feat: add Exo distributed AI cluster as LLM provider

[stephanj]

Summary

• Add Exo as a new local LLM provider for running large AI models across multiple Apple Silicon devices connected via Thunderbolt
• Auto-discovers downloaded models from the Exo /state API (only shows models on disk)
• Automatic instance creation with placement preview, background progress bar, and auto-recovery on disconnect
• Collapsible cluster status panel above the chat window showing nodes, memory usage, GPU%, temperature, and active instance status
• Dedicated Docusaurus documentation page with screenshots and troubleshooting guide
• Fix: MarkdownConversationRenderer missing createConversationJEditorPane method

New files

File	Purpose
ExoChatModelFactory.java	Factory creating OpenAI-compatible chat models with async instance preparation
ExoModelService.java	Manages Exo API calls: model discovery, instance lifecycle, cluster state, progress reporting
ExoClusterPanel.java	Custom-painted collapsible panel showing cluster topology and node stats
ExoModelDTO.java / ExoModelEntryDTO.java	DTOs for Exo /models API response
docusaurus/docs/llm-providers/exo.md	Full documentation page with setup, configuration, troubleshooting

Screenshots

Cluster panel in IDE:

Progress bar during instance loading:

Test plan

• Enable Exo in Settings > LLM Providers, verify URL defaults to http://localhost:52415/v1/
• Select Exo provider — only downloaded models appear in dropdown
• Select a model — background progress bar shows instance creation
• Send chat message — response received from Exo cluster
• Verify cluster panel shows nodes with memory/GPU stats
• Click cluster panel header — toggles collapsed/expanded
• Switch to another provider — cluster panel disappears
• Disconnect Exo instance — verify auto-recovery triggers on next chat
• Run ExoChatModelFactoryTest, ExoModelServiceTest, ExoModelDTOTest

🤖 Generated with Claude Code