Define model cache locations, size limits, and lifecycle

[ScottMorris]

Goal

Specify where AI models live and how they are managed over time.

Questions to answer

• Where should model files be stored per platform?
• How do we cap disk usage and handle eviction?
• How can users inspect and clear model cache safely?

Scope

• platform-specific cache paths
• cache policy (size cap, LRU/manual cleanup)
• integrity checks and corruption recovery

Acceptance Criteria

• canonical cache path policy documented for Linux/macOS/Windows
• user controls defined for status, prune, and reset
• cache safety and recovery behaviour documented