This project is an AI-powered retail assistant designed to intelligently respond to customer queries in natural language. It leverages a multi-agent system built with CrewAI, Retrieval-Augmented Generation (RAG) for accessing knowledge bases, and specialized tools for performing specific retail-related tasks. The primary goal is to provide a seamless and efficient customer experience by understanding complex queries and providing comprehensive answers.

🛍️ AI Retail Assistant Project

Features

• Multi-Agent System: Utilizes CrewAI to orchestrate a team of specialized AI agents (Query Router, Retail Assistant, Summarizer).
• Natural Language Understanding: Processes complex, multi-intent user queries.
• Retrieval-Augmented Generation (RAG):
  • Answers frequently asked questions (FAQs) by searching an embedded knowledge base.
  • Searches a product catalog semantically to find relevant items.
• Specialized Tools:
  • Order History Tool: Retrieves customer order details.
  • Return Item Tool: Processes return requests based on defined business logic.
  • FAQ Tool: Provides answers from a pre-defined set of frequently asked questions.
  • Product Search Tool: Finds products based on user descriptions.
• Conversational UI: Built with Streamlit for an interactive chat experience.
• Persistent Knowledge: Uses ChromaDB (via Docker) as a vector store for RAG data.

High-Level Architecture

1. User Interface (Streamlit): Captures user queries.
2. CrewAI Pipeline:
   • Query Router Agent: Analyzes the query and identifies distinct intents.
   • Retail Assistant Agent: Executes specific tasks based on routed intents using specialized tools (Order History, Return Item, FAQ Search, Product Search).
   • Summarizer Agent: Consolidates information from the Retail Assistant Agent and crafts a single, coherent response.
3. Tools & Data Layer:
   • Tools interact with structured data (JSON for orders, CSV for products) and unstructured data (text files for FAQs).
   • RAG tools query ChromaDB, which stores embeddings of FAQs and product information.

Setup and Installation

Prerequisites

• Python 3.10+ (Python 3.12 recommended)
• Pip (Python package installer)
• Docker Desktop (for running ChromaDB)
• An OpenAI API Key

Installation Steps

1. Clone the Repository:

   git clone [https://github.com/neelabhpant/retail-assistant-project.git](https://github.com/neelabhpant/retail-assistant-project.git)
   cd retail-assistant-project

2. Create and Activate a Virtual Environment (Recommended):

   python -m venv venv
   source venv/bin/activate

3. Install Dependencies:

   pip install -r requirements.txt

4. Set Up Environment Variables:

   • Create a .env file in the root directory of the project.
   • Add your OpenAI API key to the .env file

5. Start ChromaDB using Docker:

   • Run the following command in your terminal to start the ChromaDB container:

     docker run -d -p 8000:8000 --name chroma-db chromadb/chroma

   • To make the ChromaDB data persistent across container restarts, use a Docker volume:

     docker run -d -p 8000:8000 -v chroma_data_retail:/chroma/chroma --name chroma-db-persistent chromadb/chroma

6. Embed Data into ChromaDB:

   • Run the embedding scripts from the project root directory. These scripts will populate ChromaDB with your FAQ and product data.

     python -m rag.embed_faqs
     python -m rag.embed_products

▶How to Run

Once the setup is complete, you can run the Streamlit web application:

1. Ensure your virtual environment is activated and the ChromaDB Docker container is running.
2. From the root directory of the project, run:

   streamlit run streamlit_app.py

Future Enhancements (Potential)

• Streaming Responses: Stream agent responses token-by-token in the UI.
• Display Agent Thoughts: Provide users with insights into the agent's decision-making process.
• Action-Oriented Tools: Develop tools that can modify data or trigger external actions (e.g., actually processing a return, sending an email).
• Live Database Integration: Connect tools to real-time databases instead of static files.
• Evaluation Suite: Create a comprehensive set of tests to measure accuracy and performance.
• Deployment: Package the backend as a FastAPI and deploy frontend/backend separately.