mirror of
https://github.com/agentscope-ai/ReMe.git
synced 2026-09-07 08:26:06 +00:00
6 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
jinliyl
|
83831ec90c
|
feat(core): enhance reme4 (#281)
### 1. Agent Wrapper（统一 Agent 后端抽象） - **`base_agent_wrapper.py`**：`reply()` 返回值从 `tuple[str, Any]` 改为 `dict`（含 `session_id` / `last_message` / `result` / 可选 `structured_output`）；`reply_stream()` 改为产出统一的 `StreamChunk`。废弃 `add_tools()`，改为 `add_job_tools(names: list[str])`（按名解析 BaseJob）与 `add_skills()`；新增 `_resolve_job_tools()`、`_merged_kwargs()`、`_chunk()` 辅助方法及 `project_path` / `project_skills_root` 属性。 - **`as_agent_wrapper.py`（AgentScope 后端）**： - 会话持久化重写：`session_path` 落地到 `<vault>/<session_dir>/agentscope/`，`_load_state` 支持 `resume` / `session_id` / `fork_session`，并做 UUID 校验（`_validate_session_id`）；`_cleanup_expired_sessions` 按天数清理过期会话。 - 新增内置工具集（`BypassAnalysisBash` + Edit/Glob/Grep/Read/Write），`BypassAnalysisBash` 绕过 AgentScope 自带 Bash 静态分析以让 permission_mode 生效；`_resolve_skills()` 把配置的 skill 暴露给后端，`_load_tool_env()` 注入项目 `.env`。 - `_event_to_chunk()` 把 20+ 种 AgentScope 事件（Reply/Text/Thinking/Data/ToolCall/ToolResult/ModelCall/ExceedMaxIters）归一化为 `StreamChunk`。 - **`cc_agent_wrapper.py`（Claude Code SDK 后端，+551 行）**： - 新增 `_CcFileSessionStore`：基于 vault 的文件型会话存储，实现 append（按 uuid 去重）/ load / list / delete / list_subkeys，并对路径做 `_safe_parts` + `resolve()` 防越界校验。 - `_build_options()`：统一构建 `ClaudeAgentOptions`，处理 skills、disallowed_tools（默认禁 `WebSearch`）、`.env` 注入、Claude Code 的 API 凭据解析（`_claude_code_api_env`，多级 base_url/api_key 回退）、`CLAUDE_CONFIG_DIR` 设置、skill 目录软链接（`_ensure_claude_skill_dir`）。 - `_raw_event_to_chunk()` / `_message_content_to_chunks()`：把 Anthropic 流式事件（message_start/delta/stop、content_block_*）与 SDK 消息块（AssistantMessage/UserMessage/ResultMessage/RateLimitEvent）转换为统一 `StreamChunk`；跟踪 block_id/block_type/tool_call_name 做关联；处理尾部 `"success"` 误报异常的吞掉逻辑。 ### 2. 统一流式协议（StreamChunk / ChunkEnum） - **`stream_chunk.py`**：`StreamChunk` 扩展为承载 AS + CC 双后端完整信息的统一结构，新增 `session_id` / `block_id` / `tool_call_id` / `tool_call_name` / `media_type` / `input_tokens` / `output_tokens` 等字段，纯文本流仍保持轻量。 - **`chunk_enum.py`**：补全生命周期标记 `REPLY_START` / `REPLY_END`，并文档化两套后端事件 → ChunkEnum 的映射。 ### 3. Index 模块重构（变化批次化 + dispatch） - 新增 `_change_batch.py`：`coalesce_changes()` 把同路径多次事件折叠为最终状态（结合 path 存在性判定），`bucket_changes()` 按 watchfiles.Change 分桶。 - 新增 `init_changes.py`（`InitChangesStep`）：一次性扫描，对比 file_store / file_catalog 已索引节点计算 added/modified/deleted，写入 `context["changes"]` 后 dispatch。 - 新增 `update_changes.py`：抽象基类 `ChangeApplyStep` 统一 added/modified/deleted 处理与错误收集；`UpdateCatalogStep`（写 file_catalog）、`UpdateIndexStep`（写 file_store，含按后缀解析 chunker）。 - **`watch_changes.py`**：改用 `dispatch_step_specs`（基类提供的 `dispatch_steps()`），每批先 `coalesce_changes` 再 dispatch；默认参数调整（debounce 5000ms / step 1000ms / poll 5000ms）并暴露常量。 - 删除旧步骤：`clear_and_scan` / `foreach_dispatch` / `scan_changes` / `update_catalog`(旧) / `update_index`(旧)；`clear_store.py` 取代 clear_and_scan。 ### 4. Evolve / Dream 模块（拆分为多步 pipeline） - 删除旧的单体 `auto_dream.py` / `dream.py` / `dream.yaml`，新增 `dream/` 子包，按 5 个步骤组织： - **`extract.py`**：扫描当日 day-index + daily 笔记，对比 file_catalog 找出 changed/deleted，调用 LLM 全局抽取 `units`（procedure/personal/wiki 三桶）与 `topics`，路径与桶做清洗/路由。 - **`integrate.py`**：逐个 unit 调用 LLM 写入 digest，结构化输出 `IntegrateOutcome`（CREATE/CORROBORATE/REFINE/CORRECT），失败 unit/路径收集回写。 - **`topics.py`**：写 `daily/<date>/interests.yaml`，结合当天已有 + 近 N 天做去重（`normalize_topic`），可走 LLM 或纯规则去重两条路径。 - **`proactive.py`**：读取当日 `interests.yaml`，作为主动推荐话题的入口。 - **`finish.py`**：把变更路径落盘到 dream file_catalog（checkpoint），渲染最终汇总摘要。 - 新增 `schema.py`（`DreamState` 等跨步骤共享状态与结构化输出模型）与 `utils.py`（状态存取、扫描打包、YAML 读写、结构化回复解析等公共函数）。 - `evolve/__init__.py` 导出全部新 step。 ### 5. auto_memory / auto_resource（适配新 Agent API） - **`auto_memory.py`**：会话路径迁移到 `<session_dir>/dialog/<session_id>.jsonl`；改用 `job_tools`；新增 `source_conversation` frontmatter 反向链接（`_session_link`）；执行后刷新 day 索引（`refresh_day_index`），并对 session_id 做合法性校验。 - **`auto_resource.py`**：资源改用「同名 daily note」方案（`_compute_note_stem` 取文件 stem）；批量处理 `changes: list[dict]`（`_handle_change` 逐项处理，返回逐项结果摘要）；agent 会话 id 用稳定的 `uuid5`；同样刷新 day 索引。 ### 6. BaseStep 基类增强 - 新增 `dispatch_steps` / `dispatch_step_specs` 机制：`_resolve_dispatch_step()` 支持字符串或 dict 形式的 step spec，`dispatch_steps()` 复用当前 context 调用下游 step。 - 新增 `config_value()`：按 key 取 app config，缺失时回退 `ApplicationConfig` 默认值。 - 小幅清理：`language` 初始化、`copy()`、`Ref.__init__` 签名精简。 ### 7. Components 改动 - **`file_store/local_file_store.py`**：持久化改用 zstd 压缩（`.jsonl.zst`，通过新 `utils/jsonl_zst.py`）；upsert 时先删除旧 chunk 的 keyword 文档；embedding 复用改为 `(text, embedding)` 键控，要求文本一致才复用；新增 `_matches_search_filter()` 对 vector/keyword 搜索做 path/path_prefix/metadata 的统一后过滤。 - **`keyword_index/bm25_index.py`**：索引文件名加入组件名 + tokenizer 指纹（sha256 前 12 位），快照/恢复时校验指纹防配置漂移；空索引 dump 时删除文件，加载失败抛错而非静默。 - **`file_chunker/markdown_file_chunker.py`**：弃用 `python-frontmatter`，改用内置 YAML 解析（非法 YAML 不阻断正文索引），并修正因 frontmatter 占用行号导致的 AST 行号偏移（`line_offset`）。 - **`cron_job.py`**：大幅简化（-187 行），由原来「dispatch 外部 job/step + 多种调度模式」改为「在自身 steps 上跑 cron 表达式」；`Application` 启动顺序随之调整为 base > stream > background > cron。 - 其余小调整：service（base/http/mcp）、file_graph、file_catalog、as_llm、as_embedding、tokenizer、prompt_handler、base_component 的签名/接口微调。 ### 8. Application 生命周期 - `_start()` 启动顺序明确为 components → base → stream → background → cron，启动失败会触发 `_close()` 回滚并 re-raise（不再吞异常）。 - 启动时创建 `session_dir` 目录；新增 `update_component()`（按类型/名就地更新已存在组件，不存在则报错）。 ### 9. File IO / 路径安全 - **`_path.py`**：`resolve_path` 增加 vault 越界防护（`is_relative_to` 校验），禁止 `.` / `..` 路径分量，支持 `allow_empty`。 - **`read.py`**：大文件（超过 `MAX_FILE_READ_BYTES`）走按行读取 `read_file_lines_safe`，避免一次性载入内存。 - **`_file_io.py` / `_daily_index.py` / `_path.py`** 等支持函数补齐（如 `refresh_day_index`、`read_file_lines_safe`）。 - **`env_utils.py`**：新增 `parse_env_file()`，`load_env()` 返回加载到的键值、支持 `override`、对无路径调用做幂等缓存。 ### 10. Config - `ApplicationConfig` 新增 `session_dir`（默认 `reme_session`）。 - `config_parser.py`：环境变量展开后做类型转换（`_convert_value`）、dot-notation 与 key=value 参数校验更严格、配置文件路径支持相对 `_CONFIG_DIR` 查找、根非 dict 报错。 - `default.yaml`：作业编排改用 `init_changes_step` + `dispatch_steps`（index/resource/digest 三个 watch loop 与 reindex）；新增 `auto_dream`（4 步）、`proactive` 作业，移除旧 `dream`；file_catalog 增配 `resource` / `digest` / `dream` 实例；LLM 默认值与 Claude Code 凭据配置调整（tool_result_limit 50000、thinking_enable=false 等）。 ### 11. 其它 - 新增 `steps/common/add.py`（`AddStep` 算术 demo）、`channel/__init__.py` 与 common `__init__` 导出整理。 - 新增 4 篇文档：`docs4/auto_dream_logic_and_step_refactor.md`、`docs4/watch_loop_step_refactor_plan.md`、`docs4/todo.md`，以及 `reme_design.md` 更新。 ** |
||
|
jinliyl
|
041f957a7f
|
refactor(components) components and file I/O, fix method calls and validation (#268)
* refactor(components): extract shared component state into mixin - Introduce ComponentMixin class with shared state for components and steps - Move identity, config, and vault path functionality to ComponentMixin - Update BaseComponent to inherit from ComponentMixin - Update BaseStep to inherit from ComponentMixin - Consolidate vault path helper methods in ComponentMixin - Remove duplicate vault path implementations from BaseComponent and BaseStep - Add ComponentMixin to components module exports * refactor(file_io): implement path locks cache eviction mechanism - Add _PATH_LOCKS_MAX constant set to 1024 for cache size limit - Implement cache eviction logic when locks exceed maximum capacity - Remove half of unlocked entries when cache limit is reached - Use list comprehension to identify unlocked locks for removal - Maintain existing path normalization and locking behavior fix(edit): correct method call from public to private fail method - Change self.fail to self._fail for internal error handling - Maintain consistent private method usage within class fix(mcp_client): change pop to get for optional command and args - Replace kwargs.pop with kwargs.get to avoid removing keys - Preserve original kwargs dictionary contents - Maintain default empty string and list values feat(reme): add client backend validation with error raising - Check if client_cls is None before instantiation - Raise ValueError with descriptive message for unknown backends - Provide clear error feedback for invalid backend configurations * fix(components): move directory creation to start method - Moved component_metadata_path.mkdir call from __init__ to _start in base_keyword_index - Moved component_metadata_path.mkdir call from __init__ to _start in local_file_graph - Moved component_metadata_path.mkdir call from __init__ to _start in local_file_store - Ensures directory creation happens after component initialization - Prevents potential issues with path creation during object construction * fix(steps): replace assertions with runtime errors for app_context validation - Replace assert statements with explicit RuntimeError exceptions when app_context is None - Add descriptive error messages for better debugging when resolving components - Replace assert in resolve_component method with proper exception handling - Replace assert in get_file_parser method with proper exception handling - Maintain same functionality while improving error reporting clarity * refactor(file_io): split file IO utilities into modular components - Move daily note helpers to separate _daily_index module - Extract path validation and resolution to new _path module - Remove unused code and imports from _file_io module - Update import statements across affected modules - Introduce WikilinkHandler utility for link parsing - Replace regex-based link extraction with WikilinkHandler - Add integration JSONL files to gitignore - Consolidate file locking mechanism in _file_io module * style(formatter): fix spacing issues in file IO and chunked file parser - Fixed whitespace around colon in slice notation in file_io.py - Corrected spacing around colon in slice notation in chunked_file_parser.py - Applied consistent formatting for array slicing operations - Improved code readability by standardizing space placement in ranges * refactor(steps): replace property-based component resolution with Ref descriptor - Introduce Ref descriptor class for lazy component dependency resolution - Replace _resolve method and individual properties with Ref descriptors - Add as_llm, as_llm_formatter, as_token_counter, file_store, and embedding Ref attributes - Remove legacy property methods and resolve logic from BaseStep - Add cache clearing mechanism for Ref values during step calls - Update UpdateCatalogStep to use Ref instead of property-based resolution |
||
|
jinliyl
|
a4efc0f776
|
refactor(reme4): restructure steps packages (#258)
* fix(bm25_index): 修正BM25索引计算中的文档长度归一化问题 修复了在计算BM25相似度时对文档长度进行不正确归一化的bug，确保所有查询都能得到准确的相关性评分。 * up * up * up * up * up * up * up * up * up * up * up * up * up * up * up * up * up * refactor(steps): Rename and adjust indexing step logic - Rename `scan_changes.py` and `reindex.py` to `clear_and_scan.py` - Update implementation details of `ScanChangesStep` and `ClearAndScanStep` - Modify the scheduling mechanism in `WatchChangesStep` - Adjust step registration and parameter configuration in config files - Update related tests to align with the new interface changes * up * feat(daily): replace daily CRUD operations with slug provisioning approach * refactor(tests): migrate CRUD step tests from HTTP server to direct LocalFileStore * up * up * up * up --------- Co-authored-by: huangsen <huangsen.huang@alibaba-inc.com> |
||
|
Sen Huang
|
7d0bec60be
|
feat: rename working_dir to vault_dir and update documentation (#254)
* feat: rename working_dir to vault_dir and update documentation - Rename working_dir to vault_dir across the application - Update documentation to reflect vault_dir instead of working_dir - Change FileFrontMatter title field to name field - Update .gitignore to include vault directory - Modify file path descriptions to reference vault instead of working_dir - Update related configuration and property names accordingly * refactor(steps): rename working_path to vault_path in CRUD operations - Rename parameter from `working_path` to `vault_path` in `resolve_path` function - Update all usages in append, edit, read, and write steps to use `self.vault_path` - Update documentation comments to reflect the new parameter name - Update docstring in read.py to mention `vault_dir` instead of `vault` test(chunked_file_parser): update frontmatter field from title to name - Change frontmatter field from `title` to `name` in test cases - Update comment in background steps test to reference `vault_path` instead of `working_path` * refactor(schema): remove unused ResourceEntry import * feat(file_graph): add link scope filtering to get_inlinks/get_outlinks * feat(file-store): add scope parameter to link methods |
||
|
jinliyl
|
71e42dbad0
|
refactor(reme4): replace file_watcher component with background steps pipeline (#251)
* up * up * up * up * up * up * up * up * up * up * up * feat(config): add daily_dir configuration and background job logging - Added daily_dir setting with default value 'memory' to config - Implemented logging for background job startup events - Enhanced component start logic to handle background backend type - Updated default YAML configuration structure * refactor(file_parser): replace _get_relative_path with to_vault_relative method - Remove redundant working_dir property from base file parser - Add to_vault_relative method to base component for path resolution - Update bare_file_parser to use new to_vault_relative method - Update default_file_parser to use new to_vault_relative method - Update linked_file_parser to use new to_vault_relative method - Make working_path absolute in base_component and steps - Simplify index_changes step by removing redundant base variable - Consolidate path relative logic in single shared method * docs(reme4): update report with detailed architecture sections - Add comprehensive Markdown kernel section covering Obsidian compatibility - Include detailed explanation of YAML front matter and wikilink formats - Document smart slicing mechanism using Markdown AST instead of fixed tokens - Explain graph indexing with bidirectional links and multiple backends - Restructure sections with proper numbering from 4 to 7 - Move Markdown kernel section to appear before self-evolution features - Add detailed explanations of auto-memory, auto-dream, and auto-link processes - Document three-way hybrid search with RRF fusion and progressive expansion - Include engineering value explanations for keyword indexing in Chinese context |
||
|
jinliyl
|
e411eeb4c0
|
dev/reme4 init merge (#236) |