[2/7][multi-lora]: adapter controller - registry state machine, backend, control-plane HTTP API, named Ray actor

[mathewjhan]

#1638

[gemini-code-assist]

Code Review

This pull request introduces a multi-LoRA training framework, implementing a controller Ray actor, an HTTP control-plane server, an adapter registry, and configuration parsing to manage the lifecycle of concurrent LoRA adapters. The code review identified several critical issues and improvement opportunities: a potential stale actor handle caching bug in the controller, possible crashes in the HTTP server when handling null configurations or malformed YAML files, silent validation bypasses for empty data paths, the risky use of assert statements for production argument validation, and unnecessarily long timeouts on time-sensitive abort requests.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[yushengsu-thu]

Reviewed extensively on the umbrella PR #1638 (2xH200 E2E validated); layer verified independently (imports + fast suites green on this branch's own tree).