Internal technical demonstration
AI Meta App Hub — Vertical Multi-Agent Orchestration Platform
AI Meta App Hub is a vertical-industry AI agent platform that ingests structured, unstructured, real-time and batch data — including files and IoT streams — and lets operators configure multiple agents with defined skills and tools.
Technology stack
- Java 8
- Spring Boot 2.7
- Spring Cloud Alibaba
- MyBatis-Plus
- React 19
- TypeScript
- Vite
- Ant Design
- Docker Compose
- Nginx
- Nacos
- Sentinel
- MySQL
- Redis
- MinIO
- Nexus
- Milvus
- etcd
- Attu
- NebulaGraph
Challenge
Vertical businesses need one platform where a single spoken instruction can trigger complex, multi-step task orchestration — data reporting, OA approval workflows — across heterogeneous data sources, instead of stitching together separate reporting tools, approval systems and data connectors.
Approach
A Spring Cloud Alibaba microservice backend (Nacos for registry/config, Sentinel for traffic control) exposes agent, skill and tool configuration through a React 19 + Ant Design console; Milvus (vector search) and NebulaGraph (graph database) back retrieval and relationship reasoning over ingested data, with MinIO holding files.
Architecture
Microservices behind Nginx, containerized with Docker Compose; Nacos and Sentinel handle service registry/config and traffic control; Milvus and NebulaGraph form the retrieval/reasoning layer the agents draw on.
Outcome
A working multi-agent platform where a single voice instruction can drive task orchestration, data reporting and OA approval flows across connected data sources.