fix: owned_by field in GET /v1/models returns list instead of string

[nileshpatil6]

What does this fix?

In ModelManager.get_gen_models(), the owned_by field in the model list response was incorrectly set to a list instead of a string.

# Before (bug): returns a list like ["mistralai", "Mistral-7B-v0.1"]
author = repo.repo_id.split("/") if "/" in repo.repo_id else ""

# After (fix): returns just the author string "mistralai"
author = repo.repo_id.split("/")[0] if "/" in repo.repo_id else ""

str.split("/") returns a list of all segments. Without [0], the owned_by field in the GET /v1/models response contained a Python list (e.g. ["mistralai", "Mistral-7B-v0.1"]) instead of the expected author string ("mistralai"). This breaks OpenAI-compatible clients that parse the models list and expect owned_by to be a string.

Changes

• src/transformers/cli/serving/model_manager.py: add [0] index to split("/") call so only the author segment is used for owned_by.