@sonicbotman/lobster-press 3.2.4 → 3.2.6

package/mcp_server/README.md

@@ -0,0 +1,207 @@
+# LobsterPress MCP Server
+
+> **推荐方式**：如果你在使用 OpenClaw，请优先使用
+> [插件安装模式](../README.md#-openclaw-插件推荐)，
+> 更简单，无需手动管理进程。
+>
+> 本文档描述直接运行 MCP Server 的高级用法，适合非 OpenClaw 环境
+> 或需要自定义集成的场景。
+
+基于 Model Context Protocol (MCP) 的 OpenClaw 会话压缩服务。
+
+## 安装
+
+```bash
+cd lobster-press/mcp_server
+pip install -r requirements.txt
+```
+
+## 使用方法
+
+### 1. 与 Claude Desktop 集成
+
+在 Claude Desktop 配置文件中添加：
+
+**macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`  
+**Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
+
+```json
+{
+  "mcpServers": {
+    "lobster-press": {
+      "command": "python3",
+      "args": ["/path/to/lobster-press/mcp_server/lobster_mcp_server.py"]
+    }
+  }
+}
+```
+
+### 2. 与 Cursor 集成
+
+在 Cursor 设置中添加 MCP 服务器配置。
+
+### 3. 测试模式
+
+```bash
+python3 lobster_mcp_server.py --test
+```
+
+## 可用工具
+
+### 1. compress_session
+
+压缩 OpenClaw 会话历史。
+
+```json
+{
+  "session_id": "abc123",
+  "strategy": "medium",
+  "dry_run": false
+}
+```
+
+**参数：**
+- `session_id`: 会话 ID（文件名，不含扩展名）
+- `strategy`: 压缩策略（light/medium/aggressive）
+- `dry_run`: 是否仅预览（默认 false）
+
+**返回：**
+```json
+{
+  "status": "success",
+  "original_messages": 1000,
+  "compressed_messages": 500,
+  "tokens_saved": 50000,
+  "compression_ratio": "50%"
+}
+```
+
+### 2. preview_compression
+
+预览压缩效果。
+
+```json
+{
+  "session_id": "abc123",
+  "strategy": "medium"
+}
+```
+
+### 3. get_compression_stats
+
+获取压缩统计数据。
+
+### 4. update_weights
+
+更新消息类型权重配置。
+
+```json
+{
+  "weights": {
+    "decision": 0.3,
+    "error": 0.25,
+    "config": 0.2,
+    "preference": 0.15
+  }
+}
+```
+
+### 5. list_sessions
+
+列出所有可压缩的会话。
+
+```json
+{
+  "min_tokens": 10000
+}
+```
+
+## 压缩策略
+
+| 策略 | 保留比例 | 适用场景 |
+|------|---------|---------|
+| light | 70% | 保留大部分信息 |
+| medium | 50% | 平衡压缩与保留 |
+| aggressive | 30% | 最大压缩 |
+
+## 消息评分
+
+消息按以下权重评分：
+
+| 类型 | 权重 | 示例 |
+|------|------|------|
+| decision | 0.3 | "决定采用方案A" |
+| error | 0.25 | "发生错误：连接超时" |
+| config | 0.2 | "API Key 已更新" |
+| preference | 0.15 | "用户偏好使用 GLM-4" |
+| context | 0.05 | "当前系统版本 2026.3.2" |
+| chitchat | 0.02 | "哈哈，不错" |
+
+## 资源访问
+
+MCP 客户端可以通过以下 URI 访问会话资源：
+
+```
+lobster://sessions/{session_id}
+```
+
+## 示例
+
+### Python 客户端
+
+```python
+import json
+import subprocess
+
+def call_mcp_tool(tool_name, arguments):
+    request = {
+        "method": "tools/call",
+        "params": {
+            "name": tool_name,
+            "arguments": arguments
+        }
+    }
+    
+    proc = subprocess.Popen(
+        ["python3", "lobster_mcp_server.py"],
+        stdin=subprocess.PIPE,
+        stdout=subprocess.PIPE,
+        text=True
+    )
+    
+    stdout, _ = proc.communicate(json.dumps(request))
+    return json.loads(stdout)
+
+# 列出会话
+result = call_mcp_tool("list_sessions", {"min_tokens": 10000})
+print(result)
+
+# 预览压缩
+result = call_mcp_tool("preview_compression", {
+    "session_id": "abc123",
+    "strategy": "medium"
+})
+print(result)
+
+# 执行压缩
+result = call_mcp_tool("compress_session", {
+    "session_id": "abc123",
+    "strategy": "medium"
+})
+print(result)
+```
+
+## 安全性
+
+- ✅ 自动备份原文件（带时间戳）
+- ✅ 仅本地操作，不上传数据
+- ✅ 支持预览模式（dry_run）
+
+## 许可证
+
+MIT License
+
+## 相关
+
+- Issue #42: v2.0 架构演进：向 MCP Server 演进
+- Issue #28: MCP 协议支持

package/mcp_server/lobster.config.json

@@ -0,0 +1,56 @@
+{
+  "name": "lobster-press",
+  "version": "1.2.8",
+  "description": "LobsterPress - 智能 OpenClaw 会话压缩服务 (MCP Server)",
+  "main": "lobster_mcp_server.py",
+  "type": "mcp-server",
+  "tools": [
+    {
+      "name": "compress_session",
+      "description": "压缩 OpenClaw 会话历史，保留重要信息"
+    },
+    {
+      "name": "preview_compression",
+      "description": "预览压缩效果，显示将要保留和删除的消息"
+    },
+    {
+      "name": "get_compression_stats",
+      "description": "获取压缩统计数据"
+    },
+    {
+      "name": "update_weights",
+      "description": "更新消息类型权重配置"
+    },
+    {
+      "name": "list_sessions",
+      "description": "列出所有可压缩的会话"
+    }
+  ],
+  "resources": [
+    {
+      "uri": "lobster://sessions/{session_id}",
+      "description": "访问 OpenClaw 会话数据"
+    }
+  ],
+  "config": {
+    "sessions_dir": "~/.openclaw/agents/main/sessions",
+    "default_strategy": "medium",
+    "strategies": {
+      "light": {
+        "retention_ratio": 0.7,
+        "description": "轻度压缩，保留 70% 消息"
+      },
+      "medium": {
+        "retention_ratio": 0.5,
+        "description": "中度压缩，保留 50% 消息"
+      },
+      "aggressive": {
+        "retention_ratio": 0.3,
+        "description": "激进压缩，保留 30% 消息"
+      }
+    }
+  },
+  "author": "LobsterPress Team",
+  "repository": "https://github.com/SonicBotMan/lobster-press",
+  "license": "MIT"
+}

package/mcp_server/lobster_mcp_server.py

@@ -0,0 +1,658 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+LobsterPress MCP Server v1.0.0
+基于 Model Context Protocol (MCP) 的压缩服务
+
+Issue: #42 - v2.0 架构演进：向 MCP Server 演进
+Author: LobsterPress Team
+Version: v1.3.0
+"""
+
+import sys
+import json
+import asyncio
+import os
+import re
+from pathlib import Path
+from typing import Dict, List, Optional, Any
+from dataclasses import dataclass, asdict
+from datetime import datetime
+
+
+@dataclass
+class MCPTool:
+    """MCP 工具定义"""
+    name: str
+    description: str
+    input_schema: Dict[str, Any]
+
+
+class LobsterPressMCPServer:
+    """LobsterPress MCP Server"""
+    
+    def __init__(self, sessions_dir: str = None, db_path: str = None, llm_provider: str = None, llm_model: str = None):
+        """初始化 MCP Server
+        
+        Args:
+            sessions_dir: 会话目录
+            db_path: LobsterPress 数据库路径
+            llm_provider: LLM 提供商
+            llm_model: LLM 模型名称
+        """
+        self.sessions_dir = Path(sessions_dir or os.path.expanduser("~/.openclaw/agents/main/sessions"))
+        self.sessions_dir.mkdir(parents=True, exist_ok=True)
+        
+        # v3.2.2: OpenClaw 插件支持
+        self.db_path = db_path or os.path.expanduser("~/.openclaw/lobster.db")
+        self.llm_provider = llm_provider
+        self.llm_model = llm_model
+        self._db = None  # 懒加载数据库连接
+        
+        # 统计数据
+        self.stats = {
+            "total_compressions": 0,
+            "total_messages_processed": 0,
+            "total_tokens_saved": 0,
+            "last_compression": None
+        }
+        
+        # 配置
+        self.config = {
+            "weights": {
+                "decision": 0.3,
+                "error": 0.25,
+                "config": 0.2,
+                "preference": 0.15,
+                "context": 0.05,
+                "chitchat": 0.02,
+                "other": 0.03
+            },
+            "strategy": "medium",
+            "max_tokens": 800000
+        }
+        
+        # 注册工具
+        self.tools = self._register_tools()
+    
+    def _register_tools(self) -> List[MCPTool]:
+        """注册 MCP 工具"""
+        return [
+            MCPTool(
+                name="compress_session",
+                description="压缩 OpenClaw 会话历史，保留重要信息",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "session_id": {
+                            "type": "string",
+                            "description": "会话 ID（文件名，不含扩展名）"
+                        },
+                        "strategy": {
+                            "type": "string",
+                            "enum": ["light", "medium", "aggressive"],
+                            "description": "压缩策略：light（轻度）、medium（中度）、aggressive（激进）",
+                            "default": "medium"
+                        },
+                        "dry_run": {
+                            "type": "boolean",
+                            "description": "是否仅预览，不实际压缩",
+                            "default": False
+                        }
+                    },
+                    "required": ["session_id"]
+                }
+            ),
+            MCPTool(
+                name="preview_compression",
+                description="预览压缩效果，显示将要保留和删除的消息",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "session_id": {
+                            "type": "string",
+                            "description": "会话 ID"
+                        },
+                        "strategy": {
+                            "type": "string",
+                            "enum": ["light", "medium", "aggressive"],
+                            "default": "medium"
+                        }
+                    },
+                    "required": ["session_id"]
+                }
+            ),
+            MCPTool(
+                name="get_compression_stats",
+                description="获取压缩统计数据",
+                input_schema={
+                    "type": "object",
+                    "properties": {},
+                    "required": []
+                }
+            ),
+            MCPTool(
+                name="update_weights",
+                description="更新消息类型权重配置",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "weights": {
+                            "type": "object",
+                            "description": "消息类型权重",
+                            "properties": {
+                                "decision": {"type": "number"},
+                                "error": {"type": "number"},
+                                "config": {"type": "number"},
+                                "preference": {"type": "number"},
+                                "context": {"type": "number"},
+                                "chitchat": {"type": "number"}
+                            }
+                        }
+                    },
+                    "required": ["weights"]
+                }
+            ),
+            MCPTool(
+                name="list_sessions",
+                description="列出所有可压缩的会话",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "min_tokens": {
+                            "type": "integer",
+                            "description": "最小 Token 数阈值",
+                            "default": 10000
+                        }
+                    },
+                    "required": []
+                }
+            ),
+            # v3.2.2: OpenClaw 插件工具
+            MCPTool(
+                name="lobster_grep",
+                description="在 LobsterPress 记忆库中全文搜索历史对话（FTS5 + TF-IDF 重排序）",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "query": {"type": "string", "description": "搜索关键词或短语"},
+                        "conversation_id": {"type": "string", "description": "限定搜索范围的会话 ID（可选）"},
+                        "limit": {"type": "integer", "description": "最多返回条数，默认 5", "default": 5}
+                    },
+                    "required": ["query"]
+                }
+            ),
+            MCPTool(
+                name="lobster_describe",
+                description="查看 LobsterPress 的 DAG 摘要层级结构",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "conversation_id": {"type": "string", "description": "会话 ID（可选，留空查全局）"}
+                    },
+                    "required": []
+                }
+            ),
+            MCPTool(
+                name="lobster_expand",
+                description="将 DAG 摘要节点展开，还原其对应的原始消息（无损检索）",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "summary_id": {"type": "string", "description": "要展开的摘要节点 ID"},
+                        "max_depth": {"type": "integer", "description": "最大展开层数，默认 2", "default": 2}
+                    },
+                    "required": ["summary_id"]
+                }
+            )
+        ]
+    
+    async def handle_request(self, request: Dict[str, Any]) -> Dict[str, Any]:
+        """处理 MCP 请求
+        
+        Args:
+            request: MCP 请求
+        
+        Returns:
+            MCP 响应
+        """
+        method = request.get("method")
+        params = request.get("params", {})
+        
+        try:
+            if method == "tools/list":
+                return {
+                    "tools": [asdict(tool) for tool in self.tools]
+                }
+            
+            elif method == "tools/call":
+                tool_name = params.get("name")
+                arguments = params.get("arguments", {})
+                
+                result = await self._call_tool(tool_name, arguments)
+                return {
+                    "content": [
+                        {
+                            "type": "text",
+                            "text": json.dumps(result, ensure_ascii=False, indent=2)
+                        }
+                    ]
+                }
+            
+            elif method == "resources/list":
+                return await self._list_resources()
+            
+            elif method == "resources/read":
+                return await self._read_resource(params.get("uri"))
+            
+            else:
+                return {"error": f"Unknown method: {method}"}
+        
+        except Exception as e:
+            return {"error": str(e)}
+    
+    async def _call_tool(self, tool_name: str, arguments: Dict[str, Any]) -> Dict[str, Any]:
+        """调用工具"""
+        if tool_name == "compress_session":
+            return await self._compress_session(
+                arguments.get("session_id"),
+                arguments.get("strategy", "medium"),
+                arguments.get("dry_run", False)
+            )
+        
+        elif tool_name == "preview_compression":
+            return await self._preview_compression(
+                arguments.get("session_id"),
+                arguments.get("strategy", "medium")
+            )
+        
+        elif tool_name == "get_compression_stats":
+            return self._get_stats()
+        
+        elif tool_name == "update_weights":
+            return self._update_weights(arguments.get("weights", {}))
+        
+        elif tool_name == "list_sessions":
+            return await self._list_sessions(arguments.get("min_tokens", 10000))
+        
+        # v3.2.2: OpenClaw 插件工具
+        elif tool_name == "lobster_grep":
+            from agent_tools import lobster_grep
+            db = self._get_db()
+            results = lobster_grep(
+                db,
+                arguments["query"],
+                conversation_id=arguments.get("conversation_id"),
+                limit=arguments.get("limit", 5)
+            )
+            return {"results": results}
+        
+        elif tool_name == "lobster_describe":
+            from agent_tools import lobster_describe
+            db = self._get_db()
+            return lobster_describe(db, conversation_id=arguments.get("conversation_id"))
+        
+        elif tool_name == "lobster_expand":
+            from agent_tools import lobster_expand
+            db = self._get_db()
+            return lobster_expand(db, arguments["summary_id"])
+        
+        else:
+            raise ValueError(f"Unknown tool: {tool_name}")
+    
+    def _get_db(self):
+        """获取数据库连接（懒加载）"""
+        if self._db is None:
+            # 添加 src 目录到 path
+            src_dir = Path(__file__).parent.parent / "src"
+            if str(src_dir) not in sys.path:
+                sys.path.insert(0, str(src_dir))
+            
+            from database import LobsterDatabase
+            self._db = LobsterDatabase(self.db_path)
+        return self._db
+    
+    def _validate_session_id(self, session_id: str) -> str:
+        """验证 session_id 防止路径遍历攻击（修复 Issue #12）
+        
+        Args:
+            session_id: 会话 ID
+        
+        Returns:
+            验证后的会话 ID
+        
+        Raises:
+            ValueError: 无效的会话 ID
+        """
+        if not session_id:
+            raise ValueError("Session ID cannot be empty")
+        
+        # 只允许字母、数字、下划线、连字符
+        if not re.match(r'^[a-zA-Z0-9_-]+$', session_id):
+            raise ValueError(f"Invalid session_id: {session_id} (only alphanumeric, underscore, hyphen allowed)")
+        
+        # 防止路径遍历
+        if '..' in session_id or '/' in session_id or '\\' in session_id:
+            raise ValueError(f"Invalid session_id: {session_id} (path traversal detected)")
+        
+        # 长度限制
+        if len(session_id) > 255:
+            raise ValueError(f"Session ID too long: {len(session_id)} > 255")
+        
+        return session_id
+    
+    async def _compress_session(self, session_id: str, strategy: str, dry_run: bool) -> Dict[str, Any]:
+        """压缩会话"""
+        # 验证 session_id（修复 Issue #12）
+        session_id = self._validate_session_id(session_id)
+        session_file = self.sessions_dir / f"{session_id}.jsonl"
+        
+        if not session_file.exists():
+            raise FileNotFoundError(f"Session not found: {session_id}")
+        
+        # 读取会话
+        messages = []
+        with open(session_file, 'r', encoding='utf-8') as f:
+            for line in f:
+                if line.strip():
+                    messages.append(json.loads(line))
+        
+        total_messages = len(messages)
+        
+        # 估算 Token 数
+        estimated_tokens = self._estimate_tokens(messages)
+        
+        # 根据策略决定保留比例
+        retention_ratios = {
+            "light": 0.7,
+            "medium": 0.5,
+            "aggressive": 0.3
+        }
+        retention_ratio = retention_ratios.get(strategy, 0.5)
+        
+        # 计算要保留的消息数
+        keep_count = int(total_messages * retention_ratio)
+        
+        # 评分并排序（简化版）
+        scored_messages = []
+        for i, msg in enumerate(messages):
+            score = self._score_message(msg)
+            scored_messages.append((i, score, msg))
+        
+        # 按分数排序，保留高分消息
+        scored_messages.sort(key=lambda x: x[1], reverse=True)
+        keep_indices = set(x[0] for x in scored_messages[:keep_count])
+        
+        # 按原始顺序保留
+        compressed_messages = [msg for i, msg in enumerate(messages) if i in keep_indices]
+        
+        # 计算压缩后的 Token 数
+        compressed_tokens = self._estimate_tokens(compressed_messages)
+        tokens_saved = estimated_tokens - compressed_tokens
+        
+        if dry_run:
+            return {
+                "status": "preview",
+                "session_id": session_id,
+                "original_messages": total_messages,
+                "compressed_messages": len(compressed_messages),
+                "original_tokens": estimated_tokens,
+                "compressed_tokens": compressed_tokens,
+                "tokens_saved": tokens_saved,
+                "compression_ratio": f"{(tokens_saved / estimated_tokens * 100):.1f}%" if estimated_tokens > 0 else "0%",
+                "strategy": strategy
+            }
+        
+        # 实际写入（备份原文件）
+        backup_file = self.sessions_dir / f"{session_id}.backup.{datetime.now().strftime('%Y%m%d%H%M%S')}"
+        session_file.rename(backup_file)
+        
+        with open(session_file, 'w', encoding='utf-8') as f:
+            for msg in compressed_messages:
+                f.write(json.dumps(msg, ensure_ascii=False) + '\n')
+        
+        # 更新统计
+        self.stats["total_compressions"] += 1
+        self.stats["total_messages_processed"] += total_messages
+        self.stats["total_tokens_saved"] += tokens_saved
+        self.stats["last_compression"] = datetime.now().isoformat()
+        
+        return {
+            "status": "success",
+            "session_id": session_id,
+            "original_messages": total_messages,
+            "compressed_messages": len(compressed_messages),
+            "original_tokens": estimated_tokens,
+            "compressed_tokens": compressed_tokens,
+            "tokens_saved": tokens_saved,
+            "compression_ratio": f"{(tokens_saved / estimated_tokens * 100):.1f}%" if estimated_tokens > 0 else "0%",
+            "strategy": strategy,
+            "backup_file": str(backup_file)
+        }
+    
+    async def _preview_compression(self, session_id: str, strategy: str) -> Dict[str, Any]:
+        """预览压缩效果"""
+        return await self._compress_session(session_id, strategy, dry_run=True)
+    
+    def _get_stats(self) -> Dict[str, Any]:
+        """获取统计数据"""
+        return {
+            "stats": self.stats,
+            "config": self.config
+        }
+    
+    def _update_weights(self, weights: Dict[str, float]) -> Dict[str, Any]:
+        """更新权重配置"""
+        # 验证权重
+        for key, value in weights.items():
+            if key in self.config["weights"]:
+                self.config["weights"][key] = value
+        
+        return {
+            "status": "success",
+            "updated_weights": self.config["weights"]
+        }
+    
+    async def _list_sessions(self, min_tokens: int) -> Dict[str, Any]:
+        """列出会话"""
+        sessions = []
+        
+        for session_file in self.sessions_dir.glob("*.jsonl"):
+            if session_file.name.endswith((".backup.", ".reset.", ".deleted.")):
+                continue
+            
+            # 估算 Token 数
+            messages = []
+            with open(session_file, 'r', encoding='utf-8') as f:
+                for line in f:
+                    if line.strip():
+                        messages.append(json.loads(line))
+            
+            estimated_tokens = self._estimate_tokens(messages)
+            
+            if estimated_tokens >= min_tokens:
+                sessions.append({
+                    "session_id": session_file.stem,
+                    "messages": len(messages),
+                    "estimated_tokens": estimated_tokens,
+                    "file_size": session_file.stat().st_size,
+                    "modified": datetime.fromtimestamp(session_file.stat().st_mtime).isoformat()
+                })
+        
+        # 按 Token 数排序
+        sessions.sort(key=lambda x: x["estimated_tokens"], reverse=True)
+        
+        return {
+            "sessions": sessions,
+            "total": len(sessions)
+        }
+    
+    async def _list_resources(self) -> Dict[str, Any]:
+        """列出资源"""
+        resources = []
+        
+        for session_file in self.sessions_dir.glob("*.jsonl"):
+            if not session_file.name.endswith((".backup.", ".reset.", ".deleted.")):
+                resources.append({
+                    "uri": f"lobster://sessions/{session_file.stem}",
+                    "name": session_file.stem,
+                    "mimeType": "application/jsonl"
+                })
+        
+        return {"resources": resources}
+    
+    async def _read_resource(self, uri: str) -> Dict[str, Any]:
+        """读取资源"""
+        if not uri.startswith("lobster://sessions/"):
+            raise ValueError(f"Invalid resource URI: {uri}")
+        
+        session_id = uri.replace("lobster://sessions/", "")
+        session_file = self.sessions_dir / f"{session_id}.jsonl"
+        
+        if not session_file.exists():
+            raise FileNotFoundError(f"Session not found: {session_id}")
+        
+        with open(session_file, 'r', encoding='utf-8') as f:
+            content = f.read()
+        
+        return {
+            "contents": [
+                {
+                    "uri": uri,
+                    "mimeType": "application/jsonl",
+                    "text": content
+                }
+            ]
+        }
+    
+    def _estimate_tokens(self, messages: List[Dict]) -> int:
+        """估算 Token 数"""
+        total_chars = 0
+        for msg in messages:
+            content = msg.get("content", "")
+            if isinstance(content, str):
+                total_chars += len(content)
+            elif isinstance(content, list):
+                for part in content:
+                    if isinstance(part, dict) and "text" in part:
+                        total_chars += len(part["text"])
+        
+        # 简单估算：3 字符 = 1 token
+        return total_chars // 3
+    
+    def _score_message(self, msg: Dict) -> float:
+        """评分消息重要性"""
+        content = msg.get("content", "")
+        if isinstance(content, list):
+            content = " ".join(p.get("text", "") for p in content if isinstance(p, dict))
+        
+        content_lower = content.lower()
+        score = 0.0
+        
+        # 关键词评分
+        patterns = {
+            "decision": ["决定", "方案", "选择", "决定采用", "decide", "choose"],
+            "error": ["错误", "失败", "异常", "error", "fail", "exception"],
+            "config": ["配置", "设置", "更新", "config", "setting", "update"],
+            "preference": ["偏好", "喜欢", "希望", "prefer", "like", "want"],
+        }
+        
+        for category, keywords in patterns.items():
+            for keyword in keywords:
+                if keyword in content_lower:
+                    score += self.config["weights"].get(category, 0.1)
+        
+        return score
+
+
+def emit(obj: Dict[str, Any]) -> None:
+    """发送 JSON 响应到 stdout"""
+    sys.stdout.write(json.dumps(obj, ensure_ascii=False) + "\n")
+    sys.stdout.flush()
+
+
+async def main():
+    """主入口"""
+    import argparse
+    import time
+    
+    parser = argparse.ArgumentParser(description="LobsterPress MCP Server")
+    parser.add_argument("--sessions-dir", help="会话目录")
+    parser.add_argument("--db", dest="db_path", help="LobsterPress 数据库路径")
+    parser.add_argument("--provider", dest="llm_provider", help="LLM 提供商")
+    parser.add_argument("--model", dest="llm_model", help="LLM 模型名称")
+    parser.add_argument("--test", action="store_true", help="测试模式")
+    
+    args = parser.parse_args()
+    
+    server = LobsterPressMCPServer(
+        sessions_dir=args.sessions_dir,
+        db_path=args.db_path,
+        llm_provider=args.llm_provider,
+        llm_model=args.llm_model
+    )
+    
+    if args.test:
+        # 测试模式
+        print("=== LobsterPress MCP Server 测试 ===")
+        print("\n可用工具:")
+        for tool in server.tools:
+            print(f"  - {tool.name}: {tool.description}")
+        
+        # 测试 list_sessions
+        result = await server._call_tool("list_sessions", {"min_tokens": 1000})
+        print(f"\n找到 {result['total']} 个可压缩会话")
+        for session in result["sessions"][:3]:
+            print(f"  - {session['session_id']}: {session['estimated_tokens']} tokens")
+        
+        print("\n✅ MCP Server 正常工作")
+    else:
+        # Phase 1 (Issue #115): 发送 ready handshake
+        emit({"type": "lobster-press/ready", "ts": time.time()})
+        
+        # MCP 协议模式（读取 stdin）
+        for line in sys.stdin:
+            line = line.strip()
+            if not line:
+                continue
+            
+            request_id = None
+            try:
+                req = json.loads(line)
+                request_id = req.get("requestId") or req.get("id")
+                method = req.get("method")
+                
+                if method == "tools/call":
+                    params = req.get("params", {})
+                    tool_name = params.get("name")
+                    arguments = params.get("arguments", {})
+                    result = await server._call_tool(tool_name, arguments)
+                    emit({
+                        "requestId": request_id,
+                        "status": "ok",
+                        "result": result,
+                    })
+                else:
+                    # 其他方法使用原有处理逻辑
+                    response = await server.handle_request(req)
+                    emit({
+                        "requestId": request_id,
+                        "status": "ok",
+                        "result": response,
+                    })
+            except json.JSONDecodeError as e:
+                emit({
+                    "requestId": request_id,
+                    "status": "error",
+                    "error": f"Invalid JSON: {e}",
+                })
+            except Exception as e:
+                emit({
+                    "requestId": request_id,
+                    "status": "error",
+                    "error": str(e),
+                })
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

package/mcp_server/requirements.txt

@@ -0,0 +1,11 @@
+# LobsterPress MCP Server Dependencies
+
+# 核心依赖
+# Python 3.8+ required
+
+# 可选依赖（用于精确 Token 计数）
+tiktoken>=0.5.0
+
+# 测试依赖
+pytest>=7.0.0
+pytest-asyncio>=0.21.0

package/package.json

@@ -1,12 +1,13 @@
 {
   "name": "@sonicbotman/lobster-press",
-  "version": "3.2.4",
-  "description": "Cognitive Memory System for AI Agents — OpenClaw Plugin",
+  "version": "3.2.6",
+  "description": "Cognitive Memory System for AI Agents \u2014 OpenClaw Plugin",
   "type": "module",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
   "files": [
     "dist/",
+    "mcp_server/",
     "openclaw.plugin.json",
     "README.md"
   ],
@@ -29,7 +30,6 @@
     "openclaw": ">=2026.3.0"
   },
   "devDependencies": {
-    "@sinclair/typebox": "^0.34.48",
     "openclaw": "^2026.3.0",
     "typescript": "^5.4.0"
   },
@@ -42,6 +42,11 @@
   },
   "homepage": "https://github.com/SonicBotMan/lobster-press#readme",
   "openclaw": {
-    "extensions": ["./dist/index.js"]
+    "extensions": [
+      "./dist/index.js"
+    ]
+  },
+  "dependencies": {
+    "@sinclair/typebox": "^0.34.48"
   }
 }