fix(vllm): only pass documents via chat_template_kwargs, not top-level

fix(vllm): strip document-role messages before sending to vLLM

fix: conditional Ollama timeout — 5s when vLLM primary, 240s otherwise

fix: vLLM readiness probe + fast Ollama fallback timeout (app/llm.py)

fix(llm): raise vLLM first-token timeout to 360s for RunPod cold start

deploy(l4): self-contained Riprap mirror