Run a council of LLM value agents over your journal to surface decisions you have not made. Local-first via Ollama, OpenRouter, or any OpenAI-compatible endpoint.