Create app.py

app.py

@@ -0,0 +1,78 @@
+import streamlit as st
+from llama_cpp import Llama
+from huggingface_hub import hf_hub_download
+import os
+
+# 配置：使用3B以下量化模型（CPU友好）
+MODEL_REPO = "Qwen/Qwen2.5-1.5B-Instruct-GGUF"
+MODEL_FILENAME = "qwen2.5-1_5b-instruct-q4_k_m.gguf"
+
+@st.cache_resource
+def load_model():
+    """下载并加载GGUF模型（首次启动自动下载）"""
+    model_path = hf_hub_download(
+        repo_id=MODEL_REPO,
+        filename=MODEL_FILENAME,
+        local_dir="models"
+    )
+    
+    # CPU优化配置：限制线程数避免OOM
+    llm = Llama(
+        model_path=model_path,
+        n_ctx=2048,          # 上下文长度
+        n_threads=4,         # CPU线程数（Spaces通常4核）
+        n_gpu_layers=0,      # 0=纯CPU推理
+        verbose=False
+    )
+    return llm
+
+# Streamlit UI
+st.title("🦙 本地LLM (CPU-only)")
+st.caption("基于Qwen2.5-1.5B量化模型 · 无外部API调用")
+
+if "llm" not in st.session_state:
+    with st.spinner("⏳ 首次加载模型（约2-3分钟）..."):
+        st.session_state.llm = load_model()
+    st.success("✅ 模型加载完成！")
+
+# 聊天界面
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+for message in st.session_state.messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+if prompt := st.chat_input("输入您的问题..."):
+    # 显示用户消息
+    st.session_state.messages.append({"role": "user", "content": prompt})
+    with st.chat_message("user"):
+        st.markdown(prompt)
+    
+    # 生成回复（CPU推理较慢，需显示加载状态）
+    with st.chat_message("assistant"):
+        message_placeholder = st.empty()
+        full_response = ""
+        
+        # Qwen2.5的对话模板
+        messages = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            *[
+                {"role": m["role"], "content": m["content"]}
+                for m in st.session_state.messages
+            ]
+        ]
+        
+        # 流式生成
+        for chunk in st.session_state.llm.create_chat_completion(
+            messages=messages,
+            max_tokens=512,
+            temperature=0.7,
+            stream=True
+        ):
+            if "content" in chunk["choices"][0]["delta"]:
+                full_response += chunk["choices"][0]["delta"]["content"]
+                message_placeholder.markdown(full_response + "▌")
+        
+        message_placeholder.markdown(full_response)
+        st.session_state.messages.append({"role": "assistant", "content": full_response})