Fix expected string or bytes-like object TypeError in format_thinking_tags by ensuring text is returned and type-checked

Silence duckduckgo_search package rename warning in src/tools.py

Add top-tier open, ungated models (DeepSeek R1 Distill, Qwen Coder, Phi-3.5) to configurations

Add Google Gemma 2 (2B, 9B, and 27B) models to available configurations

Set Qwen 2.5 72B Instruct as default for HF Serverless API mode to avoid gated repo issues

Import missing re module in src/engine.py

Fix ValueError: Language plaintext not supported. in gr.Code

Handle gated repository errors gracefully and support hf_token for local loading

Fix UI compilation issues, rename app to Saffan Chat, add AI Persona & Skill presets

Escape LaTeX curly braces in SYSTEM_PROMPT to fix string formatting KeyError

Initialize chatbot state to list and add safety checks for NoneType history in Gradio 5

Set Chatbot type to tuples for Gradio 5 compatibility

Implement thread-free KV-cached inference to prevent Zero-GPU thread hang and quota drain

Deploy premium Claude-style LLM chatbot with web search, scraping, memory, and access control