Intelligent Support Assistant: Multi-Agent Orchestration System

A sophisticated AI-powered customer support platform that uses a specialized multi-Assistant architecture to handle complex workflows with precision, context-awareness, and seamless user experience.

System Architecture

This system employs a specialized ensemble of AI models, each engineered for optimal performance in specific tasks:

Assistant	Specialization	Purpose
🤖 Intent Classifier	Natural Language Understanding	Analyzes user queries to determine intent and route to appropriate specialist
✂️ Cancellation Agent	Transaction Execution	Handles complete cancellation workflows with parameter validation
🔄 Rescheduling Agent	Logistics & Scheduling	Manages date/time extraction and booking modifications
❓ Context-Aware Q&A	Knowledge Retrieval	Answers general questions from custom knowledge base

✨ Key Features

Multi-Assistant Architecture: Specialized models for intent classification, cancellation, and rescheduling processes

Intelligent Parameter Validation: Extracts and validates all required parameters before executing API calls

Context-Aware Conversations: Asks targeted clarifying questions when parameters are missing

Pre-Flight Checks: Ensures all data is confirmed before irreversible actions (e.g., cancellations)

Seamless User Experience: Maintains conversation flow through intelligent state management

Technology Stack

• AI Orchestration OpenAI Assistants API with specialized agents
• Backend Framework Python + Flask
• Data Persistence: Firebase Firestore for conversation state management
• Validation: Custom validation layers with parameter verification
• Deployment: Google Cloud Functions, Google Cloud Scheduler
• Authentication: Firebase Auth

Navigate to

http://127.0.0.1:5000

Style guide for Python code

https://peps.python.org/pep-0008/

🔄 Workflow Process

The system follows a rigorous validation workflow:

1. Intent Classification: User message is analyzed to determine purpose

2. Specialist Routing: Conversation is routed to appropriate Assistant

3. Parameter Extraction: Required parameters are identified and extracted

4. Validation Check: All parameters are validated before proceeding

5. Targeted Clarification: Missing parameters trigger specific questions

6. Action Execution: API calls are made only after full validation

7. Confirmation: User receives confirmation of completed action

Installation

1. Install Python 3.11.7:

   • Windows 64-bit
   • macOS

2. Clone the repository:

   git clone <repository-url>
   cd <project-directory>

3. Create and activate a virtual environment (recommended):

   python -m venv venv
   # On Windows:
   .\venv\Scripts\activate
   # On Unix or MacOS:
   source venv/bin/activate

4. Install dependencies:

   pip install -r requirements.txt

5. Set up environment variables: Create a .env file based on the provided example and add your API keys:

   • OpenAI API Key
   • Google Service Account JSON credentials
   • Social Media Platform API tokens

Running the Application Locally

Run the development server:

flask run