fix(inference): validate base_url, add trailing slash, clear OPENAI_* env conflicts, upgrade openai version

made some final changes

chore: update inference dependencies and refactor model loading logic

first commit