model-selection

name	model-selection
description	Get the LLM models for evolutions from the workspace agent subscription tier config
version	2.0.0
author	kai-agent
metadata	{"kai":{"tags":["kai","evolve","models","subscription","tier"]}}

Model Selection for Evolutions

Evolution models are defined by the workspace's agent subscription tier config. Do NOT pick models yourself — the server enforces the tier's model list regardless of what you pass.

How to get the models

Call check_agent_subscription (in the budget category) to see the tier config:

activate_category("budget")
check_agent_subscription(workspaceId)

The response includes a models field:

{
  "models": {
    "evolve": ["anthropic/claude-sonnet-4.6", "openai/gpt-4.1"],
    "security": { "root": "...", "analyzer": "...", "verifier": "..." },
    "agent": "anthropic/claude-opus-4.6"
  }
}

Use the models.evolve array in your config.llm.models when calling start_code_optimization. The server will override your choices with these models anyway, but passing them avoids confusion in logs.

Rules

1. NEVER hardcode model names — always read from the subscription tier
2. The server enforces models — whatever you pass in config.llm.models gets overridden by the tier's evolve list
3. Weights are yours to set — the server preserves the weights you pass, only the model names are overridden
4. If no subscription exists, the server uses whatever models you pass as a fallback