What's Changed
11 New AI Providers (16 Total, need more testing):
Groq, Mistral AI, xAI (Grok), OpenRouter, Qwen (DashScope), Kimi (Moonshot), Together AI, Fireworks AI, SambaNova, Cohere, Doubao (ByteDance)
Bug Fixes:
Fixed streaming on macOS - Resolved intermittent hang when streaming responses (TCP warmup before fork)
Fixed Anthropic extended thinking - Response parsing now correctly handles thinking models
Fixed Gemini thinking models - Proper MAX_TOKENS detection for thinking responses
Updated xAI model names - Corrected to match current API
Developer:
Added standalone test framework for provider testing (lua tests/run_tests.lua)
Unit tests for system prompts, streaming parser, and response parsing
Full Changelog: v0.6.1...v0.6.2