Use when a user wants to add or update Google provider models in opencode.json.
This skill adds or updates Google AI model definitions in the OpenCode configuration file (~/.config/opencode/opencode.json), under the provider.google key.
Trigger this skill when the user asks to "add Google models to OpenCode" or similar.
To use Google models with OpenCode, you must enable the Gemini API in your Google Cloud project:
cloudaicompanion.googleapis.com) and click Enable.Read Current Config:
~/.config/opencode/opencode.json.provider.google key.Fetch Latest Reference Configuration:
Identify Models to Add/Update:
provider.google.models.provider.google.models.<model-id>.id, name, reasoning, release_date, limit (context, output), cost (input, output, cache_read), modalities, and variants (thinking-level presets).Merge Strategy:
provider.google does not exist, create it with "npm": "@ai-sdk/google" and an empty models object first.release_date field for all models during the merge.release_date).Write Back:
~/.config/opencode/opencode.json, pretty-printed with 2-space indent.Verify:
provider.google.models.The canonical Google provider structure based on this example:
{
"plugin": ["opencode-google-antigravity-auth"],
"provider": {
"google": {
"npm": "@ai-sdk/google",
"models": {
"gemini-3-pro-preview": {
"id": "gemini-3-pro-preview",
"name": "Gemini 3 Pro",
"reasoning": true,
"limit": { "context": 1000000, "output": 64000 },
"cost": { "input": 2, "output": 12, "cache_read": 0.2 },
"modalities": {
"input": ["text", "image", "video", "audio", "pdf"],
"output": ["text"]
},
"variants": {
"low": { ... },
"medium": { ... },
"high": { ... }
}
}
}
}
}
}
npm field must always be "@ai-sdk/google".variants define thinking-level presets (minimal, low, medium, high). Not all models support all levels.gemini-2.5-flash-lite) can omit limit, cost, modalities, and variants.