 QihuAI Design
|
bc2899ed09
|
前台8项体验修复: 知识库上传自动刷新; 资产PDF归类; Token驾驶舱媒体统计; AI模糊意图澄清选择; ChatPage增加MCP/附件上传/语音; 技能市场安装态修复+清理演示数据; MCP协议应用改名MCP应用; Agent单应用接入(名称/介绍/API Key)
|
2026-07-15 10:33:43 +08:00 |
|
|
QihuAI Design
|
463824480b
|
fix: 知识库文档上传(尾斜杠+FormData+真文件接收); AI回复强制Markdown; 平台全功能知识库注入系统提示
|
2026-07-14 23:04:30 +08:00 |
|
|
QihuAI Design
|
5d1370156e
|
fix: 停止迁移复活默认模型供应商/模型；AI聊天媒体生成改为异步(修复视频请求失败)
- migrations/002_init_data.sql: 移除 model_providers/models 的 INSERT IGNORE 种子，重启容器不再复活 deepseek/openai/通义千问等默认供应商与模型，改由用户在后台大模型库自行维护
- chat.py: 图像/视频生成分支改为立即返回占位(generating=true)，后台 asyncio 任务轮询 AGNES 并回写消息内容，避免最长 5 分钟同步轮询阻塞 HTTP 请求导致前端「发送失败：请求失败」
- ChatPage.tsx: 发送后若 assistant_message.generating，每 3s 轮询 getMessages 直至内容更新；页面刷新时若仍有「⏳」占位继续轮询；卸载清理定时器
|
2026-07-14 21:48:39 +08:00 |
|
|
QihuAI Design
|
a1669374e6
|
feat(kb): 接入 Dify 知识库，移除 FastGPT；AI 聊天产出+用户文档自动入库量化
- 新增 app/dify_kb.py Dify v1 客户端(连通测试/检索/文本入库)
- kb_router 仅保留 Dify 外部检索分支
- knowledge_admin 去掉 FastGPT/RAGFlow，新增 dify 批量入库端点
- chat.py 注入 Dify RAG 上下文 + 助手回答 fire-and-forget 入库
- 前端 Knowledge.tsx 改为 Dify 单服务 UI(可通讯/可用验证+入库按钮)
- DB: external_kb_services 删除 fastgpt，qihuai 设为 https+connected
|
2026-07-14 20:56:53 +08:00 |
|
|
QihuAI Design
|
d179cc2f02
|
Round6: 前台图片缩略图+下载/视频修复; 后台连接测试单次+自动消失; 导航重组; 存储中文化; 自动化去掉创建按钮; 短信验证码(注册/重置/充值通知)
|
2026-07-14 18:49:27 +08:00 |
|
|
QihuAI Design
|
ba09017884
|
fix: AGNES图像/视频生成强制经智能路由代理(172.18.0.1:7890)，修复容器内直连被重置导致图像生成异常
|
2026-07-14 17:11:12 +08:00 |
|
|
QihuAI Design
|
945eeb8ba0
|
feat: 模型库能力默认模型UI + 前台Auto默认模型 + AGNES直连修复 + 意图识别增强
|
2026-07-14 05:47:57 +08:00 |
|
|
qyyfw
|
ab9cf55321
|
recover: chat.py Auto routing + models.py capability-defaults route (lost in 5d7f32d)
|
2026-07-12 07:12:59 +08:00 |
|
|
qyyfw
|
5d7f32db0a
|
refactor: sync running-container source back to repo for reproducible docker compose build; remove dead code (orphan pipeline.py + 104 stale files from iterative dev) and drop unused app/ mirror dir
|
2026-07-11 06:29:32 +08:00 |
|
|
qyyfw
|
f5968d0f61
|
feat: Auto模式按用户输入意图自动路由视频/图像/语音生成模型
|
2026-07-10 22:04:16 +08:00 |
|
|
qyyfw
|
4c372746d5
|
feat: 补全模型能力选项(11种) + 按能力指定默认模型 + Auto自动模式
|
2026-07-10 20:04:38 +08:00 |
|
|
qyyfw
|
d62848ee98
|
feat(chat): support AGNES image/video models in AI chat - backend _generate_media (images/generations + async videos polling) and frontend MarkdownContent media rendering + model capability badges; nginx proxy_read_timeout raised to 600s for long media generation
|
2026-07-10 15:47:03 +08:00 |
|
|
qyyfw
|
774f1648dc
|
fix(chat/wechat): messages order by created_at,id (sub-second DATETIME(3)); container+MySQL TZ->Asia/Shanghai; correct UTC timestamps to CST; rebuild WeXin Bot history
|
2026-07-10 10:50:06 +08:00 |
|
|
qyyfw
|
7b7871a14d
|
feat(im): mirror WeChat bot chats into AI chat as WeXin Bot conversation; add conversations.source; migrate history; remove redundant history page
|
2026-07-10 10:09:49 +08:00 |
|
|
qyyfw
|
5d1ad26874
|
fix: skills variable init + ChatPage show all published apps + populate skill content
|
2026-07-03 12:10:37 +08:00 |
|
|
qyyfw
|
5d9d420525
|
feat: Phase 2.5 - Skill Execution Pipeline + Chat Workflow Enhancement - Stage A: MarketApp->Skill ID mapping, real skill execution engine, SkillVersion snapshot fix - Stage B: AgentConfig dynamic injection, multi-turn context persistence, Function Calling loop (5 rounds) - Stage C: KnowledgeDocumentChunk model, BM25 text retrieval, smart KB context injection - D: PipelineScheduler LLM-based skill pattern extraction replacing mock
|
2026-07-03 10:37:02 +08:00 |
|
|
qyyfw
|
8ec808b515
|
fix: 6 bug fixes - timezone CST, knowledge_ids mismatch, skill uninstall, dashboard stats, chat knowledge retrieval, skill selection
|
2026-07-03 09:18:04 +08:00 |
|
|
qyyfw
|
78ff261b5a
|
初始提交: Qihu-AIFlow v2.0.1 完整项目
- 前端 React+TypeScript (源码+构建产物)
- 后端 Python FastAPI
- 数据库迁移 001-015
- Docker部署配置
- 微信ClawBot API频道绑定
- 项目文档 README.md
|
2026-07-02 14:40:46 +08:00 |
|