Fix composer hangs and JSON failures (no timeout on the LLM call)

app.py

@@ -501,15 +501,27 @@ def _llm_json(system, user):
 
     from openai import OpenAI
 
-    client = OpenAI(base_url="https://router.huggingface.co/v1", api_key=os.environ["HF_TOKEN"])
-    completion = client.chat.completions.create(
-        model="MiniMaxAI/MiniMax-M3:fastest",
-        messages=[{"role": "system", "content": system}, {"role": "user", "content": user}],
-    )
-    text = completion.choices[0].message.content.strip()
-    if text.startswith("```"):
-        text = text.split("```")[1].removeprefix("json").strip()
-    return _json.loads(text)
+    # Bounded timeout: a hung provider must fail over, not freeze the UI at the composing step.
+    client = OpenAI(base_url="https://router.huggingface.co/v1", api_key=os.environ["HF_TOKEN"], timeout=60, max_retries=0)
+    last_error = None
+    for model in ("MiniMaxAI/MiniMax-M3:fastest", "MiniMaxAI/MiniMax-M3", "MiniMaxAI/MiniMax-M3:fireworks-ai"):
+        try:
+            completion = client.chat.completions.create(
+                model=model,
+                messages=[{"role": "system", "content": system}, {"role": "user", "content": user}],
+            )
+            text = completion.choices[0].message.content or ""
+            # Tolerate fences/preambles and reject truncated replies: parse the outermost {...} span.
+            start, end = text.find("{"), text.rfind("}")
+            if start == -1 or end <= start:
+                raise ValueError(f"no JSON object in composer reply (finish_reason={completion.choices[0].finish_reason})")
+            return _json.loads(text[start : end + 1])
+        except Exception as e:
+            last_error = e
+    raise gr.Error(
+        "The MiniMax-M3 composer is overloaded right now — try again in a moment, "
+        "or write the lyrics and structured prompt directly in the Studio tab."
+    ) from last_error
 
 
 def compose_song(description, duration):