Commit Graph

15 Commits

Author SHA1 Message Date
310d82f4b8 perf(debug): 添加多智能体模式内存监控埋点
- 新增 modules/memory_debug.py 统一内存监控模块
- 在子智能体任务、主对话 build_messages、多智能体消息注入、
  API 请求准备、深压缩等关键路径记录 RSS/VMS 和消息规模
- 日志输出到 ~/.astrion/astrion/host/logs/memory_debug.log
2026-07-15 17:02:33 +08:00
351cd311c2 fix(multi-agent): 修复压缩404、prompt冻结串扰、子智能体状态串扰及/multiagent/new跳转问题 2026-07-14 23:22:06 +08:00
4fcd0ccc4d feat(multi-agent): 子智能体系统升级
- prompt改造:子智能体注入AGENTS.md/执行环境/工作区信息/skills/项目记忆
- 工具增加:read_skill/recall_project_memory/todo_create/todo_update_task/save_webpage
- 上下文压缩:深度压缩机制,记录current_context_tokens,默认150k阈值可配置
- 模型升级:sub_agent_models.json支持thinkmode_status/extra_parameter,与主智能体对齐
- 角色管理三层结构:源码树预设multi_agent_roles/ + 运行态host/web预设 + web按用户隔离
- 启动同步:initialize_system调用sync_preset_roles同步预设到host和web运行态
- 模式判断:API用session.host_mode,工具用data_dir路径判断,不依赖IS_HOST_MODE
- 前端:个人空间新增子智能体管理页(角色CRUD/压缩阈值/模型选择),复用个人空间样式
- 入口改造:登录页移除多智能体按钮,QuickMenu加模式切换项,运行中对话禁止切换
- 工具调整:多智能体模式create_sub_agent去掉timeout/deliverables_dir参数
- skill禁止:sub-agent-guide在多智能体模式下禁止阅读
- .agents/统一为.astrion/路径修复
2026-07-14 02:28:45 +08:00
72a49a7c8d fix: 修复对话计时器持久化、子智能体工具美化与状态查询,统一后台通知池
- 新增 server/work_timer.py,在对话真正空闲(无前台任务、无后台子智能体/后台命令/压缩)时持久化 work_timer,并同步内存副本,解决刷新后计时器回退问题。
- server/chat_flow_task_main.py / chat_flow.py / tasks/models.py 在任务结束/取消时按空闲判定决定是否持久化。
- 前端 lifecycle.ts 在仍有后台任务时不提前停止计时器。
- 子智能体工具(create/terminate/get_status)渲染美化,task 参数放在顶部元信息区。
- 子智能体状态查询支持返回「已完成」「已终止」「不存在」;修复 wait_for_completion 在 final_result 就绪前返回导致的「ID 被占用」误报。
- 统一后台完成通知池 poll_completion_notifications,合并子智能体与后台 run_command 两路轮询,避免逐条触发工作循环与单工作区互斥冲突。
- 删除本次新增的各类 debug_log / notify_pool_log 调用及辅助脚本。
2026-06-25 04:16:44 +08:00
f75e2f07a3 refactor(deep_compression): 改造深度压缩消息结构、提示词加载与前端展示
- 删除 compact 文件和 inject guide 中的最近工具记录
- 新结构:用户所有输入(按压缩轮次分段)+ 历次压缩总结 + 最近一次输入
- 用户输入分段标记改为 <第N次压缩> 和 <当前触发的第X次压缩>
- 压缩总结提示词迁移到 prompts/deep_compression_summary.txt
- deep_compression_records 增加 user_inputs_before 和 summary 字段
- 压缩后清空全部 frozen prompt 缓存
- 修复 wait 模式 in-place 压缩前端不刷新问题
- 更新手动压缩确认弹窗文案
- 修复 _apply_workspace_personalization_preferences 测试 mock 签名
- 清理 context.py 中误导性的主提示词构建参数
2026-06-18 14:50:16 +08:00
ee9ac28f26 feat(compression): 对话压缩提示词新增当前任务进度与下一步行动条目 2026-06-10 19:24:18 +08:00
820ec69354 feat(memory): 新增项目记忆系统,整合总体+项目记忆注入
- 新建 .agents/memory/ 项目记忆目录,支持 YAML frontmatter + Markdown
- 新增 update_project_memory / recall_project_memory 工具(复用 write_file/read_file)
- 新建 memory_system.txt prompt,总体长期记忆 + 项目记忆统一注入
- 记忆从 workspace_system.txt 中移出,独立为 memory_prompt 注入
- 重排 system prompt 注入顺序:主→权限→环境→对话→个性化→工作区→AGENTS.md→skills→记忆→禁用
- 压缩后选择性 invalidate 4 个 frozen keys(skills/workspace/personalization/memory)
- 前端:notebook/notebook-pen 图标、动画、状态文案、展开内容渲染
- 优化三个记忆工具的 description 和 memory_system prompt,引导模型主动调用
2026-06-10 15:14:01 +08:00
3fd2214da0 feat(compression): 深压缩改为 in-place 标记前缀,不再切换新对话
深压缩从"建新对话+切过去"重构为标记当前对话历史前缀(deep_compacted):
原文保留并照常显示,仅在 build_messages 构建请求时排除。conversation_id
不变,避免任务状态/目标模式/前端对话切换的大量适配带来的 bug。

- deep_compression.py: _mark_history_compacted 打标 + 重置 current_context_tokens
  防自动续接死循环;总结请求传入正常 tools 以 100% 命中前缀缓存
- context.py build_messages: 跳过 deep_compacted 消息
- conversation.py /compress: 去掉切对话,按 compress_behavior 决定续接/等待
- 新增个性化设置 deep_compress_form(file/inject) 与
  deep_compress_behavior(continue/wait,仅手动压缩生效)
- 前端去掉强制切换对话,改为重载当前对话刷新展示
- AGENTS.md/CLAUDE.md: 补充默认中文交流约定等说明

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-30 20:56:15 +08:00
8a1e17b9a0 feat(goal): add autonomous goal mode workflow
Introduce workspace-level goal state persistence, goal prompt injection, and after-turn review handling so an active task can continue until the configured completion conditions are met.

Add a dedicated goal review agent with readonly and active evidence modes, configurable model settings, review prompt, token/turn boundaries, idle-no-tool protection, and progress/completed/stopped events.

Wire goal_mode through task creation, task restoration, compression handoff, runtime user messages, API message sanitization, and tool-call ordering so goal continuations survive long-running tasks and deep compression.

Add Vue UI for arming goal mode from the quick menu, showing running/completed banners, displaying progress metrics, restoring running goal state, and exposing personalization settings for review mode and stop limits.

Include goal mode research notes and default goal review configuration.
2026-05-30 12:51:42 +08:00
398ee6e760 feat(memory): refactor proactive memory and conversation recall
Refactor conversation persistence into workspace-scoped storage with host workspace legacy migration, current workspace filtering, and active host workspace state protection.

Add proactive recent-conversation prompt injection with personal-space controls, configurable item count, empty-conversation filtering, and frozen per-conversation prompt metadata to keep request caches stable.

Introduce conversation_search and conversation_review tools for current-workspace history: multi-keyword search, list mode, current conversation exclusion, message/tool counts, formatted model output, read/save review modes, long-review fallback, and .agents/review storage.

Move internal workspace artifacts under .agents, including uploads, skills, compact results, and reviews; update prompt file-tree visibility, compression output, upload paths, and hidden versioning exclusions.

Add create_skill workflow and a Skills tool category, move read_skill into Skills, validate/归档 skill folders, support host global skills and Docker/API private skills, and sync private skills into .agents/skills and skill prompts.

Update frontend tool icons, statuses, and expanded renderers for conversation recall and skills workflows; add workspace conversation and skills-manager tests.
2026-05-27 03:09:42 +08:00
8395cf8b4b fix(prompt): merge disabled-tool notice into leading system prompts 2026-05-09 18:56:49 +08:00
eba36c46da fix(versioning): 压缩继承版本控制并忽略系统目录 2026-04-11 17:02:39 +08:00
a21df8f7fb feat: persist deep compression history and include it in guide messages 2026-04-07 11:50:54 +08:00
b29e6a67f8 fix: keep compression toast persistent and align summary layout 2026-04-06 17:09:36 +08:00
7e03e3cd5f feat: stabilize context compression and deep-compact flow 2026-04-06 15:12:25 +08:00