aiagent/misc/issues_open.json

[{"id":84,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/66","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/66","number":66,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 知识进化仪表盘 — 知识图谱/进化历史/效果度量","body":"## 目标\n前端页面：可视化展示知识自进化的全过程，让用户看到平台在变聪明。\n\n## 页面模块\n\n### 1. 知识增长曲线\n- 知识条目数量随时间增长\n- 按类别分布（bug_fix/best_practice/optimization）\n- 知识被检索次数排行\n\n### 2. 进化历史时间线\n- 展示每次自动优化：瓶颈发现 → 优化方案 → 应用 → 效果对比\n- 展示每次策略调整：反馈触发 → 策略更新 → 下次准确率\n\n### 3. 效果度量\n- Agent 成功率趋势（进化前 vs 进化后）\n- 平均响应时间趋势\n- 用户满意度趋势\n- 知识复用率\n\n### 4. 知识图谱\n- D3/vis.js 可视化知识之间的关系\n- 节点：知识条目，边：引用/相似/演化关系\n\n## 新增文件\n- frontend/src/views/KnowledgeEvolution.vue\n- frontend/src/stores/knowledge.ts\n- 路由：/knowledge-evolution","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:20+08:00","updated_at":"2026-05-10T15:32:20+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":83,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/65","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/65","number":65,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 新工具自动发现与集成 — 监控市场，自动学习接入","body":"## 目标\n当平台新增工具或外部工具市场发布新工具时，系统自动发现、评估、集成。\n\n## 发现渠道\n- 平台内：工具市场上新（已有 tools 表）\n- 外部：GitHub trending / MCP 工具市场 / API 发布\n- 用户请求：\"能不能接一个XX工具\"（计入需求池）\n\n## 自动评估\n- 读取工具描述/API 文档\n- LLM 分析：这个工具能解决平台当前哪些痛点？\n- 匹配度评分（结合平台现有 Agent/工作流场景）\n- 生成集成方案（需要的 adapter、参数映射）\n\n## 自动集成\n- 匹配度 \u003e 80%：自动生成 tool adapter 代码\n- 生成测试用例\n- 注册为 builtin tool\n- 通知管理员确认后上线\n\n## 新增文件\n- backend/app/services/tool_discovery.py\n- backend/app/tools/auto_adapter.py\n- Celery 定时：每6小时扫描一次","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:20+08:00","updated_at":"2026-05-10T15:32:20+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":82,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/64","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/64","number":64,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 用户反馈学习 — 点踩/修改后自动调整 Agent 策略","body":"## 目标\n用户对 Agent 输出点踩或手动修改后，系统自动学习，不在同样场景犯同样错误。\n\n## 反馈信号采集\n- 点踩（thumbs down）\n- 手动修改 Agent 输出后保存\n- 用户说\"不对\"\"重新来\"\"换个方式\"\n- 同一问题反复问（暗示上次回答不满）\n- Task 被驳回（approval rejected）\n\n## 学习机制\n- 收集 (输入, 输出, 反馈) 三元组\n- 聚类相似场景（embedding 相似度）\n- 生成策略调整：避免产生类似被否定的输出\n- 更新 Agent system prompt（追加反例）\n- 或在知识库中记录\"这种场景不要这样做\"\n\n## 策略调整方式\n- 追加 negative examples 到 prompt\n- 调整 Agent 的 temperature（踩多→降温）\n- 替换使用的工具链\n- 推荐更换模型\n\n## 新增文件\n- backend/app/services/feedback_learner.py\n- backend/app/models/feedback_record.py\n- Celery 定时：每天汇总反馈，生成策略调整建议","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:19+08:00","updated_at":"2026-05-10T15:32:19+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":81,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/63","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/63","number":63,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 工作流自动优化 — 瓶颈节点自动调整 DAG 结构","body":"## 目标\n检测到瓶颈后，系统自动生成优化方案供用户确认。\n\n## 优化策略\n| 瓶颈类型 | 自动优化方案 |\n|---------|------------|\n| 单节点耗时长 | 拆分为多个并行子节点 |\n| 频繁失败 | 增加前置校验节点 + 重试逻辑 |\n| LLM Token 过高 | 增加摘要/压缩节点 |\n| 依赖链过长 | 分析可并行化的节点，重组 DAG |\n| 空转等待 | 增加缓存节点，跳过重复计算 |\n\n## 工作流\n1. bottleneck_detector 识别瓶颈\n2. optimization_engine 生成优化方案（DAG diff）\n3. 通过飞书/站内通知用户确认\n4. 用户确认后自动应用（创建新版本工作流）\n5. A/B 对比新旧工作流效果\n\n## 新增文件\n- backend/app/services/workflow_optimizer.py\n- 前端：优化方案预览页（DAG 对比视图）","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:19+08:00","updated_at":"2026-05-10T15:32:19+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":80,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/62","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/62","number":62,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 工作流瓶颈自动检测 — 统计分析节点耗时/失败率","body":"## 目标\n自动分析工作流执行数据，发现经常卡住的节点，给出优化建议。\n\n## 检测指标\n- 节点平均耗时（P50/P95/P99）\n- 节点失败率\n- 节点重试率\n- 节点间等待时间\n- 资源消耗（LLM token 用量、工具调用次数）\n\n## 瓶颈判定规则\n- 某节点 P95 耗时 \u003e 所有节点平均 P95 的 3 倍 → 瓶颈节点\n- 某节点失败率 \u003e 20% → 问题节点\n- 某节点工具调用次数 \u003e 10 → 效率低下\n\n## 输出\n- 瓶颈节点列表（按严重程度排序）\n- 优化建议（拆分子任务 / 更换模型 / 增加超时 / 添加重试）\n\n## 新增文件\n- backend/app/services/bottleneck_detector.py\n- Celery 定时：每天凌晨分析昨日数据\n- 前端：工作流详情页新增瓶颈分析 tab","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:18+08:00","updated_at":"2026-05-10T15:32:18+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":79,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/61","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/61","number":61,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 知识检索增强 — Agent 执行前自动检索相关知识","body":"## 目标\n每个 Agent 执行任务前，自动从知识库检索相关经验，注入到 system prompt 或 context 中。\n\n## 实现方案\n- AgentRuntime.run() 执行前触发 pre_execute hook\n- 用用户输入 + 任务描述做语义检索（RAG）\n- 检索到相关知识条目后，拼接到 system prompt 末尾\n- 检索结果附带相似度分数，低于阈值的不注入\n\n## 配置项\n- knowledge_retrieval_enabled: bool（默认true）\n- knowledge_top_k: int（默认3）\n- knowledge_min_score: float（默认0.7）\n\n## 修改文件\n- backend/app/agent_runtime/core.py — 添加 pre_execute hook\n- backend/app/services/knowledge_retriever.py — 向量检索服务\n- Agent 配置页新增知识增强开关","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:18+08:00","updated_at":"2026-05-10T15:32:18+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":78,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/60","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/60","number":60,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] 经验自动沉淀管道 — LLM 提取关键经验 → RAG 入库","body":"## 目标\n从执行日志中自动提取可复用的知识，让一个 Agent 踩过的坑所有 Agent 都能避开。\n\n## 管道流程\nAgent执行日志 → 筛选高质量执行（success=True + user_feedback\u003e=4）→ LLM 提取经验（做了什么、怎么做的、为什么成功、注意事项）→ 结构化知识条目 → Embedding → RAG 向量库 → 原始日志关联（可回溯）\n\n## 知识条目结构\n- title: 一句话标题（如\"处理MySQL 死锁的重试策略\"）\n- category: bug_fix / best_practice / workaround / optimization\n- tags: [\"mysql\", \"deadlock\", \"retry\"]\n- situation: 什么场景下遇到的问题\n- solution: 怎么解决的\n- caveats: 注意事项\n- source_execution_ids: [原始执行ID]\n\n## 新增文件\n- backend/app/services/knowledge_extractor.py\n- backend/app/models/knowledge_entry.py\n- Celery 定时任务：每小时运行一次提取管道","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:18+08:00","updated_at":"2026-05-10T15:32:18+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":77,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/59","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/59","number":59,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[知识进化] Agent 执行日志结构化提取","body":"## 目标\n将每个 Agent 的每次执行结构化为可分析的数据。\n\n## 结构化字段\n- agent_id / agent_name / agent_version\n- goal_id / task_id（关联上下文）\n- input: 用户输入完整文本\n- output: Agent 输出完整文本\n- tool_chain: [{tool_name, input, output, duration_ms}]\n- llm_calls: [{model, prompt_tokens, completion_tokens, duration_ms}]\n- success: bool\n- error_message: str\n- user_feedback: {rating, comment}\n- latency_ms: 总耗时\n\n## 技术方案\n- AgentRuntime.run() 返回值扩展为 ExecutionLog\n- 写入独立表 agent_execution_logs\n- 异步写入（Celery task）\n\n## 新增文件\n- backend/app/models/agent_execution_log.py\n- backend/app/services/execution_logger.py\n- Alembic 迁移","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:32:18+08:00","updated_at":"2026-05-10T15:32:18+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":76,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/58","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/58","number":58,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 数字分身管理页面 — 创建/训练/授权/审计","body":"## 目标\n前端页面：用户管理自己的数字分身，查看训练进度，配置授权等级，审计决策历史。\n\n## 页面功能\n### 1. 分身概览\n- 当前阶段（观察/影子/代理/自主）\n- 各场景准确率卡片\n- 行为数据统计（观察了多少条数据）\n\n### 2. 训练中心\n- 启动/暂停行为学习\n- 影子模式开关\n- 场景解锁状态\n\n### 3. 授权配置\n- L0-L4 授权矩阵配置\n- 场景化授权规则\n\n### 4. 审计日志\n- 所有代理决策列表\n- 用户确认/驳回记录\n- 决策回溯\n\n## 新增文件\n- frontend/src/views/DigitalTwin.vue\n- frontend/src/stores/digital-twin.ts\n- 路由：/digital-twin","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:07+08:00","updated_at":"2026-05-10T15:28:07+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":75,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/57","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/57","number":57,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 决策授权体系 — 低风险自动执行，高风险举手","body":"## 目标\n建立分层授权体系，让数字分身在合适的风险等级下自主决策。\n\n## 授权等级\n| 等级 | 风险 | 示例 | 需要确认 |\n|------|------|------|---------|\n| L0 完全自主 | 零风险 | 格式化代码、修正拼写 | 否 |\n| L1 自动+通知 | 低风险 | 合并小PR、回复常规邮件 | 否（通知） |\n| L2 建议+确认 | 中风险 | 修改API接口、调整配置 | 是（一键确认） |\n| L3 举手 | 高风险 | 修改数据库schema、涉及资金 | 是（详细审批） |\n| L4 禁止 | 极高风险 | 删除生产数据、修改权限 | 始终禁止 |\n\n## 技术方案\n- 每个操作预定义风险等级（可配置）\n- 数字分身执行前检查授权矩阵\n- 需要确认时通过飞书/站内发送确认卡片\n- 所有决策记录审计日志\n\n## 新增文件\n- backend/app/models/auth_matrix.py\n- backend/app/services/auth_gate.py","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:07+08:00","updated_at":"2026-05-10T15:28:07+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":74,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/56","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/56","number":56,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 需求评估代理 — 拉历史数据给工时预估","body":"## 目标\n用户说\"这个需求评估一下\"，数字分身拉取历史类似需求数据，给出工时预估和风险提示。\n\n## 输入\n- 需求描述（自然语言）\n- 需求类型（功能/修复/优化）\n- 涉及模块\n\n## 输出\n- 预估工时（基于历史同类需求的P50/P90）\n- 历史类似需求列表及实际工时\n- 风险提示（涉及模块复杂度、依赖方数量）\n- 建议方案（分阶段/简化方案/参考实现）\n\n## 技术方案\n- 从 Gitea Issue / 执行历史中提取需求→工时映射\n- LLM 语义匹配需求相似度\n- 统计模型给出工时区间\n\n## 新增文件\n- backend/app/services/requirement_estimator.py\n- 前端：需求评估输入页","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:06+08:00","updated_at":"2026-05-10T15:28:06+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":73,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/55","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/55","number":55,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 代码评审代理 — \"这个PR你看下\" 自主审核","body":"## 目标\n用户说\"这个PR你看下\"，数字分身自动完成代码审核，标注问题，给出合入建议。\n\n## 输入\n- PR URL (Gitea API)\n- PR diff 内容\n- 关联 Issue 上下文\n\n## 输出\n- 审核意见列表（按严重程度排序）\n  - 阻塞：安全漏洞、数据丢失风险\n  - 建议：性能问题、可读性改进\n  - 赞赏：好的实现模式\n- 合入建议：Approve / Request Changes / Comment\n- 审核摘要（3句话）\n\n## 技术方案\n- 利用用户历史审核指纹（偏好权重）指导审核重点\n- 使用 parse_test_result_file + create_gitea_issue 工具链\n- 审核结果作为 PR comment 回写 Gitea\n\n## 新增文件\n- backend/app/services/code_review_agent.py\n- backend/app/tools/code_review_tools.py\n- 前端：PR审核页面","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:06+08:00","updated_at":"2026-05-10T15:28:06+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":72,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/54","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/54","number":54,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 影子模式执行引擎 — 生成建议不执行，对比人类决策","body":"## 目标\n在代理执行之前，先以影子模式运行：生成建议但不自动执行，与人类实际决策对比，计算准确率。\n\n## 核心逻辑\n用户触发操作 → 数字分身同步生成建议 → 建议仅对用户可见（影子）→ 用户实际操作 vs 影子建议 → 对比记录 → 准确率达标（如 \u003e90%）→ 解锁代理模式\n\n## 影子模式场景\n| 场景 | 影子动作 | 对比维度 |\n|------|---------|---------|\n| PR审核 | 生成审核意见 | 相同问题是否被指出 |\n| 邮件回复 | 草拟回复 | 意图是否一致 |\n| 需求评估 | 给出工时 | 偏差是否在20%内 |\n\n## 准确率仪表\n- 每个场景独立计算影子准确率\n- 准确率 \u003e 90% 持续 2 周 → 自动解锁代理模式\n- 准确率 \u003c 70% → 回退到观察模式\n\n## 新增文件\n- backend/app/services/shadow_executor.py\n- backend/app/models/shadow_comparison.py\n- 前端：数字分身页面新增影子准确率卡片","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:05+08:00","updated_at":"2026-05-10T15:28:05+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":71,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/53","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/53","number":53,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 行为指纹模型 — 向量化 + 偏好权重学习","body":"## 目标\n将原始行为日志转化为可量化的用户\"数字指纹\"。\n\n## 技术方案\n\n### 1. 行为序列向量化\n- 使用 embedding 模型将行为序列转为向量\n- 行为：{action, context, decision, timestamp} → sentence → embedding\n- 存储到向量数据库（pgvector 或 Qdrant）\n\n### 2. 偏好权重模型\n- 从决策历史中提取用户偏好权重\n- 示例：PR审核权重 = {security: 0.4, performance: 0.3, readability: 0.2, style: 0.1}\n- 使用 Bradley-Terry 模型或简单频率统计\n\n### 3. 决策规则提取\n- 从历史决策中提取 if-then 规则\n- 示例：IF 改动文件\u003e10 AND 无测试 THEN 要求补充测试\n\n## 新增文件\n- backend/app/models/user_fingerprint.py\n- backend/app/services/fingerprint_engine.py\n- 定时任务：每周更新指纹模型","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:28:05+08:00","updated_at":"2026-05-10T15:28:05+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":70,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/52","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/52","number":52,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 合同/文档审查代理 — 对照历史，标记风险条款","body":"## 目标\n用户上传合同或文档，数字分身对照历史审查记录，自动标记风险条款和异常变更。\n\n## 输入\n- 合同/文档文件（PDF/Word）\n- 审查上下文（合同类型、对手方、金额范围）\n\n## 输出\n- 风险条款列表（按风险等级排序）\n  - 🔴 高风险：违约责任不对等、无限担保\n  - 🟡 中风险：付款条件模糊、验收标准不明确\n  - 🟢 低风险：格式问题、标点错误\n- 对比历史同类合同的差异\n- 修改建议（引用历史合同中的条款作为参考）\n\n## 技术方案\n- PDF/Word 文本提取（PyMuPDF + python-docx）\n- 条款级 diff（对比历史合同库）\n- 风险规则库（用户可配置）\n- 结合用户历史审查偏好\n\n## 新增文件\n- `backend/app/services/document_review_agent.py`\n- `backend/app/tools/document_tools.py` — `extract_pdf`, `extract_docx`, `compare_clauses`","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:26:59+08:00","updated_at":"2026-05-10T15:26:59+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0},{"id":66,"url":"http://101.43.95.130:3001/api/v1/repos/admin/aiagent/issues/48","html_url":"http://101.43.95.130:3001/admin/aiagent/issues/48","number":48,"user":{"id":1,"login":"admin","login_name":"","source_id":0,"full_name":"","email":"263303411@qq.com","avatar_url":"http://101.43.95.130:3001/avatars/9f89f41683d9730fc0abe566f120891a","html_url":"http://101.43.95.130:3001/admin","language":"zh-CN","is_admin":true,"last_login":"2026-04-30T00:13:32+08:00","created":"2025-12-19T11:20:47+08:00","restricted":false,"active":true,"prohibit_login":false,"location":"","website":"","description":"","visibility":"public","followers_count":0,"following_count":0,"starred_repos_count":0,"username":"admin"},"original_author":"","original_author_id":0,"title":"[数字分身] 用户行为数据采集器 — 四大观察维度","body":"## 目标\n让天工观察用户的日常工作模式，建立行为基线。\n\n## 四大观察维度\n\n### 1. 邮件处理模式\n- 采集：邮件分类逻辑（哪些先回、哪些已读不回、哪些转发）\n- 采集：回复风格（正式/简洁/模板化）\n- 采集：响应时间分布（紧急30分钟、普通当天、低优先级3天）\n\n### 2. 代码审核模式\n- 采集：PR 审核关注点（安全 \u003e 性能 \u003e 可读性？）\n- 采集：审核评论风格（直接修改/提建议/问问题）\n- 采集：合入标准（测试覆盖率/CI通过/至少1个approve）\n\n### 3. 文档撰写风格\n- 采集：文档结构偏好（先结论后分析/先背景后方案）\n- 采集：语言风格指纹（用词习惯、句式长度、图表偏好）\n\n### 4. 决策历史关联\n- 采集：决策输入（什么信息触发决策）→ 决策过程（查了哪些数据）→ 决策结果\n- 构建决策依赖图\n\n## 技术方案\n- 新增 `UserBehaviorLog` 模型：user_id, category, action, context, result, timestamp\n- 各业务端点埋点采集（非侵入式 middleware）\n- 飞书/邮件/PR 通过 webhook 回调写入\n\n## 产出\n- `backend/app/models/user_behavior.py` — 行为日志模型\n- `backend/app/services/behavior_collector.py` — 采集服务\n- Alembic 迁移","ref":"","assets":[],"labels":[],"milestone":null,"assignee":null,"assignees":null,"state":"open","is_locked":false,"comments":0,"created_at":"2026-05-10T15:25:10+08:00","updated_at":"2026-05-10T15:25:10+08:00","closed_at":null,"due_date":null,"time_estimate":0,"pull_request":null,"repository":{"id":27,"name":"aiagent","owner":"admin","full_name":"admin/aiagent"},"pin_order":0}]