fischer
/
fischer-agentkit
1
0
Fork 0
Code Issues Pull Requests Actions 87 Packages Projects Releases Wiki Activity
296 Commits 1 Branch 0 Tags 20 MiB
Commit Graph

183 Commits

Author SHA1 Message Date
chiguyong 31c65e01b8 fix(security): P0 安全加固 + 多实例部署一致性 (U1-U4 + U5c)
Deploy to Production / deploy (push) Has been cancelled Details 
U1: LLM gateway KB 缓存 fail-closed — 异常时默认禁用缓存防止 KB 数据泄漏
U2: MCP 危险工具黑名单过滤 — 6+1 端点覆盖，防止绕过 chat confirmation
U3: SecretsStore Redis 迁移 — 多 worker 共享凭证，内存降级保留开发模式
U4: channels webhook Redis 状态 — ZSET 滑动窗口限流 + nonce dedup + backpressure
U5c: ce-code-review 修复批次:
  - P0: 统一 MCP 黑名单与 publisher.py 一致 (terminal_execute -> terminal, +file_read)
  - P1: ZSET 限流 member 加 uuid 后缀避免同时间戳碰撞
  - P1: SecretsStore redis 参数 Any -> aioredis.Redis | None (AGENTS.md 合规)
  - P1: Redis client 添加 socket_timeout 防止单点故障请求挂死

测试: 171 scoped tests pass, ruff clean
 2026-06-26 04:05:33 +08:00
chiguyong 53faa60472 fix(review): ce-code-review P1+P2 修复 — 安全/可靠性/性能 
P1 安全与可靠性（4 项）：
- wecom: verify_signature 增加时间戳新鲜度校验（5 分钟窗口防重放）
- cache: should_cache 在 per_user_namespace 开启时拒绝 user_id=None
  匿名请求，避免跨用户缓存泄漏（安全要求 a/e）
- channels: webhook receive_message 异常兜底，防止 500 触发平台重试风暴
- app: shutdown 调用 close_all_adapters + await _pending_webhook_tasks，
  防止 httpx 连接泄漏和丢失 IM 回复

P2 效率与可维护性（5 项）：
- feishu: _TOKEN_CACHE_TTL 300 → 6900（2h 减 5min 余量，避免 24x 过频刷新）
- channels: _pending_webhook_tasks 有界化（2x 并发上限时 429 拒绝）
- gateway: quota 检查每 period 单次 get_usage，复用 summary 检查 token+cost
- cache_key: generate_cache_key 合并为单次 SHA-256（消除 8-10 次冗余哈希）
- config: ProviderConfig.get_api_key 移除未用的 secrets_store 参数

P3 去重（1 项）：
- channels: _process_inbound_message DIRECT_CHAT 路径提取 _direct_chat 辅助函数

测试：
- test_wecom: 时间戳改用 int(time.time())，新增 test_expired_timestamp_rejected
- test_cache: should_cache 测试覆盖匿名拒绝 + namespace_off 兼容
- test_config_migration: get_api_key 测试适配新签名
- channels/config_migration/quota_enforcement 测试全部通过
 2026-06-26 01:40:31 +08:00
chiguyong 1ccaf56b9a refactor: ce-simplify-code 审查修复 — 去重 + 效率 + 死代码清理 
3 个审查代理（复用/质量/效率）发现 15 个问题，全部修复：

效率与安全（6 项）：
- MCPClient 缓存 MultiServerMCPClient 单例 + aclose()，修复连接/子进程泄漏
- _rate_limits 清理空 IP 条目，修复 X-Forwarded-For 欺骗下内存泄漏
- _seen_nonces 改用 OrderedDict，O(1) 摊销过期清理
- webhook 后台任务加 Semaphore(20) + 任务引用追踪，限制无界并发
- _build_adapter 用 asyncio.gather 并行解密 secrets
- 适配器实例缓存（_adapter_cache），token TTL 缓存跨请求命中

去重（4 项）：
- header_get 提取到 channels/base.py，4 个适配器统一 import
- _get_client/close() 移入 MessageAdapter 基类，子类继承
- URLVerificationChallenge 统一到 base.py，feishu/slack/wecom 共用
- Transport ABC 添加 endpoint_url 属性，from_transport 不再访问私有字段

死代码与类型安全（5 项）：
- detect_cache_hit 死方法替换为 record_cache_result 公开 API
- execution_mode.value == "direct_chat" 改用枚举比较
- 删除 yielded_any 死变量、重复 from fastapi import Request、
  多余 getattr 防御

453 tests passed, ruff clean（预存 F841 非本次引入）
 2026-06-25 23:54:14 +08:00
chiguyong 793476cafa feat(llm): U17 — LiteLLM 语义缓存替换 + per-user/ACL scope 安全隔离 
- 新增 LitellmCacheManager：配置 litellm.cache 全局，三级后端 fallback
  (RedisSemanticCache -> RedisCache -> InMemoryCache)，redisvl lazy import
- cache_key 扩展 user_id + kb_acl_hash 参数（安全要求 a/b/e）
- gateway 集成：读取 KB caching_disabled flag（安全要求 c），构建带 scope
  的 cache_key，命中时 cost=0
- LLMResponse 新增 cache_hit 字段；LLMRequest 新增 cache 参数
- litellm_provider 透传 cache 参数 + 检测 _hidden_params 缓存命中
- 33 个新测试覆盖 13 场景（含 User A != User B 缓存隔离）
- 旧 InMemoryLLMCache/RedisLLMCache 保留向后兼容
 2026-06-25 22:49:59 +08:00
chiguyong 86541d7172 feat(mcp): U16 — langchain-mcp-adapters client replacement + transport deprecation 
- 重写 MCPClient：URL scheme 自动检测（stdio/http/sse）→ langchain config
- 旧 Transport 注入路径保留（DeprecationWarning），向后兼容
- transport.py 模块级弃用警告
- 28 个新测试覆盖 URL 检测、list_tools、call_tool、legacy 路径、ImportError
- 修复 manager.py / transport.py 预存 F401/F841
 2026-06-25 22:04:37 +08:00
chiguyong 069dbc22b1 feat(llm): U15 — LiteLLM unified provider + api_key encrypted secrets migration 2026-06-25 21:41:15 +08:00
chiguyong 13c516a54f feat(mcp): U14 — Skill/Team MCP publish with admin auth + dangerous-tool opt-in 2026-06-25 21:10:06 +08:00
chiguyong 16c33be295 feat(mcp): U13 — refactor MCPServer to route factory + mount at /api/v1/mcp with auth 2026-06-25 20:58:41 +08:00
chiguyong 8998f94c42 feat(channels): U12 — DingTalk/WeCom/Slack adapters + multi-channel webhook dispatch 2026-06-25 20:45:43 +08:00
chiguyong 4b58e8f661 feat(channels): U11 — Feishu IM adapter end-to-end (webhook + signature + AES-CBC decrypt + chat integration) 2026-06-25 20:24:21 +08:00
chiguyong 5572387c01 feat(channels): U10 — message adapter ABC + AES-256-GCM secrets store + channel CRUD routes 2026-06-25 20:13:37 +08:00
chiguyong e3ae2f3a56 feat(rag_platform): U8 — TaskIQ async task integration 
Add tasks.py: TaskManager with vectorize/batch_index tasks, per-user concurrency limits, degraded mode (sync execution without broker), WorkerSweeper for timeout detection, error message sanitization
Add taskiq>=0.11 and taskiq-redis>=0.5 to pyproject.toml
Task parameter schema validation (VectorizeTaskParams, BatchIndexTaskParams)

Tests: 41 new tests, 289 total passing
 2026-06-25 12:58:51 +08:00
chiguyong d026a91f43 feat(rag_platform): U6 — hit processing mode + KB settings 
Add hit_processing.py: HitProcessor with model_opt (LLM-generated) and direct (concatenated chunks) modes, with in-process cache
Add settings.py: KBSettings/KBSettingsUpdate models, KBSettingsStore with async CRUD
Add KB settings endpoints to kb_management.py: GET/PUT /kb-management/kbs/{kb_id}/settings with owner-only modification

Tests: 43 new tests (25 hit_processing + 18 settings), 293 total passing
 2026-06-25 12:44:47 +08:00
chiguyong 5c562dbff3 feat(rag_platform): U5 — rerank + question generation + termbase 
Add rerank.py: Reranker with Cohere/BGE provider support, data export risk annotation, graceful degradation
Add question_gen.py: LLM-based question generation following ContextualChunker pattern, with caching
Add termbase.py: jieba custom dictionary management, add/remove/load terms

Tests: 58 new tests (14 rerank + 19 question_gen + 25 termbase), 205 total passing
 2026-06-25 12:31:43 +08:00
chiguyong fb9f16d6e5 feat(rag_platform): U4 — dual-index retrieval (pgvector semantic + PG fulltext jieba) 
Add fulltext.py: jieba tokenization + tsvector write/query
Add retrieval.py: RetrievalEngine with embedding/keywords/blend modes
Update models.py: add RetrievalRequest model
Tests: 35 new tests, 147 total passing
 2026-06-25 12:20:48 +08:00
chiguyong 3f9588e673 feat(rag_platform): U3+U7 — rewrite upload endpoint with sanitization + pipeline 
Rewrite upload_document() to use rag_platform sanitize + DocumentProcessor:
- File type whitelist validation (8 allowed types, reject .exe/.sh)
- File size limit (50MB) + zip bomb detection for ZIP-based formats
- DocumentProcessor.parse() (with content sanitization) + segment()
- Return chunks preview, status="segmenting" (pending vectorization)

Add POST /kb-management/documents/preview endpoint:
- Pre-upload preview with adjustable chunk_size/chunk_overlap
- Same security validation as upload, no document record created

Add POST /kb-management/documents/{id}/vectorize placeholder:
- Returns 503 — full async vectorization deferred to U8 (TaskIQ)

Test: update test_upload_document assertion (status "indexed" → "segmenting")
 2026-06-25 12:06:16 +08:00
chiguyong b55c896794 feat(rag_platform): U3+U7 — document processing pipeline + upload security 
U3: Document processing pipeline (document_processor.py)
- DocumentProcessor class wrapping parse → segment → vectorize
- parse() uses memory/document_loader.py for multi-format extraction
- segment() uses LlamaIndex SentenceSplitter
- preview() returns chunks for read-only preview (no vectorization)
- vectorize() embeds chunks and stores in pgvector (all-or-nothing)
- process() orchestrates full pipeline with status transitions:
  pending → parsing → segmenting → vectorizing → indexed | failed

U7: Upload security & content sanitization (sanitize.py)
- ALLOWED_FILE_TYPES whitelist (pdf/docx/xlsx/pptx/txt/md/csv/html)
- MAX_FILE_SIZE 50MB limit
- validate_file_type() / validate_file_size() guards
- check_zip_bomb() for ZIP-based formats (ratio > 100:1 or > 500MB)
- check_image_bomb() for pixel count > 100MP (PNG/JPEG/GIF header parsing)
- is_safe_ip() SSRF protection (loopback/RFC1918/link-local/ULA denied)
- sanitize_markdown() removes dangerous HTML tags (script/iframe/object/embed)
- sanitize_content() main entry point for text format sanitization
- parse_xml_safe() XXE protection (forbid_dtd/forbid_entities/forbid_external)

Preview API (preview.py)
- PreviewChunk / PreviewResult Pydantic models
- generate_preview() returns read-only segmentation preview

Tests: 112 tests passing (45 new + 67 existing)
- test_sanitize.py: file type/size, markdown sanitization, SSRF, zip/image bomb
- test_document_processor.py: parse/segment, preview, vectorize, failure status
 2026-06-25 11:21:42 +08:00
chiguyong c1a21f57a1 feat(rag_platform): U2 — KB persistence + per-KB ACL 
Add PostgreSQL-backed KB store replacing in-memory KnowledgeSourceStore:
- models.py: ORM models (KBModel, DocumentModel, KBAclModel) using
  SQLAlchemy 2 DeclarativeBase + Mapped style
- store.py: KBStore with async CRUD for KBs and documents,
  create_kb creates owner ACL in same transaction
- acl.py: filter_kb_by_user_acl(), grant_access(), revoke_access(),
  list_acl() — follows filter_kb_sources_by_department pattern

Schema: rag_platform_kbs, rag_platform_documents, rag_platform_kb_acl
with FK CASCADE on kb_id. UniqueConstraint on (kb_id, user_id).

Tests: 23 unit tests covering KB CRUD, document operations, ACL
filtering, grant/revoke. All 37 rag_platform tests pass.
 2026-06-25 11:01:04 +08:00
chiguyong 27d0184392 feat(rag_platform): U1 — RAG platform skeleton + LlamaIndex integration 
Create src/agentkit/rag_platform/ module with:
- models.py: Pydantic domain models (KB, Document, Chunk, QueryResult)
- indexing.py: PGVectorStore wrapper with explicit table name
  (rag_platform_kb_chunks) for schema isolation from episodic_memory
- pipeline.py: RAGPipeline wrapping LlamaIndex IngestionPipeline
  (SentenceSplitter + embedding + vector store)

Add dependencies: llama-index-core, llama-index-vector-stores-postgres,
llama-index-embeddings-openai, pgvector, jieba.

Tests: 14 unit tests covering models, indexing (URL conversion, table
name isolation, embed_dim), and pipeline (ingest, query, chunk params).
 2026-06-25 10:49:35 +08:00
chiguyong bbbf9cd40a feat(bitable): add bitable companion service with full P0-P2 fixes 
Bitable is a multi-dimensional table companion service that runs alongside
the main AgentKit server. It provides structured data storage with formula
fields, views, and ingestion pipelines.

Major components:
- Domain models (Pydantic v2): Table, Field, Record, View, RecalcTask
- SQLAlchemy 2 async ORM with independent bitable PostgreSQL schema
- Formula engine: AST parser, DAG, Kahn topological sort, safe eval
- RecalcWorker: atomic task claiming (FOR UPDATE SKIP LOCKED), topo-order
  processing, stale-threshold reaper for crash recovery
- REST API (/api/v1/bitable): tables, fields, records, views, files
- BitableTool: agent-facing tool with batch chunking (500/batch)
- CLI: agentkit bitable subcommands (create, list, import-excel, etc.)
- Frontend: Vue 3 + vxe-table grid with field management, views, filters
- Ingestion: Excel (openpyxl), database reflection, API collector

Security fixes (ce-code-review P0 + ce-debug P1):
- SQL injection prevention (field_id validation, parameterized queries)
- IDOR protection (_check_table_ownership on all table-level endpoints)
- SSRF prevention (URL scheme + private IP validation in parse_excel_url)
- OOM prevention (streaming file upload, batch delete, batch insert)
- Atomic recalc task claiming (FOR UPDATE SKIP LOCKED)
- Formula engine cache invalidation on field changes
- Composite cursor pagination for non-id sort orders
- Batch upsert (eliminates N+1 queries)
- Sync I/O offloaded to thread pool in async contexts
- Internal token auth (X-Internal-Token, hmac.compare_digest)
- PK unique index enforcement

Test coverage: 88 unit tests (95 skipped without Docker)
 2026-06-25 01:09:59 +08:00
chiguyong 567cbc9c9b refactor: simplify code across U1-U7 (bug fix + efficiency + reuse + quality) 2026-06-24 22:35:52 +08:00
chiguyong 0847c0e086 fix(checkpoint): add TTL expiration for memory fallback mode 
内存降级模式之前没有 TTL 过期机制，长期运行进程会导致内存泄漏。
现在 list_checkpoints 和 load_plan 在内存模式下会过滤/清除过期数据。

- list_checkpoints: 内存降级分支过滤过期 checkpoint
- load_plan: 内存降级分支检查 TTL 过期，过期则清除并返回 None
- 新增 _is_expired 方法检查 saved_at 是否超过 TTL
- _memory_plans 类型改为 tuple(plan_dict, timestamp) 以支持 TTL
- 新增 5 个 TTL 过期测试覆盖内存模式和 Redis 降级场景
 2026-06-24 22:04:55 +08:00
chiguyong fa152e24ac feat(skills): add progressive skill loading with disclosure_level=0 (U5) 
When disclosure_level=0, system prompt only injects skill name + description
(summary mode). SkillDetailTool is injected into the tool set, allowing the
LLM to load full instructions on-demand via skill_detail(query). This reduces
context window consumption when many skills are registered.
 2026-06-24 21:49:00 +08:00
chiguyong dfd188b1a4 feat(orchestrator): add pipeline checkpoint and crash recovery (U7) 
Add PipelineCheckpoint for stage-level crash recovery with Redis-first
+ memory fallback. TeamOrchestrator saves checkpoints after each phase
finalizes and supports resume(plan_id) to continue from the last
completed phase. New POST /api/v1/tasks/{id}/resume endpoint recreates
the team from saved plan and calls resume.
 2026-06-24 21:04:18 +08:00
chiguyong 3dfda904d7 feat(core): add middleware pipeline architecture with onion model 
U6: Unified middleware protocol (before/after) with MiddlewareChain
implementing onion model execution. Parallel integration (KTD1) —
middleware path controlled by presence of middleware_chain parameter,
existing ReActEngine path unchanged when None.

- New core/middleware.py: RequestContext, Middleware protocol,
  MiddlewareChain (onion model: before outer→inner, after inner→outer)
- 3 example middlewares: SummarizationMiddleware (U3 headroom compression),
  TokenUsageMiddleware, LoopDetectionMiddleware (request-level audit)
- ReActEngine.__init__ accepts middleware_chain parameter
- execute() branches: middleware path when chain present, existing path otherwise
- 22 tests covering ordering, error handling, state passing, backward compat
 2026-06-24 20:52:15 +08:00
chiguyong ef84e3fd53 feat(experts): add SharedWorkspace state offloading for long-horizon runs 
U4: ExpertTeam accepts redis_client, passes to SharedWorkspace. After phase
completion, full result is written to workspace and in-memory phase.result
is replaced with a 500-char summary + _ref_key. Dependency output reading
resolves offloaded content from workspace on demand, with graceful fallback
to summary on read failure.

Tests: 8 scenarios (offload creation, short content, dependency resolution,
workspace failure fallback, non-offloaded passthrough, redis_client wiring,
memory dict fallback, pipeline integration) — all pass.
 2026-06-24 20:32:10 +08:00
chiguyong 122173ec2c feat(core): add headroom-based compression trigger 
U3: ContextCompressor now accepts model_context_limit, headroom_threshold,
and min_tokens. should_compress() triggers when token ratio exceeds 0.8 of
model limit OR exceeds min_tokens (8000 fallback). ReActEngine._should_compress
delegates to compressor when available, checks is_available() first.

Tests: 6 scenarios (headroom trigger, min_tokens guard, small model,
unavailable compressor, delegation, fallback) — all pass.
 2026-06-24 20:28:14 +08:00
chiguyong 717aad1303 feat(experts): add concurrency limit to TeamOrchestrator parallel phases 
U2: Add asyncio.Semaphore to bound concurrent phase execution and debate
argument generation. Default limit=3, configurable via max_concurrent_phases.
Prevents LLM rate-limit spikes when many phases run in the same layer.

Tests: 5 scenarios (happy path, 5-phase edge case, serial mode, failure
release, debate integration) — all pass.
 2026-06-24 20:23:30 +08:00
chiguyong 018b342d96 feat(react): add loop detection to prevent repeated identical tool calls 
U1: Sliding window hash detection in ReAct loop. When the same tool is
called with identical arguments >= threshold times (default 2), injects
a correction message first, then raises LoopDetectedError if the LLM
doesn't change strategy. Covers both _execute_loop and execute_stream.
 2026-06-24 20:12:35 +08:00
chiguyong a312e584ae Merge branch 'feat/expert-team-pm-collaboration' — PM 协同模式 + 代码审查全量修复
Deploy to Production / deploy (push) Waiting to run Details 
# Conflicts:
#	src/agentkit/server/frontend/components.d.ts
 2026-06-24 18:57:37 +08:00
chiguyong 20a4c55d5b feat(skills): SkillHarness 前置条件 + 风险守卫学习增强 
- cli/skill.py: skill learn 子命令增强
- evolution/risk_guard_learner.py: 风险守卫学习改进
- memory/models.py: 记忆模型扩展
- skills/base.py + loader.py: SkillHarness 前置条件支持
- 对应测试更新
 2026-06-24 18:56:51 +08:00
chiguyong 574db8458f fix(experts): PM 协同代码审查全量修复 
P0: 跨阶段契约状态同步 — _notify_collaborators 更新接收方契约状态为 received
P0: 4 个 PM 事件加入 _VALID_TEAM_EVENT_TYPES 白名单

P1: 验收 fail-open 改标注降级原因
P1: 返工失败抛 RuntimeError 而非返回 dict
P1: 验收 prompt injection 防护 — 专家输出用 XML 标签包裹
P1: 契约字段校验 _EXPERT_NAME_RE
P1: bool("false") 修复 — 显式比较避免字符串真值陷阱
P1: _parse_risk_flags(None) 防御

P2: _notify_collaborators 移到验收通过后
P2: SharedWorkspace 写入移到验收通过后
P2: 验收贪婪正则修复
P2: 风险标记数量上限 MAX_RISK_FLAGS=10
P2: 返工 feedback 截断
P2: 前端会话隔离 — 切换会话时清除/恢复 collaborationState
P2: 前端契约状态更新 — collaboration_notice 时标记 delivered
P2: CLI 死代码标注 + 异常改 debug 日志
P2: 模块级 _RISK_FLAG_RE 预编译
 2026-06-24 18:56:27 +08:00
chiguyong 6016c087fe feat(cli): U6 CLI 协同事件 Rich 渲染 
- chat.py 新增 _render_collaboration_contracts 和 _render_pm_collaboration_event
- 4 种 PM 协同事件渲染：
  collaboration_contract_defined (cyan Panel)
  collaboration_notice (蓝→品红 文本)
  review_result (passed=green / failed=red Panel)
  risk_flagged (yellow Panel)
- plan_update 中提取 collaboration_contracts 并渲染
- _print_help 更新项目经理模式说明
- 优雅降级：字段缺失回退到 ?，空契约不输出，整体 try/except 不中断编排
- 新增 11 个测试（TestPMCollaborationRendering 9 + TestPrintHelpPMMode 2）
- ruff 通过，pytest 23 passed
 2026-06-24 14:57:49 +08:00
chiguyong 5487cca199 feat(experts): U4 专家风险标记 + risk_flagged 事件 
- orchestrator 新增 _parse_risk_flags 静态方法，正则解析 [RISK: ...] 标记
- _execute_execution_phase 在协作通知后、验收前解析风险标记
- 风险标记通过 risk_flagged 事件广播，供前端/CLI 渲染
- 无风险标记时行为不变，向后兼容
- 新增 TestRiskFlagging 7 个测试（单/多/无/格式错误/事件发出/内容/兼容）
 2026-06-24 14:17:58 +08:00
chiguyong 62fcbc0feb feat(experts): U3 Lead 验收环节 + 返工机制 
- PlanPhase 添加 rework_count 和 review_feedback 字段
- 添加 _review_phase_output 方法，Lead 用 LLM 验收阶段输出
- _execute_execution_phase 重构为返工循环（MAX_REWORKS=2）
- 验收通过/返工/失败三种路径，发出 review_result 事件
- LLM 不可用时优雅降级直接通过
- 6 个新测试，全套 449 passed 无回归
 2026-06-24 14:09:18 +08:00
chiguyong fef7ecea39 feat(skills): SkillHarness 激活前置条件 + 风险守卫学习 
基于 SkillHarness 论文（arXiv:2606.20636）与 Agent Skills 综述
（arXiv:2602.12430）引入激活前置条件（preconditions）与来源标记
（provenance），并新增从失败轨迹学习风险守卫建议的能力。

变更内容：
- U1: SkillConfig 新增 v7 preconditions/provenance 字段（base.py）
- U2: build_skill_system_prompt 注入 preconditions 软检查段落
- U3: SkillLoader 三路径记录 provenance + entry_points 危险能力告警
- U4: 10 个业务 Skill YAML 补充 preconditions（2-4 条中文短句）
- U5: RiskGuardLearner 从失败轨迹学习风险守卫建议（人工审查，不自动应用）
- U6: CLI 命令 agentkit skill learn-risk-guards

关键决策：
- KTD1: preconditions 通过 system_prompt 注入（软检查），不做硬 LLM 调用
- KTD2: RiskGuardLearner 不自动应用，需人工审查（论文显示 75% 自动学习不安全）
- KTD3: provenance 为轻量字符串，不加 hash/签名（无合规需求）

测试：39 个新增单元测试全部通过，ruff 检查通过。
 2026-06-24 13:56:37 +08:00
chiguyong c46cf06f6d feat(experts): U2 协作契约执行 — 专家可见 + 主动通知 
- _execute_execution_phase 按协作契约读取相关专家输出（可见性）
- 添加 _notify_collaborators 方法，完成后通知相关专家（可协助）
- 发出 collaboration_notice 事件，契约状态更新为 delivered
- 7 个新测试，全套 443 passed 无回归
 2026-06-24 13:54:38 +08:00
chiguyong f219c5f016 feat(experts): U1 协作契约数据模型 + Lead 生成契约 
- PlanPhase 添加 collaboration_contracts 字段（CollaborationContract dataclass）
- 修改 _decompose_task prompt，要求 Lead 分解任务时定义协作契约
- 修改 _parse_phases 解析 LLM 返回的协作契约信息
- plan_update 事件自动包含协作契约（通过 to_dict 序列化）
- 71 + 9 = 80 个新测试，全套 436 passed 无回归
 2026-06-24 13:44:50 +08:00
chiguyong b86100a0a1 feat(cli): U6 CLI 多 Agent 入口 + 辩论 Rich 渲染 
- 新增 _execute_team_cli() 处理 @team 前缀，运行 ExpertTeam 流水线
- Rich 事件渲染：team_formed/plan_update/phase_*/debate_*/team_synthesis
- 干预循环使用 select.select() 非阻塞轮询 stdin（Unix-only，ponytail 标注）
- 支持 /debate 手动触发辩论、/stop 终止团队、纯文本作为上下文注入
- 扩展 _print_help() 增加 Multi-Agent 与 Interventions 说明
- 新增 12 个单元测试覆盖路由、帮助文档、函数返回值、干预基础设施
 2026-06-24 13:03:57 +08:00
chiguyong c831e925b6 feat(experts): U4 用户干预通道 + 手动辩论触发 
建立 @team 执行期间的用户干预通道，支持 /stop、/debate <topic>、
普通文本追加上下文。

ExpertTeam (src/agentkit/experts/team.py):
- 新增 _interventions: asyncio.Queue (maxsize=64) 干预队列
- add_user_intervention(msg): 广播 + 入队
- consume_user_interventions(): 排空并返回待处理干预
- broadcast_user_message 现在同时入队干预队列

TeamOrchestrator (src/agentkit/experts/orchestrator.py):
- 新增 _user_context: list[str] 累积普通文本干预
- 新增 _process_interventions(lead, plan) 在每层执行前调用：
  * /stop → 终止执行，广播 plan_update(stopped_by_user)
  * /debate <topic> → 动态插入 DEBATE phase（受 MAX_DEBATES 限制）
  * 普通文本 → 累积到 _user_context
- _synthesize_results 将 _user_context 追加到 synthesis prompt

WS 路由 (src/agentkit/server/routes/chat.py):
- 模块级 _active_teams dict 跟踪每个 session 的活跃团队
- _execute_team_collab 执行前注册、finally 注销
- WS 消息循环：若 session 有活跃团队，message 路由为干预而非新任务
- 新增 team_intervention_ack 确认消息

测试：tests/unit/experts/test_team_intervention.py（20 测试），
覆盖队列基础、/stop、/debate、普通文本、混合消息、synthesis 影响。
同步更新 test_orchestrator_debate.py 的干预通道兼容性测试
（U4 已实现 consume_user_interventions）。

全部 418 experts 测试 + 325 server 测试通过。
 2026-06-24 12:17:09 +08:00
chiguyong 5b5bd44ac4 test(calendar): 7 integration flow tests (lifecycle, recurrence, tags, types, invitations, authz, ICS) 2026-06-24 12:04:42 +08:00
chiguyong 3fdee65979 fix(calendar): code review fixes - 23 issues (2 critical, 15 major, 6 minor) 2026-06-24 11:29:23 +08:00
chiguyong ac26d417b3 feat(experts): U3 分歧检测 + 方案评审辩论自动触发 
在 TeamOrchestrator 中新增 4 个方法实现自动辩论触发：

- _maybe_add_plan_review_debate: 任务分解后可选插入方案评审 DEBATE
  phase（phases > 2 且 LLM 判断需要时），所有执行阶段依赖它
- _detect_divergence: 每层执行后用 LLM 判断已完成阶段产出是否与其他
  阶段存在分歧，偏好 false negative
- _insert_debate_phase: 动态插入 DEBATE phase 并重 wiring 依赖
  （原依赖 trigger 的 phase 现在依赖 DEBATE）
- _check_divergence_and_insert_debates: 每层完成后的协调入口，
  受 MAX_DEBATES=3 上限保护

主循环从 `for layer in layers` 改为 `while True` + 重新计算
topological_sort()，以支持动态插入 DEBATE phase 后的依赖分层。

测试：tests/unit/experts/test_divergence_detection.py（21 测试），
覆盖 happy path / 边界 / 错误路径 / 集成分层。同步修复
test_team_orchestrator.py 的 mock gateway 以适配 U3 的额外 LLM 调用。

全部 398 测试通过。
 2026-06-24 11:09:53 +08:00
chiguyong fbe08cb1e2 feat(experts): add debate phase executor to TeamOrchestrator (U2) 
Implement _execute_debate_phase() with Lead-facilitated structured debate:
- Lead opens with divergence point + dependency context
- Experts argue in parallel per round (asyncio.gather)
- Lead summarizes each round, then adjudicates final verdict
- Verdict produces decision (adopt/compromise/shelve/inconclusive) + conclusion
- Conclusion written to SharedWorkspace for downstream phases

Escape hatches:
- debate_config.skip=true short-circuits with template text
- MAX_DEBATE_ROUNDS=4 hard cap on rounds
- User /stop intervention ends debate early (U4-compatible via getattr fallback)
- LLM unavailable falls back to template verdict, no crash

New events: debate_started, expert_argument, debate_round_summary,
debate_resolved (plus existing phase_completed for consistency).

Phase dispatcher (_execute_phase) routes by phase_type:
EXECUTION to _execute_execution_phase, DEBATE to _execute_debate_phase.

36 new tests in test_orchestrator_debate.py covering happy path (2 rounds,
2 experts), max_rounds=1 boundary, empty participants, user stop, skip
escape hatch, LLM unavailable, SharedWorkspace integration, event
broadcasting, intervention channel compatibility, and helper methods.
All 377 expert tests pass.

Also includes planning artifacts (brainstorm requirements + implementation
plan with 6 units U1-U6).
 2026-06-24 10:54:51 +08:00
chiguyong e539122314 feat(experts): add PhaseType enum and debate_config to PlanPhase 
U1: Data model foundation for structured debate collaboration.
- Add PhaseType enum (EXECUTION | DEBATE)
- Add phase_type and debate_config fields to PlanPhase
- Update to_dict/from_dict for serialization with backward compatibility
- Add tests for PhaseType, debate phase creation, serialization, and
  mixed EXECUTION+DEBATE topological sort
 2026-06-24 10:42:11 +08:00
chiguyong 8d4145ddf9 feat(calendar): U7 Outlook sync via Microsoft Graph API 
OutlookSyncProvider implementing AbstractSyncProvider for
bidirectional Outlook Calendar sync. Uses Graph API delta query
for incremental pull, auto-refreshes OAuth tokens on 401, and
converts Outlook recurrence patterns to RRULE. Same conflict
resolution as CalDAV (last-write-wins + WS notification).

- src/agentkit/calendar/sync/outlook_provider.py — OutlookSyncProvider
- tests/unit/calendar/test_sync_outlook.py — 8 tests
 2026-06-23 23:49:24 +08:00
chiguyong 40d326cd3f feat(calendar): U6 CalDAV sync provider and SyncManager 
AbstractSyncProvider interface with CalDAVSyncProvider implementation
for bidirectional Apple Calendar sync. SyncManager orchestrates all
providers (G8) — sync_all/sync_provider/resolve_conflict with
last-write-wins + WS notification on conflicts (G4). caldav library
calls wrapped in asyncio.to_thread for non-blocking operation.

- src/agentkit/calendar/sync/base.py — AbstractSyncProvider ABC
- src/agentkit/calendar/sync/caldav_provider.py — CalDAVSyncProvider
- src/agentkit/calendar/sync/manager.py — SyncManager (G8)
- pyproject.toml — added caldav>=1.3 dependency
- tests — 12 tests (9 CalDAV + 3 SyncManager)
 2026-06-23 22:52:29 +08:00
chiguyong ffb184acc7 feat(calendar): U8 iCal/ICS import and export 
ICSProvider parses .ics files (icalendar library) and creates local
CalendarEvents, skipping duplicate UIDs. Export builds an iCalendar
from events in a date range, deduplicating recurring event
occurrences back to a single VEVENT with RRULE. REST endpoints:
POST /import-ics (multipart upload), GET /export-ics (download).

- src/agentkit/calendar/sync/__init__.py — sync subpackage init
- src/agentkit/calendar/sync/ics_provider.py — ICSProvider (import/export)
- src/agentkit/calendar/db.py — added get_event_by_external_id() for dedup
- src/agentkit/server/routes/calendar.py — import-ics and export-ics endpoints
- pyproject.toml — added icalendar>=5.0 dependency
- tests/unit/calendar/test_ics_provider.py — 8 tests
 2026-06-23 22:20:07 +08:00
chiguyong 26efbb51db feat(calendar): U5 reminder subsystem with scheduler and multi-channel dispatch 
ReminderScheduler scans upcoming events every 60s, matches reminder
rules, and dispatches via client (WS), email (SMTP), or webhook
channels. Idempotent delivery (no duplicates on re-scan), retry with
exponential backoff (up to 3 attempts). Follows task_store.py
start/stop asyncio loop pattern (KTD-2 — conscious deviation from
APScheduler).

- src/agentkit/calendar/scheduler.py — ReminderScheduler (start/stop/scan_once)
- src/agentkit/calendar/reminders.py — ReminderDispatcher (strategy per channel)
- src/agentkit/calendar/db.py — added list_all_events_in_time_range() for scheduler
- tests/unit/calendar/test_scheduler.py — 8 tests
- tests/unit/calendar/test_reminders.py — 9 tests
 2026-06-23 22:19:57 +08:00
chiguyong ddcedb57b2 feat(calendar): U4 post-processing extractor with keyword gating 
Adds PostProcessingExtractor — a zero-LLM keyword gate (Chinese +
English time words) followed by LLM extraction for ambiguous cases.
Events created from extraction carry source="post_extract" so the UI
can style them distinctly (R33). LLM gateway is optional to keep the
constructor testable without a live provider.

- src/agentkit/calendar/extraction.py — PostProcessingExtractor
- tests/unit/calendar/test_extraction.py — 13 tests with MockLLMGateway
 2026-06-23 21:56:20 +08:00