Reverted to working build of chat interface.

app.py

@@ -3,6 +3,7 @@ import torch
 import importlib.util
 from tokenizers import Tokenizer
 from huggingface_hub import hf_hub_download
+import os
 
 # Download and import model components from HF Hub
 model_repo = "TimurHromek/HROM-V1"
@@ -35,53 +36,39 @@ model = load_model()
 safety = SafetyManager(model, tokenizer)
 max_response_length = 200
 
-def generate_response(model, tokenizer, input_ids, safety_manager, max_length=200, temperature=1.0):
+def generate_response(model, tokenizer, input_ids, safety_manager, max_length=200):
     device = next(model.parameters()).device
     generated_ids = input_ids.copy()
     for _ in range(max_length):
         input_tensor = torch.tensor([generated_ids], device=device)
         with torch.no_grad():
             logits = model(input_tensor)
-        
-        # Get last token logits and apply temperature
-        next_token_logits = logits[0, -1, :]
-        if temperature != 1.0:
-            next_token_logits = next_token_logits / temperature
-        probs = torch.softmax(next_token_logits, dim=-1)
-        
-        # Sample next token
-        next_token = torch.multinomial(probs, num_samples=1).item()
-        
-        # Stop if end token is generated
+        next_token = logits.argmax(-1)[:, -1].item()
         if next_token == tokenizer.token_to_id("</s>"):
             break
-            
-        # Safety check
         current_text = tokenizer.decode(generated_ids + [next_token])
         if not safety_manager.content_filter(current_text):
             break
-            
         generated_ids.append(next_token)
     return generated_ids[len(input_ids):]
 
-def process_message(user_input, chat_history, token_history, temperature, max_context_length):
+def process_message(user_input, chat_history, token_history):
     # Process user input
     user_turn = f"<user> {user_input} </s>"
     user_tokens = tokenizer.encode(user_turn).ids
     token_history.extend(user_tokens)
     
-    # Prepare input sequence with context limit
+    # Prepare input sequence
     input_sequence = [tokenizer.token_to_id("<s>")] + token_history
     
-    # Truncate based on max context length
-    max_input_len = max_context_length
+    # Truncate if needed
+    max_input_len = CONFIG["max_seq_len"] - max_response_length
     if len(input_sequence) > max_input_len:
         input_sequence = input_sequence[-max_input_len:]
         token_history = input_sequence[1:]
     
-    # Generate response with temperature
-    response_ids = generate_response(model, tokenizer, input_sequence, safety, 
-                                    max_response_length, temperature)
+    # Generate response
+    response_ids = generate_response(model, tokenizer, input_sequence, safety, max_response_length)
     
     # Process assistant response
     assistant_text = "I couldn't generate a proper response."
@@ -104,125 +91,22 @@ def process_message(user_input, chat_history, token_history, temperature, max_co
 def clear_history():
     return [], []
 
-css = """
-:root {
-    --background: white;
-    --text: black;
-    --border: #e0e0e0;
-    --button-bg: #f0f0f0;
-    --button-hover: #e0e0e0;
-    --chatbot-bg: #f8f8f8;
-}
-
-.dark {
-    --background: #1a1a1a;
-    --text: white;
-    --border: #404040;
-    --button-bg: #404040;
-    --button-hover: #505050;
-    --chatbot-bg: #262626;
-}
-
-body {
-    background: var(--background) !important;
-    color: var(--text) !important;
-    transition: all 0.3s ease;
-}
-
-.gr-box {
-    border-color: var(--border) !important;
-    background: var(--background) !important;
-}
-
-.gr-button {
-    background: var(--button-bg) !important;
-    color: var(--text) !important;
-    border-color: var(--border) !important;
-}
-
-.gr-button:hover {
-    background: var(--button-hover) !important;
-}
-
-#chatbot {
-    background: var(--chatbot-bg) !important;
-    border-color: var(--border) !important;
-    min-height: 500px;
-}
-
-.gr-textbox input {
-    color: var(--text) !important;
-}
-
-.dark .gr-markdown {
-    color: var(--text) !important;
-}
-
-.settings-panel {
-    border-left: 1px solid var(--border) !important;
-    padding-left: 20px !important;
-}
-"""
-
-with gr.Blocks(css=css, title="HROM-V1.5 Chatbot") as demo:
-    current_theme = gr.State("light")
-    
-    with gr.Row():
-        with gr.Column(scale=3):
-            gr.Markdown("# HROM-V1.5 Chatbot")
-            chatbot = gr.Chatbot(height=500, elem_id="chatbot")
-            msg = gr.Textbox(label="Your Message", 
-                           placeholder="Type your message...", 
-                           show_label=False,
-                           container=False)
-            
-        with gr.Column(scale=1, min_width=300, elem_classes="settings-panel"):
-            with gr.Accordion("⚙️ Settings", open=False):
-                with gr.Row():
-                    theme_btn = gr.Button("🌙 Dark Theme", variant="secondary")
-                with gr.Row():
-                    temperature = gr.Slider(0.1, 2.0, value=1.0, step=0.1,
-                                         label="Temperature (higher = more creative)")
-                with gr.Row():
-                    max_context = gr.Slider(100, CONFIG["max_seq_len"] - max_response_length,
-                                         value=CONFIG["max_seq_len"] - max_response_length, step=1,
-                                         label="Context Window Size")
-                with gr.Row():
-                    clear_btn = gr.Button("🧹 Clear History", variant="secondary")
-
+with gr.Blocks() as demo:
+    gr.Markdown("# HROM-V1 Chatbot")
+    chatbot = gr.Chatbot(height=500)
+    msg = gr.Textbox(label="Your Message")
     token_state = gr.State([])
-    theme_css = gr.HTML("<style></style>")
-
-    def toggle_theme(theme):
-        new_theme = "dark" if theme == "light" else "light"
-        btn_text = "🌞 Light Theme" if new_theme == "light" else "🌙 Dark Theme"
-        css = """
-        <style>
-            body { background: %s !important; color: %s !important; }
-            .dark-mode { display: %s !important; }
-        </style>
-        """ % (
-            "var(--background)", 
-            "var(--text)",
-            "block" if new_theme == "dark" else "none"
-        )
-        return new_theme, btn_text, css
-
-    theme_btn.click(
-        toggle_theme,
-        current_theme,
-        [current_theme, theme_btn, theme_css]
-    )
     
     msg.submit(
         process_message,
-        [msg, chatbot, token_state, temperature, max_context],
+        [msg, chatbot, token_state],
         [chatbot, token_state],
         queue=False
     ).then(
         lambda: "", None, msg
     )
     
+    clear_btn = gr.Button("Clear Chat History")
     clear_btn.click(
         clear_history,
         outputs=[chatbot, token_state],