oh-my-openagent-config
Model configuration profiles for oh-my-openagent (opencode). Each JSON file is a complete oh-my-openagent.json config with different model assignments tuned for cost, capability, or provider diversity.
All profiles use Amazon Bedrock (us-west-2).
Profiles
| Profile | File | Philosophy |
|---|---|---|
| Anthropic | anthropic.json |
All-Anthropic baseline. Opus on core roles, Sonnet on lightweight, Haiku on trivial. Known stable. |
| Mistral Test | mistral-test.json |
Mistral models in low-risk slots (explore, librarian, quick, git). Anthropic on everything else. |
| Qwen Test | qwen-test.json |
Qwen coders in low-risk slots. Tests Qwen streaming compatibility. |
| Multi-Provider | multi-provider.json |
Best validated models across Mistral, Qwen, and Anthropic. The "winners" profile. |
| Budget | budget.json |
Cost-optimized. Sonnet as orchestrator, Opus only for ultrawork escalation. Cheapest viable models elsewhere. |
Usage
Aliases
Add these to your ~/.aliases (or shell rc file):
# OPENCODE PROFILES (oh-my-openagent config swap)
alias omo-anthropic="command cp ~/natera/creid/oh-my-openagent-config/anthropic.json ~/.config/opencode/oh-my-openagent.json && echo 'omo: anthropic'"
alias omo-mistral="command cp ~/natera/creid/oh-my-openagent-config/mistral-test.json ~/.config/opencode/oh-my-openagent.json && echo 'omo: mistral-test'"
alias omo-qwen="command cp ~/natera/creid/oh-my-openagent-config/qwen-test.json ~/.config/opencode/oh-my-openagent.json && echo 'omo: qwen-test'"
alias omo-multi="command cp ~/natera/creid/oh-my-openagent-config/multi-provider.json ~/.config/opencode/oh-my-openagent.json && echo 'omo: multi-provider'"
alias omo-budget="command cp ~/natera/creid/oh-my-openagent-config/budget.json ~/.config/opencode/oh-my-openagent.json && echo 'omo: budget'"
Switching profiles
omo-anthropic # switch to all-Anthropic (safe default)
aws-opencode # launch opencode as usual
Each omo-* alias copies the corresponding JSON into ~/.config/opencode/oh-my-openagent.json, which opencode reads on startup. Switch profiles anytime between sessions.
Testing order (incremental trust ladder)
omo-anthropic— baseline, known goodomo-mistral— test Mistral in low-risk slotsomo-qwen— test Qwen in low-risk slotsomo-multi— combine winners from steps 2-3omo-budget— daily driver once confident