[Feature]: have predefined ocx profile?

[qwertystars]

Prerequisites

• I have searched existing issues and discussions to avoid duplicates
• This feature request is specific to oh-my-opencode (not OpenCode core)
• I have read the documentation

Problem Description

As a user,
When i hit a rate limit,
Then i need to manually change the json config which is tedious when i hit rate limits a lot.

I want to be able to make predefined profiles and use the ocx command to switch configs as easily as possible.

Proposed Solution

After removing of all hard coded llm that is #652 is merged. Make a repository for oh-my-opencode ocx profiles. Possibly integrate it with a ui for selecting the models which automatically created new ocx profiles.

Alternatives Considered

No response

Doctor Output (Optional)

Additional Context

No response

Feature Type

Configuration Option

Contribution

• I'm willing to submit a PR for this feature
• I can help with testing
• I can help with documentation

[qwertystars]

Summary

Successfully merged two PRs locally and resolved issues:

PRs Merged

• feat(config): add model variant support #677 (fix/add-variant-support): Adds optional model variant support to agents and categories
• fix(prometheus): prevent agent fallback to build in background tasks #695 (fix/prometheus-session-agent-v3): Fixes Prometheus agent fallback to Build when background tasks are injected

Merge Conflict Resolution

File: src/index.ts:448-449

Both PRs added cleanup logic in the session.deleted event handler. Resolved by keeping both cleanup calls:
if (sessionInfo?.id) {
firstMessageVariantGate.clear(sessionInfo.id); // from #677
clearSessionAgent(sessionInfo.id); // from #695
await skillMcpManager.disconnectSession(sessionInfo.id);
await lspManager.cleanupTempDirectoryClients();
}

Build Status

✅ Build successful after merge

Issue Discovered: Hardcoded Model Fallbacks

Found that several agents fall back to hardcoded anthropic/* models when not explicitly configured:

• Prometheus (Planner) → anthropic/claude-opus-4-5 (src/plugin-handlers/config-handler.ts:206)
• Sisyphus-Junior → anthropic/claude-sonnet-4-5 (src/plugin-handlers/config-handler.ts:157)
• Other agents via DEFAULT_CATEGORIES (src/tools/sisyphus-task/constants.ts)

Solution: Users must explicitly configure all agents in oh-my-opencode.json to override these defaults, especially when using non-Anthropic providers.

Testing

Tested with Antigravity provider - all agents now resolve to valid models.

[KNN-07]

#629 Merged, Kenny PR, should actually work with ocx now

[kdcokenny]

Prerequisites

• I have searched existing issues and discussions to avoid duplicates[x] This feature request is specific to oh-my-opencode (not OpenCode core)[x] I have read the documentation

Problem Description

As a user, When i hit a rate limit, Then i need to manually change the json config which is tedious when i hit rate limits a lot.

I want to be able to make predefined profiles and use the ocx command to switch configs as easily as possible.

Proposed Solution

After removing of all hard coded llm that is #652 is merged. Make a repository for oh-my-opencode ocx profiles. Possibly integrate it with a ui for selecting the models which automatically created new ocx profiles.

Alternatives Considered

No response

Doctor Output (Optional)

Additional Context

No response

Feature Type

Configuration Option

Contribution

• I'm willing to submit a PR for this feature[x] I can help with testing[x] I can help with documentation

Love this idea!

[qwertystars]

PR #866 adds the model-config command which is a step towards this feature.

What it does now:

• Detects available models via opencode models
• Generates optimal agent→model and category→model configuration based on static rankings
• --apply flag writes config to oh-my-opencode.json (respects OPENCODE_CONFIG_DIR, XDG, APPDATA)

How this helps with OCX profiles:
The model-config command provides the foundation for generating different configs. Since it now properly respects OPENCODE_CONFIG_DIR, you could potentially:

1. Set OPENCODE_CONFIG_DIR to different profile directories
2. Run model-config --apply to generate optimized config for each profile
3. Use OCX to switch between them

Next steps to fully support this:

• Add --profile <name> flag to write to specific profile directories
• Add --provider <anthropic|openai|google> flag to generate configs prioritizing specific providers (useful when one hits rate limits)
• Consider a repository of community-contributed profile templates

cc @code-yeongyu