Concepts¶

Deep dive into Xians' core concepts and patterns for building production-ready AI agents. This section covers the architectural patterns, workflows, and capabilities that power enterprise agent systems.

Page	Description
SDK Patterns	Understand key SDK design patterns for working with agents and workflows
Agents	Learn about agent registration, workflows, and multi-tenancy architecture
Multitenancy	Explore strategies for securely isolating and serving multiple tenants
Temporal Workflows	Start, execute, and communicate with Temporal-based workflows
[XiansContext]
Replying to User	Master message handling, replies, threads, scopes, and conversational patterns
Proactive Messaging	Initiate conversations, send notifications proactively from any workflow
Web SDK	Build web-based conversational UIs with the official TypeScript client
Knowledge	Manage prompts, instructions, and configuration shared between code and UI
Document DB	Store and query structured JSON data with semantic keys and metadata
Scheduling	Create time-based workflow automation with cron expressions and intervals
Agent-to-Agent (A2A)	Enable workflow communication through messages, signals, queries, and updates
Human-in-the-Loop	Implement approval workflows where agents pause for human review
Webhooks	Integrate external systems using incoming and outgoing HTTP callbacks

A Typical Agent Flow¶

Here's how features combine in a real-world scenario—an autonomous content agent:

sequenceDiagram
    participant S as ⏰ Schedule
    participant W as 🔄 Workflow
    participant K as 💡 Knowledge
    participant D as 📦 Document DB
    participant A2A as 🔗 A2A
    participant H as 👤 Human
    participant M as 💬 Messaging

    S->>W: Trigger daily content check
    W->>K: Fetch system instructions
    W->>D: Get pending content items
    W->>A2A: Send to analyzer agent
    A2A-->>W: Return analysis
    W->>H: Create HITL approval task
    H-->>W: Approve with edits
    W->>M: Notify user of publication
    W->>D: Update content status
    W->>S: Schedule next run

example: "What's happening here?"

Text Only

1. **Schedule** triggers the workflow every morning
2. **Knowledge** provides the agent's instructions and policies
3. **Document DB** stores the content items being processed
4. **A2A** delegates analysis to a specialized agent
5. **HITL** pauses for human approval before publishing
6. **Messaging** notifies stakeholders of the outcome
7. **Schedule** ensures the cycle continues autonomously