Update entrypoint.sh

Update Dockerfile

Update entrypoint.sh

Update README.md

Update chat.html

Update Dockerfile

Add markdown rendering (marked.js + highlight.js)

Switch model to IQ4_XS

Update MODEL_PATH to IQ4_XS

Render streamed reasoning_content in a collapsible Thinking box

Enable reasoning by default via --jinja --reasoning-format deepseek

Serve chat UI at / and keep /v1 API + add /api-info

Copy chat.html into image

Add browser chat UI

Set max generation to 25000 tokens (ctx 131072, leaves 106072 for prompt)

Use all CPU threads for faster inference

Remove diagnostic code; proxy confirmed working

Add diagnostic error surfacing to client

Use explicit chatml template instead of broken embedded jinja template

Fix Content-Length error: strip framing headers from request and response

Use wget instead of deprecated huggingface-cli

Fix pip install with --break-system-packages for Ubuntu 24.04

Use local pre-downloaded model and start uvicorn proxy

Add robust proxy with lifespan-based llama-server readiness check

Pre-download model, add robust proxy

Update README with API docs and app_port

Add Dockerfile for Qwen3.6-27B IQ2_M

initial commit