FashionAgent by himanishbajaj0127 - MCP Server

FashionAgent: AI-Powered Fashion Retail Forecasting

📖 Introduction

FashionAgent is an AI-driven solution designed to enhance forecasting and decision-making in the fashion retail industry. By integrating various data sources, it provides actionable insights to optimize inventory management and sales strategies.

🛠️ Setup Instructions

1. Clone the Repository

git clone https://github.com/yourusername/FashionAgent.git
cd FashionAgent_HybridSetup_Prod_Repo

2. Create a Virtual Environment

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

3. Configure Environment Variables

Edit run_mcp.sh to set your Salesforce, Azure SQL, Snowflake, and Slack credentials. The script includes placeholders for the following environment variables:

SALESFORCE_USERNAME: Your Salesforce username
SALESFORCE_PASSWORD: Your Salesforce password
SALESFORCE_TOKEN: Your Salesforce security token
AZURESQL_CONN: Your Azure SQL connection string
SNOWFLAKE_CONN: Your Snowflake connection string (format: "user=username password=password account=account warehouse=warehouse database=database schema=schema")
SLACK_WEBHOOK: Your Slack webhook URL
MCP_URL: MCP server URL (default: "http://localhost:8000")
MCP_API_KEY: Optional API key for MCP authentication
SF_OBJECT: Salesforce object to query (default: "Opportunity")
EXCEL_FOLDER: Path to Excel files (default: "/dbfs/mnt/excel_drop")
HORIZON: Forecast horizon in days (default: 28)

4. Run the MCP Server

bash run_mcp.sh

This will start a FastAPI MCP server at http://localhost:8000.

5. Upload to Databricks

Go to Databricks CE
Create a cluster (Python 3.x, small node)
Import orchestrator.py
Attach requirements.txt as a library
Run the job

6. View Forecasts & Alerts

Forecast outputs are stored in Delta tables (Databricks).
Slack alerts are sent to the configured channel.

📊 Example Workflow

Fetch CRM data from Salesforce Sandbox.
Load inventory sheets from Excel (Google Drive/local).
Query sales transactions from Azure SQL.
Fetch warehouse stock from Snowflake.
Merge data into a single feature store.
Run demand_forecaster.py to predict demand.
Compare demand vs. stock.
Trigger a Slack alert if stock shortage is predicted.

🧑‍💻 Tech Stack

Databricks (Spark Orchestration)
MCP Protocol (local FastAPI server)
Salesforce Sandbox (CRM)
Azure SQL, Snowflake, Excel (data sources)
PySpark, Pandas, MLlib (forecasting)
Slack API (alerts)

📌 Next Steps

Extend MCP to support tool manifests for external apps.
Add real-time API triggers instead of batch processing.
Deploy the MCP server on Azure App Service for scalability.

✅ With this setup, you can demo an enterprise-grade AI Agent that resonates with fashion industry recruiters and product-based companies.

📁 Project Structure

FashionAgent/
├── README.md                 # Project documentation
├── requirements.txt          # Python dependencies
├── agent/                    # Main agent module
│   ├── agent.py             # Main agent class and logic
│   ├── config.py            # Configuration management
│   ├── logging_conf.py      # Logging configuration
│   ├── mcp_client.py        # MCP client implementation
│   ├── orchestrator.py      # Workflow orchestration
│   ├── actions/             # Action handlers
│   │   └── alert.py         # Alert generation and handling
│   ├── connectors/          # Data source connectors
│   │   ├── __init__.py
│   │   ├── azuresql_connector.py    # Azure SQL connector
│   │   ├── excel_connector.py       # Excel file connector
│   │   ├── pos_connector.py         # POS system connector
│   │   ├── salesforce_connector.py  # Salesforce CRM connector
│   │   ├── snowflake_connector.py   # Snowflake data warehouse connector
│   │   └── social_connector.py      # Social media data connector
│   └── forecasting/         # Forecasting module
│       ├── __init__.py
│       ├── features.py      # Feature engineering
│       └── model.py         # ML model implementation
└── mcp_server/              # MCP server implementation
    └── app.py               # FastAPI MCP server application

This structure organizes the code into logical modules:

agent/: Core AI agent functionality
connectors/: Data source integrations
forecasting/: Machine learning and prediction logic
mcp_server/: Model Context Protocol server
actions/: Specific action implementations