Minimal isolated OpenCode profile for the local llama.cpp debugging workflow.
This config is intended to be deployed to ~/.config/opencode-llama/opencode and launched with an opencode-llama wrapper that points OpenCode at isolated XDG config/data/state/cache roots.
Goals:
- keep the default agent on
llama - avoid loading the normal OpenCode MCP-heavy profile
- preserve only the local provider config needed for llama.cpp debugging
- route OpenCode directly to
http://hero.makeitwork.cloud:8080/v1; the Forge proxy remains deployed on hero for separate experiments but is not stable enough for this profile's multi-step tool loop