Transform your stateless agent logic into a backend service. Applying Factors 5, 6, 7, and 11, this course shows how to decouple agent execution from user interfaces. Persist unified state in a relational database, orchestrate runs through REST endpoints, control Launch/Pause/Resume, and connect autonomous execution with human intervention through tool calls.
Overview
Syllabus
Exposing Agents via API
- Launch an Agent Workflow with a REST Endpoint
- Run the Agent Launch Endpoint in the Background
- Add the Agent State Retrieval Endpoint
Persisting States with Databases and Callbacks
- Persist the Initial Agent State to SQLite
- Read Agent State from SQLite
- Connect Background Agent Execution to SQLite
- Add Progress Callbacks for Durable Agent State
Pausing and Resuming Agents Through API Calls
- Teach the Progress Callback to Respect Pause Signals
- Add the Pause Endpoint
- Add Resume Support to the Agent API
Integrating Human Input Back to Agents
- Implement Human Escalation with ask_human
- Provide Human Input and Resume the Agent
Triggering Agents from Anywhere
- Your Agent Meets the Real World
