diff --git a/docs/a_memorix_sync.md b/docs/a_memorix_sync.md index f93d669e..312fe9f6 100644 --- a/docs/a_memorix_sync.md +++ b/docs/a_memorix_sync.md @@ -5,7 +5,9 @@ - A_Memorix 主线源码位于 `src/A_memorix` - 宿主接入层位于 `src/services/memory_service.py`、`src/webui/routers/memory.py` 与 dashboard 长期记忆页面 - 运行配置位于 `config/a_memorix.toml` -- 运行数据位于 `data/plugins/a-dawn.a-memorix/` +- 运行数据由 `storage.data_dir` 决定(当前配置模板默认 `data/a-memorix/`) +- 旧离线脚本默认目录仍可能落在 `data/plugins/a-dawn.a-memorix/`(见脚本注释/参数说明) +- Web 上传暂存目录为 `data/memory_upload_staging/` - 上游同步方式固定为 `git subtree` ## 首次接入 @@ -40,6 +42,7 @@ git subtree pull --prefix=src/A_memorix https://github.com/A-Dawn/A_memorix.git ## 同步后检查 -1. 确认 `config/a_memorix.toml` 仍指向 `data/plugins/a-dawn.a-memorix` +1. 确认 `config/a_memorix.toml` 中 `storage.data_dir` 与你的运行目录规划一致(默认模板为 `data/a-memorix`) 2. 运行 `python src/A_memorix/scripts/runtime_self_check.py --help` -3. 运行 A_memorix 相关测试或最少执行一次针对性导入验证 +3. 运行 `python -m pytest pytests/A_memorix_test/test_memory_service.py` +4. 运行 `cd dashboard && npm run test -- src/routes/resource/__tests__/knowledge-base.test.tsx` diff --git a/src/A_memorix/CHANGELOG.md b/src/A_memorix/CHANGELOG.md index 772cff46..f1a20b9d 100644 --- a/src/A_memorix/CHANGELOG.md +++ b/src/A_memorix/CHANGELOG.md @@ -1,5 +1,15 @@ # 更新日志 (Changelog) +## [Unreleased] + +### 📚 文档口径同步 + +- 补充并统一“运行时目录”和“离线脚本默认目录”的说明: + - 运行时主目录由 `storage.data_dir` 决定(当前模板默认 `data/a-memorix`); + - 部分离线脚本仍以 `data/plugins/a-dawn.a-memorix` 作为默认处理目录。 +- 修正文档中的导入示例参数,`memory_import_admin.create_paste` 的 `input_mode` 示例统一为 `text`/`json`。 +- 更新 `README.md` 关于元数据 schema 的描述,和当前代码 `SCHEMA_VERSION = 9` 保持一致。 + ## [2.0.0] - 2026-03-18 本次 `2.0.0` 为架构收敛版本,主线是 **SDK Tool 接口统一**、**管理工具能力补齐**、**元数据 schema 升级到 v8** 与 **文档口径同步到 2.0.0**。 diff --git a/src/A_memorix/CONFIG_REFERENCE.md b/src/A_memorix/CONFIG_REFERENCE.md index d55eb49b..421571c8 100644 --- a/src/A_memorix/CONFIG_REFERENCE.md +++ b/src/A_memorix/CONFIG_REFERENCE.md @@ -17,7 +17,7 @@ enabled = true [storage] -data_dir = "data/plugins/a-dawn.a-memorix" +data_dir = "data/a-memorix" [embedding] model_name = "auto" @@ -126,9 +126,14 @@ default_sample_size = 24 ### `storage` -- `storage.data_dir` (代码默认 `./data`;当前内置配置推荐 `data/plugins/a-dawn.a-memorix`) +- `storage.data_dir` (当前配置模板默认 `data/a-memorix`) : 数据目录。相对路径按 MaiBot 仓库根目录解析。 +补充说明: + +- 部分离线脚本若未显式覆盖路径,会回退到 `A_memorix.paths.default_data_dir()`(当前为 `data/plugins/a-dawn.a-memorix`)。 +- 建议在运维侧统一目录策略,避免“控制台写入目录”和“脚本处理目录”不一致。 + ### `embedding` - `embedding.model_name` (默认 `auto`) diff --git a/src/A_memorix/IMPORT_GUIDE.md b/src/A_memorix/IMPORT_GUIDE.md index 06e2834c..9d248178 100644 --- a/src/A_memorix/IMPORT_GUIDE.md +++ b/src/A_memorix/IMPORT_GUIDE.md @@ -29,6 +29,11 @@ python src/A_memorix/scripts/runtime_self_check.py --json data/plugins/a-dawn.a-memorix/raw/ ``` +说明: + +- `process_knowledge.py` 当前默认扫描上述目录。 +- 若你的运行配置使用 `storage.data_dir = "data/a-memorix"`,请在执行脚本前统一目录,避免脚本导入目录与运行目录不一致。 + 执行: ```bash @@ -52,7 +57,7 @@ python src/A_memorix/scripts/import_lpmm_json.py ## 2.3 LPMM 数据转换 ```bash -python src/A_memorix/scripts/convert_lpmm.py -i -o data/plugins/a-dawn.a-memorix +python src/A_memorix/scripts/convert_lpmm.py -i -o data/a-memorix ``` ## 2.4 历史数据迁移 @@ -115,7 +120,7 @@ python src/A_memorix/scripts/audit_vector_consistency.py --json "arguments": { "action": "create_paste", "content": "今天完成了检索调优回归。", - "input_mode": "plain_text", + "input_mode": "text", "source": "manual:worklog" } } diff --git a/src/A_memorix/MODIFICATION_POLICY.md b/src/A_memorix/MODIFICATION_POLICY.md index 19d39ab8..29031586 100644 --- a/src/A_memorix/MODIFICATION_POLICY.md +++ b/src/A_memorix/MODIFICATION_POLICY.md @@ -27,9 +27,13 @@ - `src/services/memory_service.py` - `src/webui/routers/memory.py` - `dashboard/src/routes/resource/knowledge-base.tsx` +- `dashboard/src/routes/resource/__tests__/knowledge-base.test.tsx` - `dashboard/src/routes/resource/knowledge-graph/` +- `dashboard/src/lib/memory-api.ts` - `config/a_memorix.toml` -- `data/plugins/a-dawn.a-memorix/` +- `data/a-memorix/` +- `data/plugins/a-dawn.a-memorix/`(旧脚本默认路径) +- `data/memory_upload_staging/`(Web 上传暂存) - `pytests/A_memorix_test/` - 同步脚本与同步文档,例如 `scripts/sync_a_memorix_subtree.sh` diff --git a/src/A_memorix/QUICK_START.md b/src/A_memorix/QUICK_START.md index 7ff61c38..b2b940aa 100644 --- a/src/A_memorix/QUICK_START.md +++ b/src/A_memorix/QUICK_START.md @@ -42,7 +42,7 @@ pip install -r requirements.txt --upgrade enabled = true [storage] -data_dir = "data/plugins/a-dawn.a-memorix" +data_dir = "data/a-memorix" [embedding] model_name = "auto" @@ -157,12 +157,14 @@ python src/A_memorix/scripts/runtime_self_check.py --json ### 4.1 文本批量导入 -把文本放到: +`process_knowledge.py` 当前默认扫描目录为: ```text data/plugins/a-dawn.a-memorix/raw/ ``` +若你当前运行目录使用 `storage.data_dir = "data/a-memorix"`,建议先把文本同步到脚本默认目录再执行,避免导入目录与运行目录不一致。 + 执行: ```bash @@ -181,7 +183,7 @@ python src/A_memorix/scripts/process_knowledge.py --chat-log --chat-reference-ti ```bash python src/A_memorix/scripts/import_lpmm_json.py -python src/A_memorix/scripts/convert_lpmm.py -i -o data/plugins/a-dawn.a-memorix +python src/A_memorix/scripts/convert_lpmm.py -i -o data/a-memorix python src/A_memorix/scripts/migrate_chat_history.py --help python src/A_memorix/scripts/migrate_maibot_memory.py --help python src/A_memorix/scripts/migrate_person_memory_points.py --help diff --git a/src/A_memorix/README.md b/src/A_memorix/README.md index 060bfb0d..572d01a7 100644 --- a/src/A_memorix/README.md +++ b/src/A_memorix/README.md @@ -21,7 +21,7 @@ A_Memorix 是 MaiBot 内置的长期记忆子系统。 - 旧 `components/commands/*`、`components/tools/*` 与 `server.py` 已移除。 - 统一入口为宿主侧 host service + [`core/runtime/sdk_memory_kernel.py`](core/runtime/sdk_memory_kernel.py)。 -- 元数据 schema 为 `v8`,新增外部引用与运维操作记录(如 `external_memory_refs`、`memory_v5_operations`、`delete_operations`)。 +- 元数据 schema 为 `v9`,支持外部引用与运维操作记录(如 `external_memory_refs`、`memory_v5_operations`、`delete_operations`)。 如果你还在使用旧版 slash 命令(如 `/query`、`/memory`、`/visualize`),需要按本文的 Tool 接口迁移。 @@ -140,6 +140,7 @@ enabled = true ### 配置方式 - 默认配置文件:`config/a_memorix.toml` +- 运行目录主路径:`storage.data_dir`(当前模板默认 `data/a-memorix`) - 长期记忆控制台:适合修改常用高频项,如 embedding、检索、Episode、人物画像、导入与调优开关。 - 原始 TOML:适合复制整份配置、批量粘贴参数,或编辑未在可视化表单中展示的高级项。 - 配置参考:请结合 [CONFIG_REFERENCE.md](CONFIG_REFERENCE.md) 查看各键的运行时语义与默认值。 @@ -171,7 +172,7 @@ python src/A_memorix/scripts/process_knowledge.py - `web/import.html`(导入中心) - `web/tuning.html`(检索调优) -当前分支不再内置独立 `server.py`,页面路由与 API 暴露由宿主侧 React 页面和 `/api/webui/memory/*` 接口承接。 +当前分支不再内置独立 `server.py`,页面路由与 API 暴露由宿主侧 React 页面和 `/api/webui/memory/*` 接口承接(并保留 `/api/*` 兼容路由)。 ### WebUI 验证脚本 diff --git a/src/A_memorix/core/runtime/lifecycle_orchestrator.py b/src/A_memorix/core/runtime/lifecycle_orchestrator.py index bc49a25f..a421b05a 100644 --- a/src/A_memorix/core/runtime/lifecycle_orchestrator.py +++ b/src/A_memorix/core/runtime/lifecycle_orchestrator.py @@ -4,7 +4,7 @@ from __future__ import annotations import asyncio from pathlib import Path -from typing import Any +from typing import Any, Callable, Coroutine, cast from src.common.logger import get_logger @@ -106,7 +106,8 @@ def start_background_tasks(plugin: Any) -> None: and bool(plugin.get_config("episode.generation_enabled", True)) and (episode_task is None or episode_task.done()) ): - plugin._episode_generation_task = asyncio.create_task(episode_loop()) + episode_loop_fn = cast(Callable[[], Coroutine[Any, Any, Any]], episode_loop) + plugin._episode_generation_task = asyncio.create_task(episode_loop_fn()) async def cancel_background_tasks(plugin: Any) -> None: diff --git a/src/A_memorix/core/runtime/sdk_memory_kernel.py b/src/A_memorix/core/runtime/sdk_memory_kernel.py index 13e55bfa..0da4e02d 100644 --- a/src/A_memorix/core/runtime/sdk_memory_kernel.py +++ b/src/A_memorix/core/runtime/sdk_memory_kernel.py @@ -7,7 +7,7 @@ import time import uuid from dataclasses import dataclass from pathlib import Path -from typing import Any, Awaitable, Callable, Dict, Iterable, List, Optional, Sequence +from typing import Any, Callable, Coroutine, Dict, Iterable, List, Optional, Sequence from src.common.logger import get_logger @@ -83,7 +83,7 @@ class _KernelRuntimeFacade: async def execute_request_with_dedup( self, request_key: str, - executor: Callable[[], Awaitable[Dict[str, Any]]], + executor: Callable[[], Coroutine[Any, Any, Dict[str, Any]]], ) -> tuple[bool, Dict[str, Any]]: return await self._kernel.execute_request_with_dedup(request_key, executor) @@ -769,7 +769,7 @@ class SDKMemoryKernel: async def execute_request_with_dedup( self, request_key: str, - executor: Callable[[], Awaitable[Dict[str, Any]]], + executor: Callable[[], Coroutine[Any, Any, Dict[str, Any]]], ) -> tuple[bool, Dict[str, Any]]: token = str(request_key or "").strip() if not token: @@ -1761,8 +1761,22 @@ class SDKMemoryKernel: profile = manager.get_profile_snapshot() return {"success": True, "profile": profile, "toml": manager.export_toml_snippet(profile)} if act == "apply_profile": - profile = kwargs.get("profile") if isinstance(kwargs.get("profile"), dict) else kwargs - return {"success": True, **await manager.apply_profile(profile, reason=str(kwargs.get("reason", "manual") or "manual"))} + profile_raw = kwargs.get("profile") + if isinstance(profile_raw, dict): + profile_payload: Dict[str, Any] = dict(profile_raw) + else: + profile_payload = { + key: value + for key, value in kwargs.items() + if key not in {"reason", "profile"} + } + return { + "success": True, + **await manager.apply_profile( + profile_payload, + reason=str(kwargs.get("reason", "manual") or "manual"), + ), + } if act == "rollback_profile": return {"success": True, **await manager.rollback_profile()} if act == "export_profile": @@ -1999,7 +2013,11 @@ class SDKMemoryKernel: self._ensure_background_task("memory_maintenance", self._memory_maintenance_loop) self._ensure_background_task("person_profile_refresh", self._person_profile_refresh_loop) - def _ensure_background_task(self, name: str, factory: Callable[[], Awaitable[None]]) -> None: + def _ensure_background_task( + self, + name: str, + factory: Callable[[], Coroutine[Any, Any, None]], + ) -> None: task = self._background_tasks.get(name) if task is not None and not task.done(): return diff --git a/src/A_memorix/core/storage/graph_store.py b/src/A_memorix/core/storage/graph_store.py index 0a5fd95d..88574a12 100644 --- a/src/A_memorix/core/storage/graph_store.py +++ b/src/A_memorix/core/storage/graph_store.py @@ -73,7 +73,7 @@ class GraphStore: def __init__( self, - matrix_format: str = "csr", + matrix_format: Union[str, SparseMatrixFormat] = "csr", data_dir: Optional[Union[str, Path]] = None, ): """ diff --git a/src/A_memorix/core/storage/vector_store.py b/src/A_memorix/core/storage/vector_store.py index 97a9144c..787e625a 100644 --- a/src/A_memorix/core/storage/vector_store.py +++ b/src/A_memorix/core/storage/vector_store.py @@ -79,6 +79,7 @@ class VectorStore: self.quantization_type = QuantizationType.INT8 self.index_type = "sq8" self.buffer_size = buffer_size + self.min_train_threshold = self.DEFAULT_MIN_TRAIN self._index: Optional[faiss.IndexIDMap2] = None self._init_index() diff --git a/src/A_memorix/core/utils/relation_write_service.py b/src/A_memorix/core/utils/relation_write_service.py index 6fa2e621..fbaeb820 100644 --- a/src/A_memorix/core/utils/relation_write_service.py +++ b/src/A_memorix/core/utils/relation_write_service.py @@ -128,7 +128,7 @@ class RelationWriteService: predicate: str, obj: str, confidence: float = 1.0, - source_paragraph: str = "", + source_paragraph: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, *, write_vector: bool = True, diff --git a/src/webui/routers/memory.py b/src/webui/routers/memory.py index 303b2cc4..25bb1d9f 100644 --- a/src/webui/routers/memory.py +++ b/src/webui/routers/memory.py @@ -125,7 +125,8 @@ class DeletePurgeRequest(BaseModel): def _build_import_guide_markdown(settings: dict[str, Any]) -> str: - path_aliases = settings.get("path_aliases") if isinstance(settings.get("path_aliases"), dict) else {} + path_aliases_raw = settings.get("path_aliases") + path_aliases = path_aliases_raw if isinstance(path_aliases_raw, dict) else {} alias_lines = [ f"- `{name}` -> `{path}`" for name, path in sorted(path_aliases.items()) @@ -394,15 +395,7 @@ async def _memory_config_get() -> dict: async def _memory_config_get_raw() -> dict: - raw_payload_getter = getattr(a_memorix_host_service, "get_raw_config_with_meta", None) - if callable(raw_payload_getter): - raw_payload = raw_payload_getter() - else: - raw_payload = { - "config": a_memorix_host_service.get_raw_config(), - "exists": bool(a_memorix_host_service.get_config_path().exists()), - "using_default": False, - } + raw_payload = a_memorix_host_service.get_raw_config_with_meta() return { "success": True, "config": str(raw_payload.get("config", "") or ""), @@ -628,8 +621,10 @@ async def _tuning_apply_best(task_id: str) -> dict: async def _tuning_report(task_id: str, fmt: str) -> dict: - payload = await memory_service.tuning_admin(action="get_report", task_id=task_id, format=fmt) - report = payload.get("report") if isinstance(payload.get("report"), dict) else {} + payload_raw = await memory_service.tuning_admin(action="get_report", task_id=task_id, format=fmt) + payload = payload_raw if isinstance(payload_raw, dict) else {} + report_raw = payload.get("report") + report = report_raw if isinstance(report_raw, dict) else {} return { "success": bool(payload.get("success", False)), "format": report.get("format", fmt),