One-off vs Conversational

Agents can be designed for one-off tasks or multi-turn conversations. Understanding when to use each pattern helps you build better AI applications.

Two Patterns

One-off Agents (Invoke)

One-off agents process a single request and return a result. They're stateless — each request is independent.

When to Use

• Data processing and transformation
• Content generation (articles, summaries, code)
• API-like operations
• Batch processing
• Single-step automation

Use Cases

Example

POST /agent/summarizer/invoke
{
  "input": {
    "text": "Long article text here...",
    "maxLength": 100
  }
}

Response:
{
  "output": "Brief summary of the article..."
}

Conversational Agents (Chat)

Conversational agents handle multi-turn dialogues. They receive the full message history and generate contextual responses.

When to Use

• Customer support bots
• AI assistants
• Interactive tutorials
• Guided workflows
• Any dialogue-based interaction

Use Cases

Example

POST /agent/support-bot/chat
{
  "messages": [
    { "role": "user", "content": "I can't log in" },
    { "role": "assistant", "content": "I can help! What error do you see?" },
    { "role": "user", "content": "It says invalid password" }
  ]
}

Response:
{
  "message": {
    "role": "assistant",
    "content": "Try resetting your password at reminix.com/reset"
  }
}

Combining Both

Many agents benefit from both patterns:

You can implement both handlers in a single agent:

Agent("writing-assistant")
  ├── invoke → Generate content
  └── chat   → Discuss ideas

Quick Decision Guide

Need conversation history?
  ├── Yes → Use Chat
  └── No → Is it a single task?
            ├── Yes → Use Invoke
            └── No → Consider both

Next Steps

• Invoke, Chat & Streaming — Technical details
• Python Examples — See implementations
• TypeScript Examples — See implementations