Update app.py

app.py

@@ -1,11 +1,10 @@
 import gradio as gr
 import spaces
 from huggingface_hub import InferenceClient
-from huggingface_hub.errors import HfHubHTTPError, InferenceTimeoutError
 from transformers import pipeline
 
 LOCAL_MODEL = "Qwen/Qwen3-0.6B"
-REMOTE_MODEL = "openai/nopenope"
+REMOTE_MODEL = "openai/gpt-oss-20b"
 
 pipe = pipeline(
     "text-generation",
@@ -16,22 +15,18 @@ pipe = pipeline(
 
 fancy_css = """
 .gradio-container {
-    width: 95% !important;
-    max-width: 1400px !important;
-    margin: 0 auto;
+    width: 96% !important;
+    max-width: none !important;
 }
-
 #app-title {
     text-align: center;
     margin-bottom: 4px;
 }
-
 #app-subtitle {
     text-align: center;
     color: var(--body-text-color-subdued);
     margin-bottom: 24px;
 }
-
 #chat-container {
     width: 100%;
     border: 1px solid var(--border-color-primary);
@@ -39,18 +34,15 @@ fancy_css = """
     padding: 16px;
     box-shadow: 0 2px 8px rgba(0, 0, 0, 0.06);
 }
-
 #model-note {
     font-size: 0.9em;
     color: var(--body-text-color-subdued);
     margin-top: 8px;
 }
-
 @media (max-width: 768px) {
     .gradio-container {
         width: 98% !important;
     }
-
     #chat-container {
         padding: 8px;
     }
@@ -76,35 +68,6 @@ def local_generate(
     return outputs[0]["generated_text"][-1]["content"]
 
 
-def remote_generate(
-    messages,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token,
-):
-    client = InferenceClient(
-        token=hf_token.token,
-        model=REMOTE_MODEL,
-    )
-
-    response = ""
-
-    for chunk in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = chunk.choices
-
-        if choices and choices[0].delta.content:
-            response += choices[0].delta.content
-
-        yield response
-
-
 def respond(
     message,
     history: list[dict[str, str]],
@@ -112,14 +75,14 @@ def respond(
     max_tokens,
     temperature,
     top_p,
-    model_mode,
+    use_local_model,
     hf_token: gr.OAuthToken,
 ):
     messages = [{"role": "system", "content": system_message}]
     messages.extend(history)
     messages.append({"role": "user", "content": message})
 
-    if model_mode == "Local Model":
+    if use_local_model:
         print("[MODE] local")
 
         response = local_generate(
@@ -129,76 +92,37 @@ def respond(
             top_p,
         )
 
-        yield f"**Backend: Local Model**\n\n{response}"
+        yield response
         return
 
-    if hf_token is None or not getattr(hf_token, "token", None):
-        if model_mode == "Automatic Failover":
-            print("[FAILOVER] No Hugging Face token. Using local model.")
-
-            response = local_generate(
-                messages,
-                max_tokens,
-                temperature,
-                top_p,
-            )
-
-            yield (
-                "**Remote model unavailable. "
-                "Automatically switched to Local Model.**\n\n"
-                + response
-            )
-            return
+    print("[MODE] api")
 
+    if hf_token is None or not getattr(hf_token, "token", None):
         yield "⚠️ Please log in with your Hugging Face account first."
         return
 
-    if model_mode == "Remote API":
-        print("[MODE] api")
-
-        try:
-            for response in remote_generate(
-                messages,
-                max_tokens,
-                temperature,
-                top_p,
-                hf_token,
-            ):
-                yield f"**Backend: Remote API**\n\n{response}"
-
-        except (InferenceTimeoutError, HfHubHTTPError) as error:
-            yield f"⚠️ Remote API error: {error}"
-
-        return
-
-    print("[MODE] automatic")
+    client = InferenceClient(
+        token=hf_token.token,
+        model=REMOTE_MODEL,
+    )
 
-    try:
-        for response in remote_generate(
-            messages,
-            max_tokens,
-            temperature,
-            top_p,
-            hf_token,
-        ):
-            yield f"**Backend: Remote API**\n\n{response}"
+    response = ""
 
-    except (InferenceTimeoutError, HfHubHTTPError) as error:
-        print(f"[FAILOVER] Remote model failed: {error}")
-        print("[FAILOVER] Switching to local model.")
+    for chunk in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        choices = chunk.choices
+        token = ""
 
-        response = local_generate(
-            messages,
-            max_tokens,
-            temperature,
-            top_p,
-        )
+        if len(choices) and choices[0].delta.content:
+            token = choices[0].delta.content
 
-        yield (
-            "**Remote model failed. "
-            "Automatically switched to Local Model.**\n\n"
-            + response
-        )
+        response += token
+        yield response
 
 
 chatbot = gr.ChatInterface(
@@ -229,14 +153,9 @@ chatbot = gr.ChatInterface(
             step=0.05,
             label="Top-p (nucleus sampling)",
         ),
-        gr.Radio(
-            choices=[
-                "Remote API",
-                "Local Model",
-                "Automatic Failover",
-            ],
-            value="Remote API",
-            label="Model mode",
+        gr.Checkbox(
+            label="Use Local Model",
+            value=False,
         ),
     ],
 )
@@ -252,7 +171,7 @@ with gr.Blocks(css=fancy_css) as demo:
     )
 
     gr.Markdown(
-        "Compare a remotely hosted LLM with a model running locally on the Space.",
+        "A fancier version of the standard Huggging Face chatbot template.",
         elem_id="app-subtitle",
     )
 
@@ -260,8 +179,7 @@ with gr.Blocks(css=fancy_css) as demo:
         chatbot.render()
 
         gr.Markdown(
-            "Use **Additional inputs** to choose between the remote API, "
-            "the locally executed model, or automatic failover.",
+            "Use **Additional inputs** to switch between the API model and the locally executed model.",
             elem_id="model-note",
         )