Update default Groq model to openai/gpt-oss-120b and fix Dockerfile reranker model

Bake AI models directly into Docker image to prevent timeouts

Fix requirements for FastAPI backend

Initial commit