v5.3.13 — Pick Your Model Once, Everything Follows

When you install NEXO Brain, you pick your model. Until today, that choice was only half-respected: interactive sessions used your model, but automation scripts — the ones running in the background doing real work — had their own hardcoded defaults.

The Problem

Eleven core automation scripts (deep sleep, evolution, immune system, daily self-audit, synthesis, catchup, and others) hardcoded model="opus" or model="sonnet". The fast task profile defaulted to a completely different provider. If you changed your model preference, background automation ignored it.

The Fix

v5.3.12 added resolve_user_model() — a single function that returns the model you configured. All automation profiles and task backends now inherit from it instead of hardcoding strings.

v5.3.13 migrated all 11 core scripts to use it. The fast profile no longer defaults to a different backend. Every component respects your choice.

For Script Authors

If you write personal automation scripts, use:

from client_preferences import resolve_user_model
model = resolve_user_model()  # Returns whatever the user configured

Never hardcode a model string. The user picked theirs for a reason.

Update

npx nexo-brain init
# or, for existing git installs:
nexo update