A plugin-based AI operations console for managing bots, knowledge, plugins, and message platforms in group chats and multi-platform environments.

Stars

12

7-day growth

No data

Forks

5

Open issues

0

License

Apache-2.0

Last updated

2026-07-28

AI repository intelligence
FR-AI / ANALYSIS

Why it is worth attention

It offers a unified, modular control plane for message platform adapters, LLM configurations, group participation strategies, knowledge (FAQ/RAG), member memory, plugin hooks, and operational reliability (Redis Streams, Outbox, DLQ), all accessible through a React frontend and FastAPI backend with Docker Compose quick-start.

Who it is for

  • Teams operating group-chat bots across multiple messaging platforms
  • Developers building AI assistants with customizable knowledge and plugin ecosystems
  • DevOps engineers needing observable, recoverable messaging infrastructure
  • Enterprise users seeking an on-premises multi-tenant AI operations dashboard

Use cases

  • Automating FAQ responses in WeChat groups with a configured LLM and FAQ knowledge base
  • Managing a single bot personality across WeChat, Telegram, and other adapters via a central console
  • Running a group moderation system with point-based rewards, content filtering, and scheduled reports
  • Testing and debugging bot behavior with a built-in link-test tool before going live

Strengths

  • Plug-in SPI enables easy addition of new message platforms beyond the built-in WeChat adapter
  • Production-grade reliability built in with Redis Streams, transactional Outbox, retries, and DLQ replay
  • Full-stack observability: OpenTelemetry, metrics, logging, traces, and Grafana dashboards included
  • Self-contained Docker Compose stack with a fake LLM for offline development and testing

Considerations

  • Only a WeChat adapter is provided out-of-the-box; other platforms require implementing the adapter SPI
  • Knowledge features (FAQ/RAG) require an additional Qdrant deployment and embedding provider configuration
  • Production deployment demands careful secret management and independent .env.production setup

README quick start

Agent Console

面向群聊与多消息平台的插件化 AI 运营控制台。

统一管理消息平台连接、LLM、群参与策略、知识与成员记忆、插件能力、消息链路和失败恢复。当前内置 WeChat SDK 适配器,但核心平台不依赖微信,可以独立启动,也可以通过适配器 SPI 扩展其他消息平台。

快速体验 · 核心能力 · 系统架构 · 源码开发 · 部署文档

核心能力

领域能力
消息接入适配器目录、平台连接、健康探测、会话与参与者同步、租户与群聊作用域
机器人行为LLM 配置、群参与策略、回复风格、复读策略、行为模拟与链路测试
知识与记忆FAQ、文档知识库、混合 RAG、成员记忆、人物档案、群聊关系图
Agent 与插件多轮工具调用、插件 Hook、命令中心、插件安装/启停/升级与作用域策略
群运营积分、内容审核、主动参与、画图、高德地图和微信管理工具
运维与恢复Redis Streams、事务 Outbox、重试、消息队列、DLQ 重放、指标、日志与追踪

管理台按“上线向导、消息接入、机器人行为、知识、群运营、能力与集成、系统运维”组织功能,常见操作无需直接调用 API。

基本概念

  • 适配器(Adapter):让平台能够连接 WeChat 或其他消息提供商的可安装代码。
  • 连接(Connection):租户实际配置的 SDK、网关或账号实例。安装适配器不等于已经建立连接。
  • 会话(Conversation):连接下的群、频道或私聊范围。
  • 参与者(Participant):会话中的消息平台身份。

连接的非敏感配置在控制台“平台连接”页面维护;凭据只保存在部署侧的 Secret Provider 中,控制面仅保存 secret_ref

系统架构

flowchart LR
    P["消息平台"] --> A["适配器与连接"]
    A --> I["Ingress API验签 · 幂等 · 限流"]
    I --> B["Redis Streams"]
    B --> W["Inbound Worker"]
    W --> O["消息编排器"]
    O --> C["FAQ / RAG / LLM / Agent"]
    O  X["插件 Hooks 与工具"]
    C --> E["Outbox / Effect Intent"]
    E --> D["Outbound Worker"]
    D --> P

    UI["React 管理控制台"] --> API["FastAPI 管理 API"]
    API --> DB[(PostgreSQL)]
    O --> DB
    C -. 可选 .-> Q[(Qdrant)]

核心 API 与 inbound、outbound、scheduler worker 分进程运行。PostgreSQL 保存持久状态,Redis 承担队列、热状态与协调,Qdrant 仅在知识/向量能力启用时需要。

快速体验

推荐使用 Docker Compose。你只需要 Docker Desktop(Windows/macOS)或 Docker Engine + Compose v2(Linux)。默认配置使用离线 fake LLM,未配置 API Key 也能启动控制台并验证链路。

Windows

.\scripts\windows-stack.ps1 start

脚本会在缺少 .env 时从 .env.example 创建一份本地开发配置,并启动完整核心栈。

macOS / Linux

cp .env.example .env
python3 -  |
| API |  |
| API 文档 |  |
| 本地演示管理员令牌 | 上一步生成并写入 `.env` 的随机令牌 |

Windows 启动脚本会完成同样的生成并显示管理员令牌。三个管理侧
secret 均独立生成且只保存在被 Git 忽略的 `.env` 中;不要提交、共享或复用于生产。

默认 `app` profile 会启动 PostgreSQL、Redis、OpenTelemetry Collector、数据库迁移任务、API、三个核心 worker 和前端;不会启动 Qdrant 或微信 bridge,知识向量能力默认关闭。

常用命令:

```bash
docker compose --profile app ps
docker compose --profile app logs -f --tail=100
docker compose --profile app stop
docker compose --profile app down

Windows 也可以使用:

.\scripts\windows-stack.ps1 status
.\scripts\windows-stack.ps1 health
.\scripts\windows-stack.ps1 logs
.\scripts\windows-stack.ps1 stop

down 会保留命名卷中的数据。不要

Related repositories

Similar projects matched by category, topics, and programming language.

lopopolo
Featured
lopopolo GitHub avatar

harness-engineering

Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.

AI & Machine LearningAI Agents
2,390
MoonshotAI
Featured
MoonshotAI GitHub avatar

Kimi-K3

Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.

AI & Machine LearningAI Agents
3,348
mshumer
Featured
mshumer GitHub avatar

Claude-of-Duty

A browser-based first-person shooter built entirely with procedural generation and orchestrated AI agents, featuring 55k lines of Three.js/WebGL2 code and no art assets.

AI & Machine LearningAI Agents
1,234