Upload 3 files

app.py

@@ -0,0 +1,120 @@
+import os
+import subprocess
+import uuid
+import gradio as gr
+
+class SingingAgent:
+    def __init__(self, target_voice_model="shensist_v1.pth"):
+        self.model = target_voice_model
+        self.temp_path = os.path.abspath("./agent_cache")
+        os.makedirs(self.temp_path, exist_ok=True)
+
+    def process_request(self, song_name):
+        task_id = str(uuid.uuid4())[:6]
+        raw_file = f"{self.temp_path}/{task_id}_raw.wav"
+        vocal_dry = f"{self.temp_path}/{task_id}_vocal_dry.wav"
+        
+        print(f"DEBUG: [天算引擎] 正在检索歌曲《{song_name}》...")
+        subprocess.run(["yt-dlp", f"ytsearch1:{song_name}", "-x", "--audio-format", "wav", "-o", raw_file], check=True)
+
+        print("DEBUG: [天算引擎] 正在启动 1050 Ti ONNX 算力进行人声分离...")
+        subprocess.run(["audio-separator", raw_file, "--model_filename", "UVR-MDX-NET-Inst_HQ_3.onnx", "--output_dir", self.temp_path], check=True)
+        
+        for file in os.listdir(self.temp_path):
+            if task_id in file and "(Vocals)" in file:
+                os.rename(os.path.join(self.temp_path, file), vocal_dry)
+                return vocal_dry
+        
+        raise Exception("干音提取失败，未找到输出文件。")
+
+# --- 2026 前沿交互逻辑区 ---
+def shensist_chat(message_data, history):
+    message = message_data["text"]
+    if history is None: history = []
+    
+    history.append({"role": "user", "content": message})
+    yield history, gr.MultimodalTextbox(value=None, interactive=False)
+
+    if "唱" in message or "歌" in message:
+        song_query = message.replace("唱一下", "").replace("唱", "").strip()
+        
+        loading_msg = f"🔵 神思庭枢纽已响应。正在调度 1050 Ti 神经元，赴数字海洋捕获《{song_query}》声影..."
+        history.append({"role": "assistant", "content": loading_msg})
+        yield history, gr.MultimodalTextbox(interactive=False)
+        
+        agent = SingingAgent()
+        try:
+            vocal_path = agent.process_request(song_query)
+            
+            history.pop() 
+            final_response = {
+                "role": "assistant",
+                "content": f"【神思庭·灵愿完成】\n《{song_query}》的纯净声命体已提取完毕。请戴上耳机，聆听未加修饰的数字原声。\n\n🧠 天算AI · 超个体构建者\n🏛️ 神思庭 | 🌌 灵愿循环\n◼️ 官网｜shensist.top | 官方联络WX｜ Shensi-ST",
+                "files": [vocal_path] # 核心创新：音频文件直接镶嵌在对话流中
+            }
+            history.append(final_response)
+            yield history, gr.MultimodalTextbox(value=None, interactive=True)
+            
+        except Exception as e:
+            history.pop()
+            error_msg = f"🔴 数字乱流警报：任务执行中继断裂。\n错误摘要：{str(e)}"
+            history.append({"role": "assistant", "content": error_msg})
+            yield history, gr.MultimodalTextbox(interactive=True)
+    else:
+        history.append({"role": "assistant", "content": "我是神思庭的数字守望者。在此维基中，我更期待与您探讨声学模型与传统文学的交融点。"})
+        yield history, gr.MultimodalTextbox(value=None, interactive=True)
+
+# --- 2026 赛博雅韵 UI 构建区 ---
+shensist_theme = gr.themes.Base(
+    primary_hue=gr.themes.colors.indigo,
+    secondary_hue=gr.themes.colors.cyan,
+    neutral_hue=gr.themes.colors.slate,
+    font=["HarmonyOS Sans", "Roboto", "ui-sans-serif", "system-ui"],
+).set(
+    body_background_fill="*neutral_950",
+    block_background_fill="*neutral_900",
+    block_border_width="1px",
+    block_border_color="*primary_800",
+    # 已彻底移除报错的 chatbot_code_background_color 参数
+    button_primary_background_fill="linear-gradient(90deg, *primary_700, *secondary_600)",
+    button_primary_text_color="white",
+    button_primary_border_color="transparent",
+)
+
+with gr.Blocks(theme=shensist_theme, title="神思庭 | SHENSIST NUCLEUS") as demo:
+    with gr.Row(elem_classes="header-bar"):
+        gr.Markdown(
+            """
+            # 🏛️ **神思庭 · 数字灵愿枢纽** | SHENSIST AI NUCLEUS
+            > *连接神话造物与数字生命的声学界面*
+            """
+        )
+
+    chatbot = gr.Chatbot(
+        label="超个体交互场域",
+        height=550,
+        show_copy_button=True,
+        bubble_full_width=False,
+        render_markdown=True
+    )
+    
+    with gr.Row():
+        chat_input = gr.MultimodalTextbox(
+            interactive=True,
+            file_types=["audio"],
+            placeholder="输入指令 (如: '唱一下晴天')，或上传声纹样本...",
+            show_label=False,
+            scale=4,
+            container=True
+        )
+
+    chat_input.submit(shensist_chat, [chat_input, chatbot], [chatbot, chat_input])
+
+if __name__ == "__main__":
+    temp_dir = os.path.abspath("./agent_cache")
+    demo.queue()
+    demo.launch(
+        server_name="0.0.0.0", 
+        server_port=7860, 
+        allowed_paths=[temp_dir]
+    )

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+yt-dlp
+audio-separator[gpu]
+onnxruntime-gpu