Sales Agent using Google ADK

Project Overview

This project implements a conversational Sales Agent using Google's Agent Development Kit (ADK) in Python.

The agent:

• Activates when a new lead sends a message, simulating a form submission.
• Asks for consent before collecting any information.
• Collects specific information through a step-by-step flow:
  • Age
  • Country
  • Interested Product or Service
• Handles multiple lead interactions concurrently by managing separate sessions using session IDs.
• Saves all collected information into a CSV file (leads.csv).
• Sends a follow-up message if a lead does not respond within 10 seconds, simulating a 24-hour delay.

Project Structure

sales-agent/
├── adk-env/                  # Virtual environment
├── agent.py                  # SalesAgent class logic
├── utils.py                  # Helper function to update leads.csv
├── simulate_leads.py          # Script to automatically simulate lead interaction
├── interactive_test.py        # Script to manually interact with the agent
├── leads.csv                  # Output file where leads are saved
├── README.md                  # Project documentation

Setup Instructions

1. Clone or download the project files.
2. Create a virtual environment:

   python -m venv adk-env
   

3. Activate the virtual environment:

   adk-env\Scripts\activate
   

4. Install required dependencies:

   pip install google-adk
   pip install aioconsole
   

How It Works

• External Trigger:
  The agent is activated when a lead sends a message, simulating a form submission.

• Conversational Flow:

  1. The agent greets the user and asks for consent.
  2. If the user agrees, the agent collects information step-by-step.
  3. The collected data is saved in the leads.csv file.

• Multi-Session Handling:
  Each conversation session is handled independently using a unique session ID.

• Follow-Up Handling:
  If the user does not respond within 10 seconds after a question, the agent sends an automatic follow-up reminder.

How to Test

Simulate Leads Automatically

Run the simulation script:

python simulate_leads.py

This script automatically interacts with the agent to simulate lead form filling.

Manually Interact with the Agent

Run the manual testing script:

python interactive_test.py

Type your responses manually. If you do not respond to a question within 10 seconds, the agent will automatically send a follow-up message.

Leads Data Example

After the interaction, the leads.csv file will store lead data in the following format:

lead_id	age	country	interest	status
manual_lead	25	Pakistan	Nike Shoes	secured

Requirements Fulfilled

Task Requirement	Status
Use Google ADK in Python	Completed
Handle multiple lead interactions concurrently	Completed
Collect specific information step-by-step	Completed
Activate agent through an external trigger	Completed
Save collected information into a CSV file	Completed
Follow up with leads who do not respond	Completed
Manual and simulation-based testing	Completed

Conclusion

This project successfully implements a fully functional conversational sales agent that interacts with leads, collects necessary information, saves the data securely, and manages follow-up messaging in case of inactivity, all built using Google's Agent Development Kit in Python.