91 Commits
Author SHA1 Message Date
gozeros 66a443e915 sync update from LAN repo (source master @ c21a2a4a) 2026-11-10 14:03:50 +08:00
gozeros ea9dcff9c0 sync update from LAN repo (source master @ ca48c4ec) 2026-11-05 10:53:54 +08:00
gozeros 8bc8b0e082 sync: 鍚屾灞€鍩熺綉鏈€鏂颁唬鐮?(source master @ b69568e4) 2026-10-21 14:51:31 +08:00
董宏凯 f7e6ed65bd fix: retry httpx timeout family as transient LLM errors
Live incident: the compatible-mode provider stalled mid-stream after
200 OK (~17s without data) and raised httpx.ReadTimeout, which was not
in the retriable exception-name set, so the call failed after 1
attempt and killed the whole run. Add ReadTimeout/ConnectTimeout/
WriteTimeout/PoolTimeout/TimeoutError/ConnectError to the transient
set in LLMErrorHandlingMiddleware (retry with backoff, then a
transient fallback message instead of generic), and add ReadTimeout/
ConnectTimeout to ai_writing's classify_exception for model
switching.
2026-08-30 15:53:51 +08:00
董宏凯 4964967295 fix: writing mode first-report allows only the pipeline to write the report
- Strip write_file/str_replace/bash from the model binding in the
  first-report run: after_model rewrites only fire after a full model
  response, so any bound write tool let weak models draft the whole
  essay into tool args (write_file content or bash cat>/python
  heredocs) for tens of invisible seconds before interception
- Hallucinated hand-writes are still handled: after_model rewrites
  bash cat >/python open('w')/tee and leftover write calls to
  deep_research_report; wrap_tool_call rejects whatever reaches the
  tool layer; post-success writes are dropped, nudged to a 2-3
  sentence summary, and patched with a fallback closing message so
  the turn never ends silently
- The rewritten pipeline call carries topic = original request
  (clarification answers skipped) plus focus = clarification answers,
  so the pipeline writes the clarified subject instead of the raw
  prompt text; prompts updated to match (tools unavailable, call the
  tool directly, do not draft first)
- Frontend write-step interception and streaming preview scoping now
  applies only in writing mode; normal mode shows/previews writes
  exactly as before
2026-08-30 15:53:43 +08:00
董宏凯andCursor 49c01cb72b fix: writing mode uses chat search then pipeline write only
Keep deep_research_report as a write-only deliverable from conversation materials, block post-success rewrite loops with a summary nudge, fill thin/empty pools from model knowledge, and open the sandbox only on real report.md content.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-30 12:41:04 +08:00
董宏凯 babc5e95c6 代码提交 2026-08-24 20:47:31 +08:00
董宏凯andCursor 541bc3280f fix: 岗位会商本轮只允许 present_files 一次,避免重复产物卡
弱模型写完报告后仍会再调用一次 present_files,前端会画出两张相同卡片。上限同时覆盖呈现,已呈现则只催收尾。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-23 21:26:57 +08:00
董宏凯andCursor b81d200eca fix: 协助卡停轮与提交防 409,隐藏配置向导会话,并限制岗位会商产物重复写入
澄清中间件改为 ToolMessage + after_model 停轮,前端提交前取消残留 run;系统/向导 thread_type 从对话列表排除;岗位总结席位加 write_file 上限。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-23 20:29:59 +08:00
董宏凯andCursor 51afebb394 feat: 报告结构支持检索方向/来源与智能新增,并修复协助卡停轮与宿主跳转
为深度研究报告结构增加检索方向与技能来源、对话式智能新增页与详情排版;澄清中间件在 after_model 可靠结束本轮,避免弱模型出卡后继续生成;任务按钮补充 host 跳转与相关深链能力。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-22 17:46:06 +08:00
董宏凯 33b0af333c Merge branch 'master-nj' into 'master'
增强 Deep Research 可靠性,并接入 8BF 任务卡片与会商上下文。

See merge request knowledge-base-bj/deerflow-code!25
2026-08-18 09:55:17 +00:00
董宏凯 819e537aa3 Merge branch 'master' of http://192.168.0.200/knowledge-base-bj/deerflow-code into master-nj 2026-08-17 19:30:37 +08:00
吉鑫升 68c4896088 fix: stabilize auth sync and stream recovery for ji 2026-08-17 19:27:39 +08:00
董宏凯andCursor 771957429a 增强 Deep Research 可靠性,并接入 8BF 任务卡片与会商上下文。
修复 MySQL 读写分离下的持久化与 SSE 重连问题,报告完成前先落库;8BF 嵌入页支持 taskId 任务卡与独立意图 metadata;前端统一 createClientId 并更新 runtime 配置。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-17 18:16:05 +08:00
吉鑫升 94cf37a2ae fix: normalize system messages before model calls 2026-08-17 10:12:43 +08:00
吉鑫升 7e7d3c9846 fix: truncate rwfx 3q context 2026-07-02 11:05:12 +08:00
吉鑫升 36a3851233 feat: inject 3q context into rwfx chat 2026-07-02 10:03:34 +08:00
Administrator 2d3033434b Fix async thread title generation flow 2026-06-28 18:09:48 +08:00
AdministratorandClaude Opus 4.8 acc29c8121 fix(middlewares): 标题模型推理不再泄漏到 UI 流 + open/ephemeral 会话避免无谓 DB 访问
- title: 标题模型用隔离 config(不继承父级 streaming callbacks),force_disable_thinking,<think> 块清洗,避免推理 token 进入实时流/运行日志
- uploads: 无文件的普通对话轮次直接跳过,不再解析线程归属(避免命中 DB)
- thread_data: ephemeral 会话用当前登录用户解析数据目录
- thread_paths: open-/__skill-review__ 等合成线程视为 local-only,直接用当前用户
- frontend: 新增 build:test 构建脚本

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-24 21:20:12 +08:00
吉鑫升andClaude Opus 4.8 6aff165d09 fix(db): 路由跨 loop 的 DB 访问回引擎主 loop,消除 asyncmy "attached to a different loop"
会商/定时任务/沙箱工具等在 worker 线程的临时 loop 上用 asyncio.run 跑全局引擎连接池的
DB 操作,连接被绑定到随即关闭的临时 loop;池子回收/释放该连接时 asyncmy
_terminate_graceful_close 在错误的 loop 上触发 future,持续刷
"got Future attached to a different loop"。

- engine.py: 新增 run_db_blocking(fire_and_forget_db_write 的带返回值版本)。离开主 loop
  时用 run_coroutine_threadsafe 把协程送回主 loop 执行、阻塞调用线程拿结果;仅在无主 loop
  或正同步处于主 loop 内(送回会死锁)时退回临时 loop。
- thread_paths / prompt / skill_tools: 三处同步→异步阻塞读桥(任务owner解析/常驻技能/
  技能可见性)改走 run_db_blocking,删除各自重复的 _run_coro_blocking。
- llm/tool metrics 中间件: fire-and-forget 写改走 fire_and_forget_db_write。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-24 15:57:04 +08:00
fengxiangdebaozi 85aa242fd5 Merge branch 'master' into dhk-1 2026-06-23 19:45:41 +08:00
Administrator 053d3e300e Merge branch 'ji' into master (操作日志 Y-Log 上报)
# Conflicts:
#	CLAUDE.md
2026-06-23 19:22:53 +08:00
fengxiangdebaoziandClaude Opus 4.8 b8c78ba06e feat(ai-writing): 大纲先行确认 + 新文章类型识别 + 参考文献 + 编辑器修复
- 写作配置 Q&A 改为「先流式生成大纲并多轮确认 → 再确认意图」,放宽轮数上限
- 意图识别出列表外的新文章类型时带上类型名+简介(仅本次写作临时使用)
- 成稿支持参考文献:正文 [n] 角标 + 文末按出现顺序的「参考文献」列表(表单开关,默认开)
- 流式接取未闭合 <think> 思考内容,首 token 前加「正在整理…」占位
- 修复右侧编辑器输入空格只出一次、光标跳字(避免自激回灌 setContent 重置光标)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-23 19:07:03 +08:00
吉鑫升andClaude Opus 4.8 5aa66fc07d chore(backend): 放宽递归与循环检测阈值
- build_run_config recursion_limit 100 -> 250
- LoopDetectionMiddleware 默认阈值放宽(warn/hard/window/tool-freq)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-23 17:48:23 +08:00
董宏凯andClaude Opus 4.8 d383b0e9dd feat(roundtable): 第二/三步模型自动容错(出问题自动换下一个模型)
模型出问题时按 config.yaml models[] 顺序自动切到下一个模型继续,并告知用户
「『X』模型异常 → 已切换『Y』继续」。前台实时 + 后台挂起两条路径都覆盖。

- 中间件兜底/熔断 AIMessage 打 LLM_ERROR_MARKER 标记,is_llm_error_message 可识别
- roundtable_model_fallback: fallback_model_chain 候选链 + reason_text + 总开关
  env ROUNDTABLE_MODEL_FALLBACK
- 触发: 异常(start_run失败/idle卡死) / 兜底错误文案 / 交付为空(仅seat/leader)
- 后台 InProcessRoundtableGateway._run_turn_with_fallback 包住每轮 run
- 前台 _special_run/_leader_run 收尾检测失败→model_switch 帧+换模型重流
- 结构化抽取 flow_extract round1 astream 同款换模型循环
- Step2 单席位异常隔离: 单席位失败不拖垮整批/整链/整个作业
- 时间线插「系统提示」气泡 + toast + 诊断 model_switch 事件
- Tests: test_roundtable_model_fallback.py

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-20 21:37:45 +08:00
AdministratorandClaude Opus 4.8 23e6685977 fix(task-deeplink): 共享任务对话的 md/产物按对话创建者解析路径,修复切换账号后找不到文件
xdfx/rwfx/3qfx 任务深链对话按 metadata.taskId 全局共享、不区分用户,但生成的
md/artifacts 与上传文件物理存在 users/{user_id}/threads/{thread_id}/ 下,user_id
之前取当前登录账号。导致 A 账号生成的文件,B 账号打开同一共享对话时后端去
users/B/ 找 → 404「找不到」。

新增 deerflow/runtime/thread_paths.py(resolve_path_user_id /
aresolve_path_user_id):带 taskId 的共享对话统一解析到对话创建者
(threads_meta.user_id),个人对话仍走当前登录(行为不变)。在所有线程文件路径关口替换:
写入源 ThreadDataMiddleware、present_files、uploads(manager+middleware+router)、
产物读写(path_utils → artifacts/skills router)、嵌入式 client.get_artifact。
正向结果进程内缓存;无 DB/内存态/无行均安全回退到当前登录。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-19 14:35:22 +08:00
AdministratorandClaude Opus 4.8 5b610fd807 feat: 上下文压缩支持状态持久化缓存,前端过滤隐藏消息
- summarization_middleware 摘要缓存写入 LangGraph thread state,跨进程/重启可复用
- 摘要消息改为 SystemMessage,提示模型静默使用、不泄露
- thread_state 新增 context_summary_cache 字段
- app_config 加载时记录 summarization 配置日志
- config.yaml 压缩触发/保留阈值调整为生产值(token 预算)
- 前端 input-box / message-list 过滤 hide_from_ui 消息

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-16 22:16:06 +08:00
吉鑫升andClaude Opus 4.8 d82b5ed884 feat: 摘要中间件增加压缩日志,提炼模板编辑框定高
- summarization_middleware: 真正压缩与复用缓存均记录 info 日志,便于无 stream writer 场景(定时任务/渠道/测试)排查
- KnowledgeManage: 提炼模板 system prompt 编辑框固定为 50vh 高度,内部滚动

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-16 17:40:14 +08:00
吉鑫升andClaude Opus 4.8 5182ea5d11 feat: 技能收藏与常驻(always-on)支持
- 新增 skill_favorites / skill_always_on 迁移与持久化模型字段
- skills 路由、存储、前端设置页支持收藏与常驻
- 调整摘要中间件与 lead agent 提示词
- 同步岗位 position_sync 及相关配置清理

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-16 16:37:43 +08:00
吉鑫升andClaude Opus 4.8 5ebbb82d66 feat: 知识库文件夹/模板、技能压缩设置与摘要窗口调优
- 知识库新增文件夹与模板支持(模型/仓储/服务/迁移),前端知识管理组件
- 岗位管理模型与接口完善
- 技能新增压缩设置页与相关存储/工具能力
- 系统设置扩展配置项
- 摘要中间件 keep 窗口改用 token 预算(16000),减少重复压缩

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-16 15:23:54 +08:00
吉鑫升andClaude Opus 4.8 3855c30edd feat: 新增岗位管理、提示词模板与资源标签,完善用户提示词与推荐问题接口
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-16 10:49:36 +08:00
吉鑫升 5daa4b3e20 Merge remote-tracking branch 'origin/master' into ji 2026-06-11 16:39:30 +08:00
xuhd 1668c1da87 Merge remote-tracking branch 'origin/master' into xuhd
# Conflicts:
#	frontend-web/.env.production
#	frontend-web/.env.sit
#	frontend-web/package.json
#	offline-backend-20260512/backend/config.yaml
2026-06-11 14:19:01 +08:00
Administrator 7fb3870247 @
feat(summarization): 非破坏性上下文压缩——保留原始对话仅在请求时瞬态压缩

重写 DeerFlowSummarizationMiddleware:将 stock before_model 的
RemoveMessage(REMOVE_ALL_MESSAGES) 改为 no-op,改在 wrap_model_call 中
瞬态压缩为 [summary, *recent],完整原始对话仍保留在 state 中,前端继续
渲染用户真实问答。summary 消息标记 hide_from_ui 且永不持久化,按 thread
缓存以避免每次模型调用都触发 LLM。前端 isHiddenFromUIMessage 额外隐藏旧
版本持久化的 name="summary" 消息。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@
2026-06-11 08:34:09 +08:00
xuhdandClaude Opus 4.7 c493639119 feat(skill-review): strict-mode prompt + raise min_tool_calls 3 to 5
The review agent was creating too many low-value skills. Three changes
tighten the funnel:

- review_worker prompt: rewritten in "保守模式". Default decision is
  do-nothing. To create, all five AND conditions must hold (>=5 tool
  calls with step dependencies, an explicit fail->fix->success episode,
  cross-scenario reusability, no >=70% overlap with skill_list, can
  produce >=300 chars of substance). Adds 7 explicit NO categories
  (single-tool loops, one-off queries, LLM common knowledge, etc.)
  and a 4-question self-check before any create. Hard cap of 1 new
  skill per review, with patch preferred over create.
- min_tool_calls: 3 -> 5. Matches the prompt's ">=5 steps" hard
  condition so trivial turns never reach the LLM reviewer.
- Updated config.yaml, the default-value assertion in
  test_skill_review_worker, and the middleware docstring's "(default
  3)" annotation to keep all three references in sync.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-06-09 16:00:29 +08:00
xuhd bd0eceda61 Merge remote-tracking branch 'origin/master' into xuhd
# Conflicts:
#	offline-backend-20260512/backend/config.yaml
2026-06-09 15:43:40 +08:00
AdministratorandClaude Opus 4.8 81e4fd92ce feat(knowledge): 阶段三 知识治理 + 阶段四 图谱增强
阶段三(治理):
- sensitive.py:写入前脱敏 token/密钥/手机号/身份证/邮箱(sensitive_redaction_enabled)
- 版本快照:更新前写 knowledge_note_versions;GET /notes/{id}/versions
- dedup.py:标题/摘要 Jaccard + 内容哈希查重;GET /duplicates、POST /merge(来源并入主条目、重复项归档)
- auto_ingest.py:独立后台队列/线程沉淀,不阻塞问答;中间件 after_agent 在 auto_ingest_enabled 时入队;价值判断(最小字符)+ skip_existing 去重;自动沉淀落 draft 待审

阶段四(图谱):
- LLM 抽取 entities/relations;service 存 knowledge_entities/knowledge_relations 并连边
- exporter.build_graph 支持实体节点 + note→entity/entity→entity 边;graph.html/json 增强
- GET /api/knowledge/graph 返回产品图谱数据

前端:知识库页「查重合并」对话框 + 相关 api/hooks
配置:sensitive_redaction_enabled / dedup_similarity_threshold / entity_extraction_enabled
测试:tests/test_knowledge.py 20 passed

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-08 18:31:33 +08:00
AdministratorandClaude Opus 4.8 683cc6d76e feat(knowledge): 阶段二 向量 RAG 与召回注入
- embeddings.py:OpenAI 兼容 embedding 客户端,未配置时关键词兜底
- chunking.py + knowledge_embeddings 表:Markdown 分块与向量索引(Python 余弦)
- search.py:keyword/vector/hybrid 三种模式,向量不可用自动降级
- 笔记创建/更新自动重建向量索引;新增 /reindex 维护接口
- KnowledgeRagMiddleware:按对话级开关 knowledge_rag_enabled 注入 <knowledge-context>,接入 lead agent 中间件链(默认随 config.knowledge.rag_enabled)
- service.build_rag_context(wiki-query 风格分层召回)
- 笔记版本快照(knowledge_note_versions)+ GET /notes/{id}/versions
- 前端:聊天页「知识库」对话级 RAG 开关 + 类型字段
- 新增 entities/relations 表(阶段四预置)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-08 18:18:04 +08:00
g zy 9ebde3b315 Merge branch 'gzy_iframedf' into 'master'
feat: iframe传 文件进入

See merge request knowledge-base-bj/deerflow-server!93
2026-06-05 07:10:59 +00:00
ZyGuo 5f6ab41798 feat: iframe传 文件进入 2026-06-05 14:15:42 +08:00
xuhd 335bc3cf48 fix(skill-review): propagate user_id through review chain for correct DB ownership 2026-06-05 13:32:19 +08:00
xuhd 5b5e4f8ff0 fix(thread-data): guard runtime.context None on line 110 2026-06-05 13:32:19 +08:00
xuhd 2a8364e902 fix(skill-review): use runtime.context to get thread_id 2026-06-05 13:32:19 +08:00
xuhd 23a18ec318 fix(skill-review): after_agent must be sync, not async 2026-06-05 13:32:19 +08:00
xuhdandClaude Sonnet 4.6 0e7a1de4db feat(skill-evolution): add SkillReviewMiddleware to trigger post-turn review
Implements SkillReviewMiddleware that hooks into after_agent to call
maybe_start_review() as a background task after each agent turn, without
blocking the HTTP response. Exceptions are caught and logged only.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-05 13:32:19 +08:00
吉鑫升 cbd07c9d1b chore: update ji branch work 2026-06-04 17:46:11 +08:00
吉鑫升 aeef1dca59 Merge remote-tracking branch 'origin/ji' into ji 2026-06-03 19:01:56 +08:00
吉鑫升 a1d460e220 Merge remote-tracking branch 'origin/master' into ji
# Conflicts:
#	frontend-web/src/components/workspace/messages/message-list-item.tsx
#	offline-backend-20260512/backend/app/gateway/app.py
#	offline-backend-20260512/backend/packages/harness/deerflow/runtime/scheduler/html_page.py
#	offline-backend-20260512/backend/packages/harness/deerflow/runtime/scheduler/service.py
2026-06-03 19:01:45 +08:00
吉鑫升 e2a962de32 chore: update ji local changes 2026-06-03 18:57:49 +08:00
fengxiangdebaozi 8f6e08eb2f Merge branch 'dhk-1' of http://192.168.0.200/knowledge-base-bj/deerflow-server into dhk-1 2026-06-03 18:54:26 +08:00