@agentunion/kite 1.3.2 → 1.5.0

package/extensions/services/model_service/entry.py

@@ -23,7 +23,113 @@ import websockets
 
 
 # ── Module configuration ──
-MODULE_NAME = "model_service"
+
+def _load_module_config() -> dict:
+    """Load module configuration from module.md frontmatter.
+
+    Returns:
+        Dict with keys: name, preferred_port, advertise_ip
+
+    Raises:
+        SystemExit: If module.md is invalid or name is non-compliant
+    """
+    _this_dir = os.path.dirname(os.path.abspath(__file__))
+    module_md = os.path.join(_this_dir, "module.md")
+
+    # Calculate relative path for error messages
+    project_root = os.environ.get("KITE_PROJECT", "")
+    if project_root and _this_dir.startswith(project_root):
+        rel_path = os.path.relpath(_this_dir, project_root)
+    else:
+        rel_path = _this_dir
+
+    # Default values (will be overridden if valid config exists)
+    result = {
+        "name": "",
+        "preferred_port": 0,
+        "advertise_ip": "0.0.0.0"
+    }
+
+    # Check if module.md exists
+    if not os.path.exists(module_md):
+        print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+        print(f"  Path: {rel_path}/module.md")
+        print(f"  Reason: File not found")
+        sys.exit(1)
+
+    try:
+        with open(module_md, encoding="utf-8") as f:
+            text = f.read()
+
+        # Extract YAML frontmatter (between --- markers)
+        import re
+        m = re.match(r'^---\s*\n(.*?)\n---', text, re.DOTALL)
+        if not m:
+            print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+            print(f"  Path: {rel_path}/module.md")
+            print(f"  Reason: Missing YAML frontmatter")
+            sys.exit(1)
+
+        # Parse YAML frontmatter
+        try:
+            import yaml
+            fm = yaml.safe_load(m.group(1)) or {}
+        except ImportError:
+            print(f"[{rel_path}] ERROR: PyYAML not installed, cannot parse module.md")
+            sys.exit(1)
+        except Exception as e:
+            print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+            print(f"  Path: {rel_path}/module.md")
+            print(f"  Reason: YAML parse error: {e}")
+            sys.exit(1)
+
+        # Validate 'name' field (required)
+        if "name" not in fm:
+            print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+            print(f"  Path: {rel_path}/module.md")
+            print(f"  Reason: Missing 'name' field")
+            sys.exit(1)
+
+        raw_name = str(fm["name"]).strip()
+
+        if not raw_name:
+            print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+            print(f"  Path: {rel_path}/module.md")
+            print(f"  Reason: Empty module name")
+            sys.exit(1)
+
+        # Validate name characters
+        sanitized = re.sub(r'[^a-zA-Z0-9_\-]', '', raw_name)
+
+        if sanitized != raw_name:
+            invalid_chars = ''.join(sorted(set(c for c in raw_name if c not in sanitized)))
+            print(f"[{rel_path}] ERROR: Invalid module configuration in module.md")
+            print(f"  Path: {rel_path}/module.md")
+            print(f"  Reason: Invalid characters in name '{raw_name}': {repr(invalid_chars)}")
+            sys.exit(1)
+
+        result["name"] = sanitized
+
+        # Extract optional fields
+        if "preferred_port" in fm:
+            try:
+                result["preferred_port"] = int(fm["preferred_port"])
+            except (ValueError, TypeError):
+                pass
+
+        if "advertise_ip" in fm:
+            result["advertise_ip"] = str(fm["advertise_ip"])
+
+    except SystemExit:
+        raise  # Re-raise exit to prevent catching by outer except
+    except Exception as e:
+        print(f"[{rel_path}] ERROR: Failed to read module.md: {e}")
+        sys.exit(1)
+
+    return result
+
+_module_config = _load_module_config()
+MODULE_NAME = _module_config["name"]
 
 
 class _SafeWriter:
@@ -264,6 +370,7 @@ def _read_stdin_kite_message(expected_type: str, timeout: float = 10) -> dict |
 # Global WS reference for publish_event callback
 _ws_global = None
 _shutting_down = False
+_exit_code = 0  # Exit code for main() to use
 
 
 def _is_auth_failure(e: Exception) -> bool:
@@ -333,7 +440,7 @@ async def main():
 
 async def _ws_loop(token: str, kernel_port: int, _t0: float):
     """Connect to Kernel with exponential backoff reconnection."""
-    global _shutting_down
+    global _shutting_down, _exit_code
     retry_delay = 0.3
     max_delay = 5.0
     max_retries = 10
@@ -349,10 +456,14 @@ async def _ws_loop(token: str, kernel_port: int, _t0: float):
             attempt += 1
             if _is_auth_failure(e):
                 print(f"[model_service] Kernel 认证失败，退出")
-                sys.exit(1)
+                _exit_code = 1
+                _shutting_down = True
+                return
             if attempt >= max_retries:
                 print(f"[model_service] 重连失败 {max_retries} 次，退出")
-                sys.exit(1)
+                _exit_code = 1
+                _shutting_down = True
+                return
             _write_crash(type(e), e, e.__traceback__, severity="error", handled=True)
             print(f"[model_service] 连接错误: {e}, {retry_delay:.1f}s 后重试 ({attempt}/{max_retries})")
         _ws_global_clear()
@@ -374,7 +485,7 @@ async def _ws_connect(token: str, kernel_port: int, _t0: float):
     ws_url = f"ws://127.0.0.1:{kernel_port}/ws?token={token}&id=model_service"
     print(f"[model_service] Connecting to Kernel: {ws_url}")
 
-    async with websockets.connect(ws_url, open_timeout=5, ping_interval=None, ping_timeout=None, close_timeout=10) as ws:
+    async with websockets.connect(ws_url, open_timeout=5, ping_interval=None, close_timeout=10) as ws:
         _ws_global = ws
         print(f"[model_service] Connected to Kernel ({_fmt_elapsed(_t0)})")
 
@@ -393,7 +504,9 @@ async def _ws_connect(token: str, kernel_port: int, _t0: float):
             "module_id": "model_service",
             "module_type": "service",
             "events_publish": {
-                "model_service.test": {"description": "Test event from model_service module"},
+                "model_service": {
+                    "test": {"description": "Test event from model_service module"},
+                }
             },
             "events_subscribe": [
                 "module.started",
@@ -405,12 +518,14 @@ async def _ws_connect(token: str, kernel_port: int, _t0: float):
 
         # Publish module.ready (every reconnect)
         if not _shutting_down:
+            startup_time = time.monotonic() - _t0
             await _rpc_call(ws, "event.publish", {
                 "event_id": str(uuid.uuid4()),
                 "event": "module.ready",
                 "data": {
                     "module_id": "model_service",
                     "graceful_shutdown": True,
+                    "startup_time": startup_time,
                 },
             })
             print(f"[model_service] module.ready published ({_fmt_elapsed(_t0)})")
@@ -419,6 +534,11 @@ async def _ws_connect(token: str, kernel_port: int, _t0: float):
         test_task = asyncio.create_task(_test_event_loop(ws))
 
         # Message loop: handle incoming RPC + events
+        # CRITICAL: RPC 死锁防范
+        # - 入站 RPC 请求必须用 create_task() 异步执行，不可 await
+        # - 原因：如果 handler 内部调用 rpc_call() 发出站请求，出站响应需要本接收循环来分发
+        # - 如果接收循环被 await handler 阻塞，出站响应永远收不到 → 超时死锁
+        # - 事件通知和 RPC 响应可以同步处理（它们不会反向调用 rpc_call）
         async for raw in ws:
             try:
                 msg = json.loads(raw)
@@ -433,8 +553,8 @@ async def _ws_connect(token: str, kernel_port: int, _t0: float):
                     # Event Notification
                     await _handle_event_notification(msg)
                 elif has_method and has_id:
-                    # Incoming RPC request
-                    await _handle_rpc_request(ws, msg)
+                    # Incoming RPC request — run in background to prevent deadlock
+                    asyncio.create_task(_handle_rpc_request(ws, msg))
                 # Ignore RPC responses (we don't await them in this simple impl)
             except Exception as e:
                 print(f"[model_service] 消息处理异常（已忽略）: {e}")
@@ -457,12 +577,32 @@ async def _publish_event(ws, event: dict):
     })
 
 
+async def _handle_ping_event(data: dict):
+    """Handle system.ping event and reply with system.pong."""
+    t1 = data.get("ping_time")
+    t2 = time.time()
+
+    await _publish_event(_ws_global, {
+        "event": "system.pong",
+        "data": {
+            "module_id": MODULE_NAME,
+            "ping_time": t1,
+            "pong_time": t2,
+        },
+    })
+
+
 async def _handle_event_notification(msg: dict):
     """Handle an event notification (JSON-RPC 2.0 Notification with method='event')."""
     params = msg.get("params", {})
     event_type = params.get("event", "")
     data = params.get("data", {})
 
+    # Handle system.ping event
+    if event_type == "system.ping":
+        await _handle_ping_event(data)
+        return
+
     # Special handling for module.shutdown
     if event_type == "module.shutdown":
         target = data.get("module_id", "")
@@ -523,29 +663,37 @@ async def _rpc_status() -> dict:
 
 
 async def _handle_shutdown():
-    """Handle module.shutdown event — exiting → ack → cleanup → ready → exit."""
+    """Handle module.shutdown event — ack → exiting → cleanup → ready → exit."""
     global _shutting_down
     print("[model_service] Received shutdown request")
     _shutting_down = True
-    # Step 0: Send module.exiting
+    # Step 1: Send ack (立即确认收到)
     await _publish_event(_ws_global, {
-        "event": "module.exiting",
-        "data": {"module_id": "model_service", "action": "none"},
+        "event": "module.shutdown.ack",
+        "data": {"module_id": "model_service"},
     })
-    # Step 1: Send ack
+    # Step 2: Send module.exiting (开始清理)
     await _publish_event(_ws_global, {
-        "event": "module.shutdown.ack",
-        "data": {"module_id": "model_service", "estimated_cleanup": 2},
+        "event": "module.exiting",
+        "data": {
+            "module_id": "model_service",
+            "type": "passive",
+            "reason": "shutdown_requested",
+            "restart": "auto",
+            "action": "none",
+            "timeout": 2.0,
+            "restart_delay": 0.0,
+        },
     })
-    # Step 2: Cleanup (nothing to clean up for model_service)
-    # Step 3: Send ready
+    # Step 3: Cleanup (nothing to clean up for model_service)
+    # Step 4: Send ready (清理完成)
     await _publish_event(_ws_global, {
         "event": "module.shutdown.ready",
         "data": {"module_id": "model_service"},
     })
     print("[model_service] Shutdown ready, exiting")
-    # Step 4: Exit
-    sys.exit(0)
+    # Step 5: Exit
+    sys.exit(_exit_code)
 
 
 async def _test_event_loop(ws):

package/extensions/services/model_service/module.md

@@ -1,22 +1,21 @@
----
-name: model_service
-display_name: Model Service
-version: "1.0"
-type: service
-state: enabled
-runtime: python
-entry: entry.py
-events:
-  - model_service.test
-subscriptions:
-  - module.started
-  - module.stopped
-  - module.shutdown
----
-
-# Model Service（大模型服务）
-
-大模型服务模块，提供统一的 LLM 调用接口。
-
-- 模型调用 — 封装多种大模型 API 的统一调用接口
-- 事件通知 — 通过 Event Hub 发布模型服务状态事件
+---
+name: model_service
+display_name: Model Service
+version: '1.0'
+type: service
+state: manual
+runtime: python
+entry: entry.py
+events:
+- model_service.test
+subscriptions:
+- module.started
+- module.stopped
+- module.shutdown
+---
+# Model Service（大模型服务）
+
+大模型服务模块，提供统一的 LLM 调用接口。
+
+- 模型调用 — 封装多种大模型 API 的统一调用接口
+- 事件通知 — 通过 Event Hub 发布模型服务状态事件

package/extensions/services/proxy/.claude/settings.local.json

@@ -0,0 +1,13 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(echo agentcp not found directly:*)",
+      "Bash(grep:*)",
+      "Bash(head:*)",
+      "Bash(find H:\\\\project\\\\evol_main\\\\evol-electron-app\\\\python_sample_backend:*)",
+      "Bash(python3:*)",
+      "Bash(python:*)",
+      "Bash(cd H:\\\\project\\\\evol_main\\\\evol-electron-app\\\\python_sample_backend:*)"
+    ]
+  }
+}

package/extensions/services/proxy/CHANGELOG_20260308.md

@@ -0,0 +1,258 @@
+# Evol Proxy 服务修改说明
+
+## 修改日期
+2026-03-08
+
+## 修改内容
+
+### 1. 端口改为系统自动分配
+
+**修改文件：**
+- `main.py`
+- `server.py`
+
+**改动说明：**
+
+原来使用固定端口 12655，容易导致端口冲突。现在改为：
+- 使用 `port=0` 让系统自动分配可用端口
+- 在启动前先获取一个可用端口
+- 启动后显示实际使用的端口号
+
+**实现方式：**
+
+```python
+# 获取系统分配的可用端口
+import socket
+sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+sock.bind(('', 0))
+actual_port = sock.getsockname()[1]
+sock.close()
+
+# 使用该端口启动服务器
+await run_server(port=actual_port)
+```
+
+### 2. 登录后自动拉取模型列表和用户信息
+
+**修改文件：**
+- `main.py`
+
+**新增功能：**
+
+在 AID 上线成功后，自动执行以下操作：
+
+1. **拉取模型列表**
+   - 调用 `get_models_list()` 获取可用模型
+   - 显示模型数量和前 5 个模型名称
+   - 示例输出：
+     ```
+     [Models] Found 15 available models:
+       - gpt-4
+       - gpt-3.5-turbo
+       - claude-3-opus
+       - claude-3-sonnet
+       - gemini-pro
+       ... and 10 more
+     ```
+
+2. **获取用户信息**
+   - 调用 `userPresenter.get_user_info()` 获取用户信息
+   - 显示手机号和用户 ID
+   - 示例输出：
+     ```
+     [User] Logged in as: 138****1234
+     [User] User ID: user_abc123
+     ```
+
+**代码位置：**
+
+在 `async_main()` 函数中，Step 3.6 部分：
+
+```python
+# Step 3.6: 拉取模型列表和用户信息
+print("\n[Info] Fetching models and user info...")
+try:
+    # 拉取模型列表
+    from evol.server.openclaw_proxy import get_models_list
+    models_data = get_models_list()
+    model_count = len(models_data.get("data", []))
+    print(f"[Models] Found {model_count} available models:")
+    for model in models_data.get("data", [])[:5]:
+        print(f"  - {model.get('id', 'unknown')}")
+    if model_count > 5:
+        print(f"  ... and {model_count - 5} more")
+
+    # 获取用户信息
+    from evol.presenter.userPresenter import userPresenter
+    user_info = userPresenter.get_user_info()
+    if user_info:
+        print(f"[User] Logged in as: {user_info.get('phone', 'unknown')}")
+        print(f"[User] User ID: {user_info.get('id', 'unknown')}")
+
+except Exception as e:
+    print(f"[Info] Warning: 获取信息失败: {e}")
+```
+
+## 启动流程变化
+
+### 修改前
+
+```
+1. 初始化配置目录
+2. 检查本地 token
+3. 若无有效 token → 控制台登录
+4. 同步用户信息 & AID 上线
+5. 获取 Evol 本地密钥
+6. 启动 FastAPI 服务器（固定端口 12655）
+```
+
+### 修改后
+
+```
+1. 初始化配置目录
+2. 检查本地 token
+3. 若无有效 token → 控制台登录
+4. 同步用户信息 & AID 上线
+5. 获取 Evol 本地密钥
+6. 拉取模型列表和用户信息 ← 新增
+7. 启动 FastAPI 服务器（系统自动分配端口）← 修改
+```
+
+## 输出示例
+
+### 修改前
+
+```
+============================================================
+  Evol Sample Backend
+  Port: 12655
+============================================================
+
+[Init] Config dir: C:\Users\xxx\.evol
+[Auth] Token valid, skip login
+[AID] Starting AID login...
+Evol: aid_xxx is online
+[Key] Evol API Key: evol-xxx
+
+[Server] Starting on port 12655...
+[Server] ANTHROPIC_BASE_URL = http://127.0.0.1:12655/claude-proxy
+[Server] OPENAI_BASE_URL    = http://127.0.0.1:12655/openclaw-proxy/v1
+[Server] Health check       = http://127.0.0.1:12655/health
+[Server] Evol API Key       = evol-xxx
+
+Press Ctrl+C to stop.
+```
+
+### 修改后
+
+```
+============================================================
+  Evol Sample Backend
+  Port: Auto-assigned by system
+============================================================
+
+[Init] Config dir: C:\Users\xxx\.evol
+[Auth] Token valid, skip login
+[AID] Starting AID login...
+Evol: aid_xxx is online
+[Key] Evol API Key: evol-xxx
+
+[Info] Fetching models and user info...
+[Models] Found 15 available models:
+  - gpt-4
+  - gpt-3.5-turbo
+  - claude-3-opus
+  - claude-3-sonnet
+  - gemini-pro
+  ... and 10 more
+[User] Logged in as: 138****1234
+[User] User ID: user_abc123
+
+[Server] Starting server...
+
+[Server] Server will start on port 54321
+[Server] ANTHROPIC_BASE_URL = http://127.0.0.1:54321/claude-proxy
+[Server] OPENAI_BASE_URL    = http://127.0.0.1:54321/openclaw-proxy/v1
+[Server] Health check       = http://127.0.0.1:54321/health
+[Server] Evol API Key       = evol-xxx
+
+Press Ctrl+C to stop.
+```
+
+## 优点
+
+### 1. 端口自动分配
+
+- **避免端口冲突** - 不会与其他服务冲突
+- **支持多实例** - 可以同时运行多个代理服务
+- **更灵活** - 适应不同的部署环境
+
+### 2. 自动拉取信息
+
+- **验证登录状态** - 确认模型列表可以正常获取
+- **显示可用模型** - 用户一目了然有哪些模型可用
+- **确认用户身份** - 显示当前登录的用户信息
+- **便于调试** - 启动时就能发现配置问题
+
+## 注意事项
+
+1. **端口号变化** - 每次启动端口可能不同，需要从启动日志中获取
+2. **防火墙规则** - 如果需要外部访问，需要动态配置防火墙
+3. **错误处理** - 如果拉取模型列表失败，只会显示警告，不会中断启动
+
+## 测试建议
+
+1. **基本功能测试**
+   ```bash
+   # 启动服务
+   python main.py
+
+   # 查看启动日志，获取实际端口（假设是 54321）
+   # 测试健康检查
+   curl http://127.0.0.1:54321/health
+
+   # 测试模型列表
+   curl http://127.0.0.1:54321/openclaw-proxy/v1/models \
+     -H "Authorization: Bearer evol-xxx"
+   ```
+
+2. **多实例测试**
+   ```bash
+   # 同时启动多个实例，验证端口不冲突
+   python main.py &
+   python main.py &
+   ```
+
+3. **端口冲突测试**
+   ```bash
+   # 验证不会出现端口占用错误
+   # 之前固定端口时可能出现的问题现在应该不会发生
+   ```
+
+## 回滚方案
+
+如果需要回滚到固定端口模式，修改 `main.py`：
+
+```python
+# 改回固定端口
+PORT = 12655  # 或其他固定端口
+
+# 恢复端口清理逻辑
+kill_port(PORT)
+
+# 直接使用固定端口启动
+await run_server(port=PORT)
+```
+
+## 相关文件
+
+- `main.py` - 主程序入口
+- `server.py` - FastAPI 服务器
+- `console_auth.py` - 控制台认证
+- `evol/server/openclaw_proxy.py` - OpenClaw 代理实现
+- `evol/presenter/userPresenter.py` - 用户信息管理
+
+---
+
+**修改者：** Claude (Kite Assistant)
+**审核者：** 待审核