Echo AI

Multi-agent voice assistant backend using Google ADK agentic approaches.

Frontend: https://echo-frontend-406300707114.us-central1.run.app

Backend: https://echo-backend-406300707114.us-central1.run.app

hosted and deployed in Cloud Run Services

Architecture

The backend uses a declarative workflow engine for agentic orchestration:

• WorkflowEngine: Executes agents in a declarative, conditional workflow
• IntentAgent: Classifies user queries using Gemini
• CalendarAgent: Retrieves calendar events from Google Calendar API
• VoiceAgent: Converts text to speech using ElevenLabs and stores in GCS
• OrchestratorService: Defines and executes multi-agent workflows

Workflow Flow

1. User query → IntentAgent (always runs)
2. Conditional routing based on intent
3. Action agent (CalendarAgent, EmailAgent, etc.) if applicable
4. VoiceAgent generates TTS if response text exists