Update app.py

app.py

@@ -1,60 +1,58 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-# Pick three open-source models (note: some may require login/accept T&Cs)
+# Three open-source models to choose from
 MODEL_CHOICES = {
     "Mistral 7B Instruct (mistralai/Mistral-7B-Instruct-v0.2)": "mistralai/Mistral-7B-Instruct-v0.2",
     "Falcon 7B Instruct (tiiuae/falcon-7b-instruct)": "tiiuae/falcon-7b-instruct",
     "Zephyr 7B Beta (HuggingFaceH4/zephyr-7b-beta)": "HuggingFaceH4/zephyr-7b-beta",
 }
 
+
 def complete_text(
     prompt: str,
     max_tokens: int,
     temperature: float,
     top_p: float,
     model_choice: str,
-    hf_token: gr.OAuthToken,   # <-- token comes from LoginButton
+    hf_token: gr.OAuthToken,  # token provided after user logs in
 ):
     """
-    Stream a plain text completion from a HF-hosted open model.
+    Generate a text completion from a Hugging Face model, streamed chunk by chunk.
     """
     if not prompt:
-        yield "⚠️ Enter a prompt."
+        yield "⚠️ Please enter a prompt."
         return
 
-    model_id = MODEL_CHOICES[model_choice]
-
-    # Some models require auth; if the user isn't logged in, warn them.
-    token_str = getattr(hf_token, "token", None)
-    if token_str is None:
-        yield "🔐 This Space uses the Hugging Face Inference API. Please click **Login** (left sidebar) to authorize."
+    # If the user has not logged in, warn them (needed for gated models like Mistral / LLaMA)
+    if not getattr(hf_token, "token", None):
+        yield "🔐 Please login with your Hugging Face account (see left sidebar)."
         return
 
-    client = InferenceClient(model=model_id, token=token_str)
+    model_id = MODEL_CHOICES[model_choice]
+    client = InferenceClient(model=model_id, token=hf_token.token)
 
+    generated = ""
     try:
-        generated = ""
-        # text_generation returns an iterator of string chunks when stream=True
         for chunk in client.text_generation(
             prompt=prompt,
             max_new_tokens=int(max_tokens),
             temperature=float(temperature),
             top_p=float(top_p),
-            stream=True,
             repetition_penalty=1.0,
+            stream=True,
         ):
             generated += chunk
             yield generated
     except Exception as e:
-        yield f"❌ Inference error: {e}"
+        yield f"❌ Error while generating: {e}"
 
 
 with gr.Blocks() as demo:
-    gr.Markdown("## ✍️ Text Completion Demo (Hugging Face Inference API)")
+    gr.Markdown("## ✍️ Text Completion Demo (Open-Source LLMs via Hugging Face Inference API)")
     gr.Markdown(
-        "Pick a model, enter a prompt, and see a streamed completion. "
-        "Some models require you to **Login** (left sidebar) to access gated weights."
+        "Pick a model, enter a prompt, and stream completions. "
+        "Some models require logging in (click **Login** in the sidebar)."
     )
 
     with gr.Row():
@@ -72,13 +70,13 @@ with gr.Blocks() as demo:
                 value=list(MODEL_CHOICES.keys())[0],
                 label="Choose a model",
             )
-            submit = gr.Button("Generate")
+            submit = gr.Button("Generate Completion")
         with gr.Column(scale=3):
             with gr.Sidebar():
-                login = gr.LoginButton()  # <-- THIS provides the OAuth token
+                login = gr.LoginButton()  # supplies OAuth token at runtime
             output = gr.Textbox(label="Generated Completion", lines=15)
 
-    # Wire the login button as the input for the token parameter
+    # Wire inputs (6 args) to match function params (6 args)
     submit.click(
         fn=complete_text,
         inputs=[prompt, max_tokens, temperature, top_p, model_choice, login],