Create app.py

app.py

@@ -39,7 +39,6 @@ llm = Llama(
     n_threads=CPU_THREADS,
     n_batch=512,
     use_mmap=True,
-    use_mlock=False,
     verbose=False
 )
 
@@ -47,39 +46,36 @@ print("Model loaded successfully.")
 
 
 # ============================
-# Prompt Formatting
+# Prompt Builder
 # ============================
 
 SYSTEM_PROMPT = """You are DeepSeek Coder, an expert programming assistant.
-You write clean, efficient, production-ready code.
-Only explain if user asks.
+Write clean and efficient code.
+Only explain when asked.
 """
 
 
-def build_prompt(messages):
+def build_prompt(message, history):
 
     prompt = SYSTEM_PROMPT + "\n\n"
 
-    for msg in messages:
-        if msg["role"] == "user":
-            prompt += f"User: {msg['content']}\n"
-        elif msg["role"] == "assistant":
-            prompt += f"Assistant: {msg['content']}\n"
+    for user_msg, assistant_msg in history:
+        prompt += f"User: {user_msg}\nAssistant: {assistant_msg}\n"
 
-    prompt += "Assistant:"
+    prompt += f"User: {message}\nAssistant:"
 
     return prompt
 
 
 # ============================
-# Streaming Generator
+# Generate Response
 # ============================
 
-def generate_response(message, history):
+def chat(message, history):
 
-    messages = history + [{"role": "user", "content": message}]
+    history = history or []
 
-    prompt = build_prompt(messages)
+    prompt = build_prompt(message, history)
 
     output = ""
 
@@ -90,38 +86,18 @@ def generate_response(message, history):
         top_p=0.95,
         stream=True
     ):
-        text = token["choices"][0]["text"]
-        output += text
+        output += token["choices"][0]["text"]
         yield output
 
 
 # ============================
-# Gradio Chat Interface
-# ============================
-
-def chat(message, history):
-
-    history = history or []
-
-    assistant_response = ""
-
-    for partial in generate_response(message, history):
-        assistant_response = partial
-        yield history + [
-            {"role": "user", "content": message},
-            {"role": "assistant", "content": assistant_response},
-        ]
-
-
-# ============================
-# Launch UI
+# Launch Gradio ChatInterface
 # ============================
 
 demo = gr.ChatInterface(
     fn=chat,
     title="DeepSeek Coder 1.3B",
-    description="Production GGUF model running on llama.cpp",
-    type="messages"
+    description="Production GGUF model running on llama.cpp"
 )
 
 demo.launch(