GPT-5.6 Model & Effort Advisor

Help pick the right model tier and reasoning effort for a task, escalate only when needed, learn from real outcomes, and diagnose failures. It recommends and records; the user decides and switches.

Two independent axes

• Model tier = intelligence ceiling. A larger tier knows more and reasons deeper, at a higher price per token. GPT-5.6 tiers: Luna (economy) → Terra (balanced) → Sol (frontier).
• Reasoning effort = deliberation time. More effort spends more thinking on the same problem, at more tokens. Levels: Low → Medium → High → Max → Ultra.

These are different dials. Turning both to the maximum by default wastes money without improving quality on most tasks. Choose each axis deliberately.

Engine: registry.py

registry.py (same folder, standard library only) powers the three data functions. Always use this CLI — do not hand-edit the JSON.