An LLM-powered Sublime Text 4 plugin for inline code completion and chat-about-selection. Works with any OpenAI-compatible /v1/chat/completions endpoint like OpenAI, LMStudio, Ollama, vLLM, and others.

Fixes

• API key now sent with completion requests — cloud endpoints (OpenAI, etc.) work
• Default timeout fallback corrected (was 0.15s, now 30s)
• HTML-escape phantom preview so <, >, & in completions render correctly (C++ templates, JSX)
• First-run wizard prompts for an optional API key
• Multi-line accept preserves the model's intended indentation
• Wider model control-token cleanup (Llama 3, ChatML, FIM, GPT-style)
• CodeContinue: Configure no longer crashes when run from the Command Palette
• Errors visible in the Sublime console without enabling debug
• Network/parse error messages reliably reach the status bar