Release v2.0.3 — Image Studio Connection Testing & Research Mode Token Optimization · mdaiWorks/yengi

What's New in v2.0.3 🚀

Live Connection Test: Added a dedicated "Test Connection" button in Workspace Settings for both OpenAI-compatible endpoints (e.g., rapid-mlx, Ollama, vLLM, OpenAI) and local ComfyUI servers.
Latency & Model Discovery: Real-time round-trip latency reporting (ms) and automatic discovery of available models on the host.
Diagnostic Error Handling: Clear, contextual feedback for common connectivity pitfalls — including HTTP 401 unauthorized errors, firewall timeouts, and specific guidance for Mac rapid-mlx servers requiring --trusted-hosts "*".

Zero-Tool Casual Routing: Conversational greetings and casual messages (e.g., "hello", "how are you", "thanks") now completely bypass tool loading in Research Mode, regardless of whether the Router is enabled or disabled.
Massive Token Savings: Reduces conversational overhead from ~12k tokens to ~300 tokens per message for everyday casual banter while keeping full research capabilities on demand.

ControlNet Pose Routing: Ensured pose guidance triggers only when ControlNet is explicitly toggled on.
Local Model Timeout: Extended local generation timeout to 30 minutes to comfortably accommodate high-step local diffusion workflows.
Packaging & Setup Sync: Full synchronization with Inno Setup installer scripts and automatic updater checksums.