Update app.py

app.py

@@ -1,70 +1,47 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
 
-
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
-    """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-    """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
-
-    messages = [{"role": "system", "content": system_message}]
-
-    messages.extend(history)
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    type="messages",
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+# Baixa o modelo uncensored (só na primeira vez)
+model_path = hf_hub_download(
+    repo_id="bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF",
+    filename="Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf"
 )
 
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
-
+# Carrega o modelo em CPU (n_gpu_layers=0 força CPU)
+llm = Llama(
+    model_path=model_path,
+    n_ctx=8192,        # contexto grande
+    n_batch=512,
+    n_gpu_layers=0,    # 0 = só CPU (essencial pro basic)
+    verbose=False
+)
 
-if __name__ == "__main__":
-    demo.launch()
+# Template simples pra chat (funciona bem com Llama)
+def format_prompt(message, history):
+    prompt = "<|begin_of_text|>"
+    for user, assistant in history:
+        prompt += f"<|start_header_id|>user<|end_header_id|>\n{user}<|eot_id|>"
+        prompt += f"<|start_header_id|>assistant<|end_header_id|>\n{assistant}<|eot_id|>"
+    prompt += f"<|start_header_id|>user<|end_header_id|>\n{message}<|eot_id|>"
+    prompt += "<|start_header_id|>assistant<|end_header_id|>\n"
+    return prompt
+
+def chat(message, history):
+    prompt = format_prompt(message, history)
+    output = llm(
+        prompt,
+        max_tokens=1024,
+        temperature=0.7,
+        top_p=0.95,
+        repeat_penalty=1.15,  # evita repetição
+        stop=["<|eot_id|>"]
+    )
+    return output["choices"][0]["text"].strip()
+
+# Interface Gradio
+gr.ChatInterface(
+    chat,
+    title="Llama 3.2 3B Uncensored (CPU Básica)",
+    description="IA sem censura rodando só em CPU gratuita! Respostas em ~10-20s."
+).launch()