Thulge-Chat drops self-contained offline LLM runner

While power users gravitate toward LM Studio or Ollama, Thulge-Chat targets a very specific and underserved niche: zero-friction local LLM deployments for restricted corporate environments.

• –Bundles the llama-server binary with simple batch/shell scripts to completely bypass Docker or system-level installations
• –Binds strictly to localhost, guaranteeing zero network traffic or telemetry after the initial model download
• –Defaults to a heavily quantized Qwen2.5-1.5B but supports swapping in any GGUF model, making it a flexible drop-in solution for low-end hardware