This is an interactive chatbot demo powered by the mistralai/Mistral-7B-Instruct-v0.3 model, using Hugging Face Transformers and Gradio. It features token streaming, formatted responses, and runs efficiently on GPU (T4 or A10G recommended).

• Streaming output (token-by-token generation)
• Markdown rendering for formatting and newlines
• Fast inference with GPU acceleration
• Runs entirely on Hugging Face Spaces

This app requires the following environment setup in requirements.txt:

gradio==5.31.0                  # UI framework with streaming support
transformers==4.42.2            # Required for Mistral v0.3 + chat_template + streaming
torch==2.2.2                    # GPU-compatible for inference
accelerate                      # Optional, enables device_map="auto"
numpy<2                         # Prevents known PyTorch compatibility issues
sentencepiece                   # Required for tokenizer loading (SentencePiece-based)