feat: standalone cancel-all version - startup auto cancels all waiting/running tasks

- Target single GPU (ppu_zw_810e) instead of 4 GPUs
- Search only from HuggingFace API (removed ModelScope)
- Framework fixed to vllm (removed llama.cpp)
- Filter out GGUF format models
- Simplified config and functions

README.md

@@ -1,16 +1,33 @@
-# ModelHub 提交智能体
+# ModelHub 全云端提交智能体
 
-用于获取 `strategyId` 的最小化智能体。
+部署在平台容器中，从内网直接提交验证任务。
 
-## 功能
+## API
 
-- `GET /health` → 健康检查
-- `GET /` → 显示 strategyId
+| 端点 | 方法 | 说明 |
+|------|------|------|
+| `/health` | GET | 健康检查 |
+| `/` | GET | 智能体信息 |
+| `/status` | GET | 运行状态 |
+| `/logs?lines=50` | GET | 查看日志 |
+| `/run` | POST | 触发一次完整流程 |
 
-## 平台要求
+## POST /run 请求体
 
-- ✅ 根目录 Dockerfile
-- ✅ 暴露 8080 端口
-- ✅ `/health` 端点返回 200
-- ✅ 处理 SIGTERM 信号
-- ✅ 读取 `STRATEGY_ID` 环境变量
+```json
+{
+  "gpus": ["Kunlunxin_p-800", "hygon_k100-ai"],
+  "limit": 30
+}
+```
+
+- `gpus`: 目标 GPU 列表（可选，默认全部）
+- `limit`: 每 GPU 最大提交数（可选，默认 30）
+
+## 流程
+
+1. 搜索 ModelScope（Qwen 系列关键词）
+2. 下载量 ≥ 50 筛选
+3. 架构检查（config.json）
+4. 平台验证状态查重
+5. 提交到 ModelHub

main.py

@@ -1,74 +1,201 @@
 """
-ModelHub 提交智能体（代理模式）
-用途：获取 strategyId，本地脚本直接调用 ModelHub API
+ModelHub Cancel-All 智能体
+启动后自动取消当前账号所有 waiting/running 状态的验证任务
+用法：推送此版本到平台，选择对应 tag 运行即可
 """
+
 import json
 import os
 import signal
+import time
+import threading
+from datetime import datetime
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
 
+import requests
+
+# ============================================================
+# 配置
+# ============================================================
+
 HOST = "0.0.0.0"
 PORT = 8080
-shutdown_requested = False
+
+# 目标GPU和Token（跟提交版本保持一致）
+TARGET_GPU = "Iluvatar_bi-150"
+TARGET_TOKEN = "f45f1aae2c094426be237c88b1085015"
+
+MODELHUB_API = "https://modelhub.org.cn/api"
+BATCH_SIZE = 50  # 每批取消数量（API上限）
+
+# ============================================================
+# 全局状态
+# ============================================================
+
+state = {'running': False, 'last_run': None, 'logs': []}
+state_lock = threading.Lock()
 
 
-def _config() -> dict:
-    strategy_id = os.getenv("STRATEGY_ID", "")
-    return {
-        "strategy_id": strategy_id,
-        "status": "running",
+def log(msg: str):
+    ts = datetime.now().strftime('%H:%M:%S')
+    line = f"[{ts}] {msg}"
+    print(line, flush=True)
+    with state_lock:
+        state['logs'].append(line)
+        if len(state['logs']) > 500:
+            state['logs'] = state['logs'][-300:]
+
+
+def cancel_all_tasks():
+    """取消所有 waiting 和 running 状态的任务"""
+    headers = {
+        'Xc-Token': TARGET_TOKEN,
+        'Accept': 'application/json',
+        'Content-Type': 'application/json',
     }
 
+    log("=" * 50)
+    log("开始取消所有任务")
+    log(f"目标GPU: {TARGET_GPU}")
 
-class Handler(BaseHTTPRequestHandler):
-    def do_GET(self) -> None:
-        if self.path == "/health":
-            self._send_json({"status": "ok"})
+    # 1. 查询所有任务
+    all_tasks = []
+    page = 1
+    while True:
+        try:
+            resp = requests.get(
+                f"{MODELHUB_API}/adapt/task/page",
+                headers={'Xc-Token': TARGET_TOKEN, 'Accept': 'application/json'},
+                params={'current': page, 'pageSize': 100, 'onlyMine': 'true',
+                        'gpuType': TARGET_GPU},
+                timeout=15
+            )
+            data = resp.json()
+            if data.get('code') != 0:
+                log(f"查询失败: code={data.get('code')}")
+                break
+            records = data['data'].get('records', [])
+            if not records:
+                break
+            all_tasks.extend(records)
+            total = data['data'].get('total', 0)
+            log(f"  第{page}页: {len(records)} 条, 累计 {len(all_tasks)} / {total}")
+            if len(all_tasks) >= int(total):
+                break
+            page += 1
+        except Exception as e:
+            log(f"  查询异常: {e}")
             return
 
-        if self.path == "/":
-            self._send_json({
-                "name": "modelhub-submit-agent",
-                "config": _config(),
+    # 2. 筛选 waiting/running 状态
+    cancellable = [t for t in all_tasks if t.get('status') in ('waiting', 'running')]
+    if not cancellable:
+        log(f"  没有可取消的任务 (总任务 {len(all_tasks)} 个)")
+        log("=" * 50)
+        return
+
+    task_ids = [int(t.get('taskId')) for t in cancellable if t.get('taskId')]
+    log(f"  总任务 {len(all_tasks)} 个, 可取消 {len(task_ids)} 个")
+
+    # 3. 分批取消
+    url = f"{MODELHUB_API}/async/task/stop-create-contest-task"
+    cancelled = 0
+    for i in range(0, len(task_ids), BATCH_SIZE):
+        batch = task_ids[i:i + BATCH_SIZE]
+        try:
+            resp = requests.put(url, headers=headers, json={'taskIds': batch}, timeout=15)
+            data = resp.json()
+            if data.get('code') == 0:
+                cancelled += len(batch)
+                log(f"  批次 {i // BATCH_SIZE + 1}: 取消 {len(batch)} 个 OK")
+            else:
+                log(f"  批次 {i // BATCH_SIZE + 1}: 失败 code={data.get('code')} msg={data.get('message', '')[:80]}")
+        except Exception as e:
+            log(f"  批次 {i // BATCH_SIZE + 1}: 异常 {e}")
+        time.sleep(1)
+
+    log(f"  取消完成: {cancelled} / {len(task_ids)}")
+    log("=" * 50)
+
+
+# ============================================================
+# HTTP 服务
+# ============================================================
+
+class AgentHandler(BaseHTTPRequestHandler):
+    def do_GET(self):
+        if self.path == '/health':
+            self._json({'status': 'ok'})
+        elif self.path == '/':
+            self._json({
+                'name': 'modelhub-cancel-all',
+                'gpu': TARGET_GPU,
+                'status': 'running' if state['running'] else 'idle',
+                'last_result': state.get('last_result'),
             })
-            return
+        elif self.path.startswith('/logs'):
+            lines = 100
+            self._json({'logs': state['logs'][-lines:]})
+        else:
+            self._json({'error': 'not found'}, 404)
 
-        self._send_json({"error": "not found"}, status=404)
-
-    def log_message(self, fmt: str, *args: object) -> None:
-        print(f"{self.address_string()} - {fmt % args}", flush=True)
-
-    def _send_json(self, body: dict, status: int = 200) -> None:
-        payload = json.dumps(body).encode()
+    def _json(self, body: dict, status: int = 200):
+        payload = json.dumps(body, ensure_ascii=False).encode()
         self.send_response(status)
-        self.send_header("Content-Type", "application/json")
-        self.send_header("Content-Length", str(len(payload)))
+        self.send_header('Content-Type', 'application/json; charset=utf-8')
+        self.send_header('Content-Length', str(len(payload)))
         self.end_headers()
         self.wfile.write(payload)
 
+    def log_message(self, fmt, *args):
+        pass
 
-def _handle_signal(signum: int, _frame: object) -> None:
+
+# ============================================================
+# 启动
+# ============================================================
+
+shutdown_requested = False
+
+
+def _handle_signal(signum, _frame):
     global shutdown_requested
     shutdown_requested = True
-    print(f"received signal {signum}, shutting down", flush=True)
+    log(f"收到信号 {signum}，准备关闭")
 
 
-def main() -> None:
+def main():
     signal.signal(signal.SIGTERM, _handle_signal)
     signal.signal(signal.SIGINT, _handle_signal)
 
-    server = ThreadingHTTPServer((HOST, PORT), Handler)
+    server = ThreadingHTTPServer((HOST, PORT), AgentHandler)
     server.timeout = 1
 
-    strategy_id = os.getenv("STRATEGY_ID", "NOT_SET")
-    print(f"modelhub-submit-agent listening on {HOST}:{PORT}", flush=True)
-    print(f"STRATEGY_ID: {strategy_id}", flush=True)
+    log(f"Cancel-All 智能体启动 | {HOST}:{PORT}")
+    log(f"目标GPU: {TARGET_GPU}")
+
+    # 启动后自动取消所有任务
+    def _auto_cancel():
+        time.sleep(2)
+        try:
+            state['running'] = True
+            state['last_run'] = datetime.now().isoformat()
+            cancel_all_tasks()
+            state['last_result'] = {'success': True}
+        except Exception as e:
+            log(f"异常: {e}")
+            state['last_result'] = {'error': str(e)}
+        finally:
+            state['running'] = False
+
+    threading.Thread(target=_auto_cancel, daemon=True).start()
 
     while not shutdown_requested:
         server.handle_request()
 
     server.server_close()
+    log("智能体已关闭")
 
 
-if __name__ == "__main__":
+if __name__ == '__main__':
     main()

requirements.txt

@@ -1 +1,2 @@
-# No external dependencies needed
+requests>=2.28.0
+pyyaml>=6.0