Upload app.py with huggingface_hub

app.py

@@ -303,6 +303,14 @@ EXAMPLES_QA = [
 # ---------------------------------------------------------------------------
 # Inference functions
 # ---------------------------------------------------------------------------
+ZEROGPU_QUEUE_HINT = "No GPU was available after"
+
+
+def _is_zerogpu_queue_timeout(err: Exception) -> bool:
+    """Detect ZeroGPU queue timeout errors from the spaces runtime."""
+    return ZEROGPU_QUEUE_HINT in str(err)
+
+
 
 def _use_hf_model() -> bool:
     """Check if we should use the fine-tuned HF model."""
@@ -320,8 +328,8 @@ def classify(description: str, api_key: str) -> str:
 
     use_hf = _use_hf_model()
 
-    # If user provided an API key, prefer OpenAI (explicit choice)
-    if api_key.strip():
+    # Local-only override: allow OpenAI fallback when running outside Spaces.
+    if api_key.strip() and not IS_SPACES:
         use_hf = False
 
     if use_hf:
@@ -330,7 +338,14 @@ def classify(description: str, api_key: str) -> str:
             return json.dumps(result, indent=2)
         except Exception as e:
             logger.error(f"HF model error: {e}")
-            # Fall through to OpenAI if available
+            if _is_zerogpu_queue_timeout(e):
+                return json.dumps(
+                    {"error": "ZeroGPU queue is full right now. Try again in 1-2 minutes."},
+                    indent=2,
+                )
+            if IS_SPACES:
+                return json.dumps({"error": f"Model inference failed: {str(e)}"}, indent=2)
+            # Local fallback path only.
             key = os.getenv("OPENAI_API_KEY", "")
             if not key:
                 return json.dumps({"error": f"Model inference failed: {str(e)}"}, indent=2)
@@ -356,7 +371,7 @@ def ask(question: str, api_key: str) -> str:
 
     use_hf = _use_hf_model()
 
-    if api_key.strip():
+    if api_key.strip() and not IS_SPACES:
         use_hf = False
 
     if use_hf:
@@ -364,6 +379,11 @@ def ask(question: str, api_key: str) -> str:
             return _ask_gpu(question)
         except Exception as e:
             logger.error(f"HF model error: {e}")
+            if _is_zerogpu_queue_timeout(e):
+                return "**Error:** ZeroGPU queue is full right now. Try again in 1-2 minutes."
+            if IS_SPACES:
+                return f"**Error:** Model inference failed: {str(e)}"
+            # Local fallback path only.
             key = os.getenv("OPENAI_API_KEY", "")
             if not key:
                 return f"**Error:** Model inference failed: {str(e)}"
@@ -542,14 +562,17 @@ def build_app() -> gr.Blocks:
             </div>
             """)
 
-        # --- API Key (optional on Spaces) ---
-        with gr.Group():
-            api_key = gr.Textbox(
-                label="OpenAI API Key (Optional)" if IS_SPACES else "OpenAI API Key",
-                placeholder="sk-... (optional — the fine-tuned model runs for free)" if IS_SPACES else "sk-... (required for classification)",
-                type="password",
-                info="Your key is never stored. If provided, GPT-4o will be used instead of the fine-tuned model.",
-            )
+        # --- API Key (local mode only) ---
+        if IS_SPACES:
+            api_key = gr.State("")
+        else:
+            with gr.Group():
+                api_key = gr.Textbox(
+                    label="OpenAI API Key",
+                    placeholder="sk-... (required for OpenAI fallback)",
+                    type="password",
+                    info="Your key is never stored.",
+                )
 
         # --- Main Tabs ---
         with gr.Tabs() as tabs: