repurpose: HelpingAI/HELVETE-3B (nsfw-3b-q4_k_m.gguf)

repurpose: mradermacher/Gemma-3-Prompt-Coder-270m-it-Uncensored-GGUF (Gemma-3-Prompt-Coder-270m-it-Uncensored.Q4_K_S.gguf)

repurpose: HelpingAI/HELVETE-3B (nsfw-3b-q4_k_m.gguf)

Update README.md

Update requirements.txt

Update requirements.txt

Update README.md

Update requirements.txt

fix: bypass from_pretrained path bug with hf_hub_download

fix: separate draft_model loading to avoid path construction bug

fix: correct model filenames for SmolLM3 Q4_K_S and SmolLM2 draft model

feat: add production refinements (Phase 1-3)

SmolLM3-3B-Instruct streaming

Fix: Gradio 6.0 compatible ChatInterface with streaming

Fix: Gradio 6.0 compatibility - remove deprecated theme param and show_copy_button

Major fix: Switch to transformers (no llama-cpp build). Use Qwen2.5-3B for fast CPU inference

Fix: Switch to standard llama-cpp-python package (remove editable install)

Fix: Use editable git install for llama-cpp-python, remove version pins

Fix: Use prebuilt llama-cpp-python wheels, pin dependencies, extend timeout

Initial deployment: General Reasoning Agent

initial commit