Update app.py

Update to TinyLlama model

Switch to TinyLlama for reliable inference

Switch to Qwen2.5-Coder-0.5B with simplified prompts for reliability

Update to Phi-3 model with proper prompt formatting

FIXED: Direct HF API calls using Llama 3.2 3B - fully working now

Updated huggingface-hub version + Added simple test examples

Fixed: Use text_generation instead of chat_completion + Added Anthropic Claude API support

Fixed HF Inference API - using Qwen 2.5 1.5B with chat_completion

Updated to HF Inference API with DeepSeek-R1 - LIFETIME FREE with unlimited usage

Updated to use AJ-mini model based on DeepSeek-R1 1.5B