Core concepts overview
This chapter is for architects, technical decision-makers, and integration developers who need to understand how the platform works. For a quick start, see Quick start and the User manual.
Concept map
Rendering diagram…
Platform capability map
The Combo Agent baseline is built from these core modules:
| Module | One-line summary |
|---|---|
| Platform architecture | Layers, module relationships, and data flow |
| Agents & multi-agent | Specialized agents, sub-agent delegation, multi-agent coordination |
| Session | Session lifecycle, parallel runs, context compaction |
| Memory | Long-term memory, knowledge base, cross-session semantic search |
| Skill system | Standardized capability packaging and dynamic activation |
| Event trigger | Event-driven session scheduling (webhook / external callbacks / compound triggers) |
| Cron scheduling | One-shot / cron / interval schedules; agents can create jobs |
| Hooks | Lifecycle automation hooks for agents |
| Graft | Reference outputs across parallel sessions |
| Channel messaging | Feishu, WeCom, email, webhook, and other ingress channels |
| Storage & isolation | Local + MinIO dual layer, multi-tenant isolation |
| Multi-tenancy | Tenant / user / session identity and isolation |
| Observability | Traceable, auditable, explainable agent behavior |
Reader navigation
| You are… | Suggested reading order |
|---|---|
| Architect / technical decision-maker | Platform architecture → Agent → Session → Skills |
| Integration developer | Skills → Events → Cron → Hooks → Channels |
| Power user | Session → Memory → Graft → Cron |
| Operations | Storage → Multi-tenancy → Observability → Deploy |