fix: rewrite inference.py to use LiteLLM proxy via API_BASE_URL/API_KEY env vars

fix: add uv.lock, server entry point and openenv-core dependency for openenv validate