foxclaw 0.1.0 → 0.1.2

package/python/vendor/hmr/storage/memory_fs.py

@@ -0,0 +1,300 @@
+"""
+MemoryFS - 记忆文件系统
+修复内容：
+1. 文件写入加线程锁（防止多代理并发损坏）
+2. AgentWorkspace 持久化（进程重启不丢失）
+3. 原子写入（写临时文件再替换，防止写到一半崩溃）
+4. 启动时重建向量索引的辅助方法
+"""
+
+import json
+import threading
+from pathlib import Path
+from typing import List, Optional, Dict, Any
+from datetime import datetime
+from ..core.models import MemoryObject, RuntimeState, AgentWorkspace
+
+
+class MemoryFS:
+    """
+    记忆文件系统（线程安全版本）
+
+    目录结构：
+    hmr_data/
+    ├── memories/
+    │   ├── concepts/
+    │   ├── projects/
+    │   ├── decisions/
+    │   └── ...
+    ├── runtimes/
+    ├── workspaces/          ← 新增：AgentWorkspace 持久化
+    ├── vector_store/        ← 给 VectorStore 使用
+    └── index/
+    """
+
+    SCHEMA_VERSION = "1.1"
+
+    def __init__(self, base_path: str = "./hmr_data"):
+        self.base_path = Path(base_path)
+        self._lock = threading.Lock()   # 全局写入锁
+        self._init_directories()
+
+    def _init_directories(self):
+        dirs = [
+            self.base_path / "memories" / t
+            for t in ["concepts", "projects", "decisions", "runtimes",
+                       "tasks", "executions", "workflows", "agent_memory", "reflections"]
+        ] + [
+            self.base_path / "runtimes",
+            self.base_path / "workspaces",
+            self.base_path / "vector_store",
+            self.base_path / "index",
+        ]
+        for d in dirs:
+            d.mkdir(parents=True, exist_ok=True)
+
+        # 写入 schema 版本
+        version_file = self.base_path / "schema_version.json"
+        if not version_file.exists():
+            self._atomic_write(version_file, {"version": self.SCHEMA_VERSION})
+
+    # =========================================================================
+    # MemoryObject
+    # =========================================================================
+
+    def write_memory(self, memory: MemoryObject) -> str:
+        type_map = {
+            "concept": "concepts", "project": "projects",
+            "decision": "decisions", "runtime": "runtimes",
+            "task": "tasks", "execution": "executions",
+            "workflow": "workflows", "agent_memory": "agent_memory",
+            "reflection": "reflections"
+        }
+        sub = type_map.get(memory.type, memory.type + "s")
+        type_dir = self.base_path / "memories" / sub
+        type_dir.mkdir(exist_ok=True)
+
+        file_path = type_dir / f"{memory.id}.json"
+        data = memory.model_dump(mode="json")
+
+        with self._lock:
+            self._atomic_write(file_path, data)
+            self._update_memory_index(memory.id, memory.type, file_path)
+
+        return str(file_path)
+
+    def read_memory(self, memory_id: str) -> Optional[MemoryObject]:
+        for type_dir in (self.base_path / "memories").iterdir():
+            if not type_dir.is_dir():
+                continue
+            f = type_dir / f"{memory_id}.json"
+            if f.exists():
+                data = self._safe_read(f)
+                return MemoryObject(**data) if data else None
+        return None
+
+    TYPE_MAP = {
+        "concept": "concepts", "project": "projects",
+        "decision": "decisions", "runtime": "runtimes",
+        "task": "tasks", "execution": "executions",
+        "workflow": "workflows", "agent_memory": "agent_memory",
+        "reflection": "reflections"
+    }
+
+    def list_memories(self, memory_type: Optional[str] = None) -> List[MemoryObject]:
+        memories = []
+        base = self.base_path / "memories"
+        if memory_type:
+            mapped = self.TYPE_MAP.get(memory_type, memory_type + "s")
+            dirs = [base / mapped]
+        else:
+            dirs = list(base.iterdir())
+        for d in dirs:
+            if not d.is_dir():
+                continue
+            for f in d.glob("*.json"):
+                data = self._safe_read(f)
+                if data:
+                    try:
+                        memories.append(MemoryObject(**data))
+                    except Exception:
+                        pass
+        return memories
+
+    def search_memories(self, query: str, memory_type: Optional[str] = None) -> List[MemoryObject]:
+        """简单文本搜索（语义搜索走 VectorStore）"""
+        q = query.lower()
+        return [
+            m for m in self.list_memories(memory_type)
+            if q in m.title.lower() or q in m.content.lower()
+        ]
+
+    def delete_memory(self, memory_id: str) -> bool:
+        for type_dir in (self.base_path / "memories").iterdir():
+            if not type_dir.is_dir():
+                continue
+            f = type_dir / f"{memory_id}.json"
+            if f.exists():
+                with self._lock:
+                    f.unlink()
+                return True
+        return False
+
+    # =========================================================================
+    # RuntimeState
+    # =========================================================================
+
+    def write_runtime_state(self, state: RuntimeState) -> str:
+        file_path = self.base_path / "runtimes" / f"{state.runtime_id}.json"
+        data = state.model_dump(mode="json")
+        with self._lock:
+            self._atomic_write(file_path, data)
+            self._update_runtime_index(state.runtime_id, file_path)
+        return str(file_path)
+
+    def read_runtime_state(self, runtime_id: str) -> Optional[RuntimeState]:
+        f = self.base_path / "runtimes" / f"{runtime_id}.json"
+        data = self._safe_read(f)
+        return RuntimeState(**data) if data else None
+
+    def list_runtime_states(self) -> List[RuntimeState]:
+        states = []
+        for f in (self.base_path / "runtimes").glob("*.json"):
+            data = self._safe_read(f)
+            if data:
+                try:
+                    states.append(RuntimeState(**data))
+                except Exception:
+                    pass
+        return states
+
+    # =========================================================================
+    # AgentWorkspace 持久化（修复：之前没有持久化）
+    # =========================================================================
+
+    def write_workspace(self, workspace: AgentWorkspace) -> str:
+        file_path = self.base_path / "workspaces" / f"{workspace.agent_id}.json"
+        data = workspace.model_dump(mode="json")
+        with self._lock:
+            self._atomic_write(file_path, data)
+        return str(file_path)
+
+    def read_workspace(self, agent_id: str) -> Optional[AgentWorkspace]:
+        f = self.base_path / "workspaces" / f"{agent_id}.json"
+        data = self._safe_read(f)
+        return AgentWorkspace(**data) if data else None
+
+    def list_workspaces(self) -> List[AgentWorkspace]:
+        workspaces = []
+        for f in (self.base_path / "workspaces").glob("*.json"):
+            data = self._safe_read(f)
+            if data:
+                try:
+                    workspaces.append(AgentWorkspace(**data))
+                except Exception:
+                    pass
+        return workspaces
+
+    def delete_workspace(self, agent_id: str) -> bool:
+        f = self.base_path / "workspaces" / f"{agent_id}.json"
+        if f.exists():
+            with self._lock:
+                f.unlink()
+            return True
+        return False
+
+    # =========================================================================
+    # 统计
+    # =========================================================================
+
+    def get_statistics(self) -> Dict[str, Any]:
+        memories = self.list_memories()
+        runtimes = self.list_runtime_states()
+        workspaces = self.list_workspaces()
+
+        type_counts: Dict[str, int] = {}
+        for m in memories:
+            type_counts[m.type] = type_counts.get(m.type, 0) + 1
+
+        total_size = sum(f.stat().st_size for f in self.base_path.rglob("*.json"))
+
+        return {
+            "total_memories": len(memories),
+            "memory_types": type_counts,
+            "total_runtimes": len(runtimes),
+            "total_workspaces": len(workspaces),
+            "disk_usage_bytes": total_size,
+            "disk_usage_mb": round(total_size / (1024 * 1024), 2),
+            "schema_version": self.SCHEMA_VERSION
+        }
+
+    # =========================================================================
+    # 辅助：原子写入 & 安全读取
+    # =========================================================================
+
+    def _atomic_write(self, file_path: Path, data: dict):
+        """写临时文件再原子替换，防止写到一半崩溃导致数据损坏"""
+        tmp = file_path.with_suffix(".tmp")
+        with open(tmp, "w", encoding="utf-8") as f:
+            json.dump(data, f, ensure_ascii=False, indent=2, default=str)
+        tmp.replace(file_path)
+
+    def _safe_read(self, file_path: Path) -> Optional[dict]:
+        """安全读取，文件损坏时返回 None"""
+        if not file_path.exists():
+            return None
+        try:
+            with open(file_path, "r", encoding="utf-8") as f:
+                return json.load(f)
+        except (json.JSONDecodeError, OSError) as e:
+            print(f"[HMR MemoryFS] 文件读取失败 {file_path}: {e}")
+            return None
+
+    def _update_memory_index(self, memory_id: str, memory_type: str, file_path: Path):
+        index_file = self.base_path / "index" / "memory_index.json"
+        index = {}
+        if index_file.exists():
+            index = self._safe_read(index_file) or {}
+        index[memory_id] = {
+            "type": memory_type,
+            "path": str(file_path),
+            "updated_at": datetime.utcnow().isoformat()
+        }
+        self._atomic_write(index_file, index)
+
+    def _update_runtime_index(self, runtime_id: str, file_path: Path):
+        index_file = self.base_path / "index" / "runtime_index.json"
+        index = {}
+        if index_file.exists():
+            index = self._safe_read(index_file) or {}
+        index[runtime_id] = {
+            "path": str(file_path),
+            "updated_at": datetime.utcnow().isoformat()
+        }
+        self._atomic_write(index_file, index)
+
+    def rebuild_index(self):
+        """从文件系统重建索引（索引损坏时使用）"""
+        memory_index = {}
+        runtime_index = {}
+
+        for type_dir in (self.base_path / "memories").iterdir():
+            if not type_dir.is_dir():
+                continue
+            for f in type_dir.glob("*.json"):
+                memory_index[f.stem] = {
+                    "type": type_dir.name,
+                    "path": str(f),
+                    "updated_at": datetime.utcnow().isoformat()
+                }
+
+        for f in (self.base_path / "runtimes").glob("*.json"):
+            runtime_index[f.stem] = {
+                "path": str(f),
+                "updated_at": datetime.utcnow().isoformat()
+            }
+
+        index_dir = self.base_path / "index"
+        self._atomic_write(index_dir / "memory_index.json", memory_index)
+        self._atomic_write(index_dir / "runtime_index.json", runtime_index)
+        print(f"[HMR MemoryFS] 索引重建完成：{len(memory_index)} 条记忆，{len(runtime_index)} 条运行时")