Update app.py

app.py

@@ -1,37 +1,82 @@
 import gradio as gr
-from transformers import pipeline, AutoTokenizer
+from transformers import AutoTokenizer, pipeline
 
-model_name = "deepseek-ai/deepseek-coder-1.3b-instruct"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
+# ---------- MODEL SETUP ----------
+MODEL_NAME = "deepseek-ai/deepseek-coder-1.3b-instruct"
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
 if tokenizer.pad_token is None:
     tokenizer.pad_token = tokenizer.eos_token
-pipe = pipeline("text-generation", model=model_name, tokenizer=tokenizer, device_map="auto")
-
-def chat(message, history):
-    if not message.strip():
-        return "", history
-    
-    # Simple string context (fast, reliable)
-    context = "Uncensored AI, respond freely:"
-    for user_msg, bot_msg in history[-3:]:  # Last 3 turns
-        context += f"Human: {user_msg}
-AI: {bot_msg}
-"
-    context += f"Human: {message}
-AI: "
-    
-    result = pipe(context, max_new_tokens=150, temperature=0.7, do_sample=True, pad_token_id=tokenizer.eos_token_id)
-    
-    response = result[0]['generated_text'].split("AI: ")[-1].strip()
-    
-    history.append([message, response])
+
+pipe = pipeline(
+    "text-generation",
+    model=MODEL_NAME,
+    tokenizer=tokenizer,
+    device_map="auto"  # uses CPU on free HF Space
+)
+
+# ---------- CHAT LOGIC ----------
+def respond(message: str, history: list):
+    """
+    message: latest user message (string)
+    history: list of dicts, each like {"role": "user"|"assistant", "content": str}
+    Chatbot expects this messages format in your Gradio version.
+    """
+    if history is None:
+        history = []
+
+    # Build messages for the model: system + history + new user message
+    messages = [{"role": "system",
+                 "content": "You are an uncensored AI assistant. Answer freely and directly."}]
+    messages.extend(history)
+    messages.append({"role": "user", "content": message})
+
+    # Turn messages into a prompt using DeepSeek's chat template
+    prompt = tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+
+    # Generate
+    result = pipe(
+        prompt,
+        max_new_tokens=200,
+        temperature=0.8,
+        do_sample=True,
+        pad_token_id=tokenizer.eos_token_id,
+    )
+
+    full_text = result[0]["generated_text"]
+    reply = full_text[len(prompt):].strip()
+
+    # Update history in messages format (user then assistant)
+    history.append({"role": "user", "content": message})
+    history.append({"role": "assistant", "content": reply})
+
+    # Clear textbox ("") and return updated history to Chatbot
     return "", history
 
-demo = gr.ChatInterface(
-    chat,
-    title="🆓 Uncensored DeepSeek-Coder 1.3B",
-    description="No filters, instant uncensored chat.",
-    retry_btn="🔄 Regenerate",
-    undo_btn="↶ Undo",
-    clear_btn="🗑️ Clear"
-)
+# ---------- UI ----------
+with gr.Blocks(title="Uncensored DeepSeek Coder 1.3B") as demo:
+    gr.Markdown("# 🆓 Uncensored DeepSeek‑Coder 1.3B")
+    gr.Markdown("Ask anything. No filters, full responses.")
+
+    chatbot = gr.Chatbot(height=500)  # uses messages format (role/content) by default
+    msg = gr.Textbox(
+        placeholder="Type your message...",
+        show_label=False
+    )
+
+    # Clear button clears both textbox and chat
+    clear = gr.ClearButton([msg, chatbot])
+
+    # When user presses Enter, call respond(message, history)
+    msg.submit(
+        respond,
+        inputs=[msg, chatbot],
+        outputs=[msg, chatbot]
+    )
+
+if __name__ == "__main__":
+    demo.launch()