chore: update Dockerfile, TinyLlama converter, vite config for Space

Sync favored app: TinyLlama/Ollama config, runtime env, UX (tabs, streaming, copy/add, provider dropdown, error actions), Dockerfile and entrypoint

Fix: use /models/{HF_MODEL}/v1/chat/completions path for HF inference proxy

Fix: use chat completions API with SmolLM3-3B for HF inference

Fix: replace TinyLlama with SmolLM3-3B (available on hf-inference)

Harden proxy: safe token sed (#), nginx -t validation, intercept_errors off, safe JSON parse

Fix 502: nginx proxy resolver, SSL, timeouts; clearer 401/502 errors

Fix: use same-origin proxy for HF inference (avoid CORS)

Text Transformer: Docker app for HF Space