Dialogflow Mega Agent

A mega agent enables combining multiple agents into one unified system. The primary advantage addresses scaling constraints: organisations managing thousands of intents can exceed the 2,000-intent platform limit.

A Quick Look at Mega Agents

It's important to remember this feature is in beta. The feature allows assigning one intent per FAQ item for improved accuracy, which previously necessitated complex middleware solutions.

Important limitations: The system operates within a single GCP project with a maximum of 10 sub-agents per mega agent.

Adding a Mega Agent

Implementation is straightforward. When creating new agents, users simply toggle the mega agent setting. Agents then appear in the standard listing interface.

Adding Sub-Agents

After mega agent selection, a Sub Agent button becomes available. Configuration options include selecting specific environments and choosing whether to include knowledge bases. The system provides convenient links to each sub-agent.

Testing Functionality

Initial testing in the Dialogflow simulator revealed IAM permissions issues. The service account linked to the mega agent requires detect intent access across all sub-agents. This requires assigning the mega agent's service account to the sub-agent project with the Dialogflow API Client role. After resolving permissions, the simulator correctly returned responses from configured sub-agents.

Future Development Directions

This advancement is significant for chatbots managing high intent volumes (exceeding 2,000) or requiring multiple teams managing distinct knowledge areas by subject or topic. Key future investigation areas include leveraging contexts between agents and exploring more complex implementation scenarios.

Remaining Challenges

The primary limitation involves training pages functionality, which currently lacks support for sub-agent concept assignment. As a beta feature, additional improvements are anticipated.