Architecture

ClientCoach AI is built using a modular, multi-agent system architecture designed for flexibility, scalability, and maintainability.

🧱 System Overview

multi-agent/
├── backend/ (FastAPI + Python 3.11+)
│   ├── agents/            # Core AI agents (Customer, Observer, Phase)
│   ├── routers/           # API endpoints
│   ├── services/          # Supabase integration, evaluation logic
│   └── main.py            # API entry point
├── frontend-react/ (React 19 + TypeScript)
│   ├── components/        # UI elements: Chat, Feedback, Scenario Builder
│   ├── lib/               # API clients, speech utilities
│   └── pages/             # Main pages (index.tsx, admin.tsx)

🧠 Core AI Agents

  • Customer Agent: Simulates realistic customer behavior based on system prompts, persona, and scenario context.

  • Phase Agent: Determines when to advance, stay in, or exit a conversation phase based on evaluation outcomes.

  • Observer Agent: Evaluates user performance, checks for critical/optional/red-flag aspects, and generates score and feedback.

🗃️ Backend Stack

  • FastAPI: Handles API requests and orchestrates agent interactions.

  • Supabase: Stores user sessions, scenarios, phase configs, and evaluation logs.

  • Azure OpenAI: Powers GPT-4 for generating customer responses and feedback analysis.

💻 Frontend Stack

  • React 19: Builds a responsive and modern UI.

  • TypeScript: Ensures type safety across frontend components.

  • Azure Speech SDK: Enables text-to-speech and speech-to-text.

  • Tailwind CSS: Used for styling.

PreviousScenario BuilderNextTesting & QA

Last updated