adding examples

app.py

@@ -26,8 +26,18 @@ def respond(
     custom_model
 ):
     """
-    Respond function for ChatInterface.
+    This function handles the chatbot response. It takes in:
+    - message: the user's new message
+    - history: the list of previous messages, each as a tuple (user_msg, assistant_msg)
+    - system_message: the system prompt
+    - max_tokens: the maximum number of tokens to generate in the response
+    - temperature: sampling temperature
+    - top_p: top-p (nucleus) sampling
+    - frequency_penalty: penalize repeated tokens in the output
+    - seed: a fixed seed for reproducibility; -1 will mean 'random'
+    - custom_model: the final model name in use, which may be set by selecting from the Featured Models radio or by typing a custom model
     """
+
     print(f"Received message: {message}")
     print(f"History: {history}")
     print(f"System message: {system_message}")
@@ -35,25 +45,38 @@ def respond(
     print(f"Frequency Penalty: {frequency_penalty}, Seed: {seed}")
     print(f"Selected model (custom_model): {custom_model}")
 
+    # Convert seed to None if -1 (meaning random)
     if seed == -1:
         seed = None
 
-    # Construct the messages array
+    # Construct the messages array required by the API
     messages = [{"role": "system", "content": system_message}]
+    print("Initial messages array constructed.")
+
+    # Add conversation history to the context
     for val in history:
-        user_part = val[0]
-        assistant_part = val[1]
+        user_part = val[0]  # Extract user message from the tuple
+        assistant_part = val[1]  # Extract assistant message from the tuple
         if user_part:
             messages.append({"role": "user", "content": user_part})
+            print(f"Added user message to context: {user_part}")
         if assistant_part:
             messages.append({"role": "assistant", "content": assistant_part})
+            print(f"Added assistant message to context: {assistant_part}")
 
+    # Append the latest user message
     messages.append({"role": "user", "content": message})
+    print("Latest user message appended.")
 
-    # If user provided a model, use it; else use default
+    # If user provided a model, use that; otherwise, fall back to a default model
     model_to_use = custom_model.strip() if custom_model.strip() != "" else "meta-llama/Llama-3.3-70B-Instruct"
+    print(f"Model selected for inference: {model_to_use}")
+
+    # Start with an empty string to build the response as tokens stream in
     response = ""
+    print("Sending request to OpenAI API.")
 
+    # Make the streaming request to the HF Inference API via openai-like client
     for message_chunk in client.chat.completions.create(
         model=model_to_use,
         max_tokens=max_tokens,
@@ -64,47 +87,81 @@ def respond(
         seed=seed,
         messages=messages,
     ):
+        # Extract the token text from the response chunk
         token_text = message_chunk.choices[0].delta.content
+        print(f"Received token: {token_text}")
         response += token_text
         yield response
 
+    print("Completed response generation.")
+
 
 # -------------------------
 # GRADIO UI CONFIGURATION
 # -------------------------
 
-# Create a Chatbot component
-chatbot = gr.Chatbot(
-    height=600, 
-    show_copy_button=True,
-    placeholder="Select a model and begin chatting",
-    likeable=True,
-    layout="panel"
-)
+# Create a Chatbot component with a specified height
+chatbot = gr.Chatbot(height=600, show_copy_button=True, placeholder="Select a model and begin chatting", likeable=True, layout="panel")
+print("Chatbot interface created.")
 
-# Create textboxes/sliders for system prompt, tokens, etc.
+# Create textboxes and sliders for system prompt, tokens, and other parameters
 system_message_box = gr.Textbox(value="", label="System message")
-max_tokens_slider = gr.Slider(1, 4096, value=512, step=1, label="Max new tokens")
-temperature_slider = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
-top_p_slider = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-P")
-frequency_penalty_slider = gr.Slider(-2.0, 2.0, value=0.0, step=0.1, label="Frequency Penalty")
-seed_slider = gr.Slider(-1, 65535, value=-1, step=1, label="Seed (-1 for random)")
-custom_model_box = gr.Textbox(value="", label="Custom Model",
-                              info="(Optional) Provide a custom Hugging Face model path. Overrides any selected featured model.")
+
+max_tokens_slider = gr.Slider(
+    minimum=1,
+    maximum=4096,
+    value=512,
+    step=1,
+    label="Max new tokens"
+)
+temperature_slider = gr.Slider(
+    minimum=0.1,
+    maximum=4.0,
+    value=0.7,
+    step=0.1,
+    label="Temperature"
+)
+top_p_slider = gr.Slider(
+    minimum=0.1,
+    maximum=1.0,
+    value=0.95,
+    step=0.05,
+    label="Top-P"
+)
+frequency_penalty_slider = gr.Slider(
+    minimum=-2.0,
+    maximum=2.0,
+    value=0.0,
+    step=0.1,
+    label="Frequency Penalty"
+)
+seed_slider = gr.Slider(
+    minimum=-1,
+    maximum=65535,
+    value=-1,
+    step=1,
+    label="Seed (-1 for random)"
+)
+
+# The custom_model_box is what the respond function sees as "custom_model"
+custom_model_box = gr.Textbox(
+    value="",
+    label="Custom Model",
+    info="(Optional) Provide a custom Hugging Face model path. Overrides any selected featured model."
+)
 
 def set_custom_model_from_radio(selected):
     """
-    Update the Custom Model textbox when a featured model is selected.
+    This function will get triggered whenever someone picks a model from the 'Featured Models' radio.
+    We will update the Custom Model text box with that selection automatically.
     """
     print(f"Featured model selected: {selected}")
     return selected
 
-
-# Create a user textbox that we can reference
-# This will become our "Message" input inside the ChatInterface
-user_textbox = gr.MultimodalTextbox()
-
-# No 'examples' here—because we want to keep the user's parameters unchanged
+# IMPORTANT: Because we have 1 main user input + 7 additional inputs,
+# each example should be an 8-item list [user_text, system_prompt, max_tokens, temperature,
+# top_p, frequency_penalty, seed, custom_model].
+# You can adjust the default parameter values if desired.
 demo = gr.ChatInterface(
     fn=respond,
     additional_inputs=[
@@ -114,27 +171,42 @@ demo = gr.ChatInterface(
         top_p_slider,
         frequency_penalty_slider,
         seed_slider,
-        custom_model_box
+        custom_model_box,
     ],
     fill_height=True,
     chatbot=chatbot,
-    textbox=user_textbox,
-    multimodal=True,
-    concurrency_limit=20,
     theme="Nymbo/Nymbo_Theme",
-    # No examples parameter used
-    cache_examples=False
 )
 print("ChatInterface object created.")
 
+# Add examples to the interface
+demo.add_examples(
+    examples=[
+        ["Howdy, partner!", "You are a friendly assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
+        ["What's your model name and who trained you?", "You are a factual assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
+        ["How many R's are there in 'Strawberry'?", "You are a playful assistant.", 512, 0.7, 0.95, 0.0, -1, ""],
+    ],
+    inputs=[
+        chatbot,
+        system_message_box,
+        max_tokens_slider,
+        temperature_slider,
+        top_p_slider,
+        frequency_penalty_slider,
+        seed_slider,
+        custom_model_box,
+    ],
+)
+print("Examples added to the interface.")
+
 with demo:
-    # Featured models accordion
     with gr.Accordion("Featured Models", open=False):
         model_search_box = gr.Textbox(
             label="Filter Models",
             placeholder="Search for a featured model...",
             lines=1
         )
+        print("Model search box created.")
 
         models_list = [
             "meta-llama/Llama-3.3-70B-Instruct",
@@ -155,6 +227,7 @@ with demo:
             "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
             "microsoft/Phi-3.5-mini-instruct",
         ]
+        print("Models list initialized.")
 
         featured_model_radio = gr.Radio(
             label="Select a model below",
@@ -162,9 +235,12 @@ with demo:
             value="meta-llama/Llama-3.3-70B-Instruct",
             interactive=True
         )
+        print("Featured models radio button created.")
 
         def filter_models(search_term):
+            print(f"Filtering models with search term: {search_term}")
             filtered = [m for m in models_list if search_term.lower() in m.lower()]
+            print(f"Filtered models: {filtered}")
             return gr.update(choices=filtered)
 
         model_search_box.change(
@@ -172,34 +248,14 @@ with demo:
             inputs=model_search_box,
             outputs=featured_model_radio
         )
+        print("Model search box change event linked.")
 
         featured_model_radio.change(
             fn=set_custom_model_from_radio,
             inputs=featured_model_radio,
             outputs=custom_model_box
         )
-
-    # Example Prompts accordion
-    with gr.Accordion("Example Prompts", open=False):
-        ex1_btn = gr.Button("Example 1: 'Howdy, partner!'")
-        ex2_btn = gr.Button("Example 2: 'What's your model name and who trained you?'")
-        ex3_btn = gr.Button("Example 3: 'How many R's in Strawberry?'")
-
-        # Helper function that returns an update for user_textbox
-        def load_example(example_text):
-            return gr.update(value=example_text)
-
-        ex1_btn.click(fn=lambda: load_example("Howdy, partner!"), 
-                      inputs=[],
-                      outputs=user_textbox)
-
-        ex2_btn.click(fn=lambda: load_example("What's your model name and who trained you?"), 
-                      inputs=[],
-                      outputs=user_textbox)
-
-        ex3_btn.click(fn=lambda: load_example("How many R's are there in the word Strawberry?"), 
-                      inputs=[],
-                      outputs=user_textbox)
+        print("Featured model radio button change event linked.")
 
 print("Gradio interface initialized.")