Add return type hint to generate endpoint

Gradio's API engine was dropping the streamed JSON yields because the
function declared no return type (warnings: 'returned too many output
values (needed: 0, returned: 1). Ignoring extra values.'). Adding
'-> str' so the yielded JSON strings are forwarded to the SSE stream
and reach the JS client.

Co-Authored-By: GLM-5.2 <noreply@z.ai>

app.py

@@ -162,7 +162,7 @@ def generate(
     history: list = None,
     max_new_tokens: int = DEFAULT_MAX_NEW_TOKENS,
     temperature: float = DEFAULT_TEMPERATURE,
-):
+) -> str:
     # Stream an Ornith-1.0-9B completion; yields JSON strings with
     # {reasoning, answer, status, error}.
     if history is None: