Update app.py

app.py

@@ -17,21 +17,84 @@ if not os.path.exists(MODEL_PATH):
 
 llm = Llama(
     model_path=MODEL_PATH,
-    n_ctx=4096,
+    n_ctx=4096, # increased context window.
     n_threads=2,
     n_threads_batch=2,
     verbose=False,
 )
 
-def generate_response(message, history):
-    prompt = f"{message}"
-    output = llm(prompt, max_tokens=128, echo=False)
+def generate_response(message, history, system_prompt, max_tokens, temperature, top_p):
+    messages = [{"role": "system", "content": system_prompt}]
+    for item in history:
+        messages.append(item)
+    messages.append({"role": "user", "content": message})
+
+    prompt = "".join([f"{m['role'].capitalize()}: {m['content']}\n" for m in messages])
+
+    output = llm(
+        prompt,
+        max_tokens=max_tokens,
+        temperature=temperature,
+        top_p=top_p,
+        echo=False,
+    )
     return output["choices"][0]["text"].strip()
 
-iface = gr.ChatInterface(
-    fn=generate_response,
-    title="llama.cpp Chat",
-    description="Chat with a GGUF model.",
-)
+with gr.Blocks() as iface:
+    system_prompt_input = gr.Textbox(
+        label="System Prompt",
+        value="You are a helpful assistant.",
+        lines=3,
+    )
+    max_tokens_slider = gr.Slider(
+        label="Max Tokens",
+        minimum=32,
+        maximum=4096,
+        value=128,
+        step=32,
+    )
+    temperature_slider = gr.Slider(
+        label="Temperature",
+        minimum=0.1,
+        maximum=1.0,
+        value=0.8,
+        step=0.1,
+    )
+    top_p_slider = gr.Slider(
+        label="Top P",
+        minimum=0.1,
+        maximum=1.0,
+        value=0.9,
+        step=0.1,
+    )
+    chatbot = gr.Chatbot(type="messages")
+    message = gr.Textbox(label="Message")
+    send_button = gr.Button("Send")
+    state = gr.State([])
+
+    send_button.click(
+        generate_response,
+        inputs=[
+            message,
+            state,
+            system_prompt_input,
+            max_tokens_slider,
+            temperature_slider,
+            top_p_slider,
+        ],
+        outputs=[chatbot, state],
+    )
+    message.submit(
+        generate_response,
+        inputs=[
+            message,
+            state,
+            system_prompt_input,
+            max_tokens_slider,
+            temperature_slider,
+            top_p_slider,
+        ],
+        outputs=[chatbot, state],
+    )
 
 iface.launch()