd9270e8fd fix(desktop): address Wes review comments on #1307
FIX 1 — extract agent_env.rs to fix file-size guard:
runtime.rs was 2172 lines (gate limit 2150). Extracted
build_buzz_agent_provider_defaults + parse_agent_env_lines into a new
focused module managed_agents/agent_env.rs. Moved the 10 associated
tests (parse_agent_env_lines_*, buzz_agent_provider_defaults_*,
base64_round_trip, baked_defaults_ordering) into agent_env.rs's own
#[cfg(test)] mod. Re-exported build_buzz_agent_provider_defaults via
mod.rs pub(crate) use so agent_models.rs call site is unchanged.
runtime.rs is now 2120 lines (gate sees 2121); check-file-sizes.mjs
passes.

FIX 2 — rewrite stale E2E spec:
persona-env-vars.spec.ts 'persona model options follow the selected LLM
provider' asserted the deleted static roster (GPT-5.5, Claude Sonnet
4.6) and clicked 'Databricks (default)' (now just 'Default'). Rewrote
to assert provider-agnostic behavior: Default model present without
discovery, provider switch updates API-key field label/value, Default
provider resets model to 'Default model'. No hardcoded model IDs remain.
E2E harness requires a built app (webServer) — not runnable locally
without dist; CI validates.

FIX 3 — drop dead providers filter branch:
With the static roster gone, PersonaModelOption.providers was always
undefined. Removed the providers field from the type and simplified
getPersonaModelOptions to drop the dead option.providers?.includes()
branch. typecheck clean.

Co-authored-by: Will Pfleger <pfleger.will@gmail.com>
Signed-off-by: Will Pfleger <pfleger.will@gmail.com>
2026-06-29 19:49:16 -04:00
..