- Step1: Define individual AI sub-agents and assign them specific tasks or domains.
- Step2: Create a GraphQL schema representing composite prompt structure and expected fields.
- Step3: Configure the dynamic router with model endpoints and routing rules for each sub-agent.
- Step4: Send composite prompt requests via GraphQL queries to the router.
- Step5: Router splits the query, dispatches to selected LLMs, collects partial responses.
- Step6: GraphQL resolver merges sub-responses into a unified result.
- Step7: Consume the final composite output in your application.