T3 Code Fork Makes Local Models Default

This is a meaningful shift from AI coding subscriptions toward self-hosted inference, turning T3 Code into a local-first orchestration layer.

• –Registers inference servers and links them directly to Pi, OpenCode, or Codex without manual provider configuration
• –Queues requests per GPU server to prevent parallel coding threads from overwhelming local hardware
• –Adds health monitoring, fallback routing, token metrics, cache statistics, and latency tracking
• –Keeps titles, commits, pull-request text, and other background generation on local models
• –The main caveat is distribution: this is an independent fork, not an upstream T3 Code release