California Procurement Agent

 

About   |   Features   |   Technologies   |   Architecture   |   Requirements   |   Starting   |   License   |   Contact

🎯 About

California Procurement Agent is an intelligent analytical platform designed to explore and analyze California's state procurement data from the eSCPRS (Electronic California Procurement Reporting System). The system combines Retrieval-Augmented Generation (RAG) technology with advanced data visualization to provide insights into government spending patterns, supplier analysis, and procurement trends.

The platform features an AI-powered chat interface that allows users to ask natural language questions about procurement data and receive detailed text-based analysis, combined with comprehensive data exploration notebooks and interactive visualizations. The system processes real procurement data including purchase orders, supplier information, spending categories, and acquisition methods.

Key Data Insights:

• Analysis of $X+ in state procurement spending
• Examination of supplier diversity and qualification programs
• Contract vs non-contract spending patterns
• Department-wise spending analysis
• CalCard (state credit card) usage patterns
• Geographic distribution of procurement activities

Kaggle Notebook: For a detailed exploratory data analysis, check out our comprehensive notebook: California State Procurement EDA

---

✨ Features

AI-Powered Chat Interface:

• Natural Language Queries: Ask questions about procurement data in plain English
• Intelligent Agent: LangChain-powered agent with MongoDB query tools for complex data analysis
• Intelligent Responses: Get contextual answers with proper formatting and data insights
• Chat History: Save and manage conversation history with procurement analysis

Data Exploration & Visualization:

• Interactive Charts: Plotly-powered visualizations for spending patterns, supplier analysis, and trends (available in Jupyter notebook)
• Comprehensive EDA: Jupyter notebook with statistical analysis and data insights

Procurement Data Analysis:

• Supplier Analysis: Identify top suppliers, qualification status, and diversity metrics
• Spending Patterns: Analyze contract vs non-contract spending, CalCard usage
• Department Insights: Compare spending across different state departments
• Acquisition Methods: Understand procurement methods and their distribution

Technical Features:

• MongoDB Integration: Efficient storage and querying of large procurement datasets
• Data Normalization: Consistent field naming and data type handling
• API Endpoints: RESTful APIs for data access and analysis
• Modern UI: React-based frontend with responsive design

---

🚀 Technologies

The following tools and frameworks were used in this project:

• Backend:

  • FastAPI - High-performance web framework for APIs
  • MongoDB - NoSQL database for procurement data storage
  • Google Gemini - AI model for natural language processing
  • OpenAI - AI model for text generation and understanding
  • LangChain - Framework for LLM applications and agent orchestration
  • Pandas - Data manipulation and analysis
  • Plotly - Interactive data visualization
  • Matplotlib - Static visualizations
  • Seaborn - Statistical data visualization

• Frontend:

  • React - UI library for modern web applications
  • Vite - Fast build tool and development server
  • Axios - HTTP client for API communication

• Data Analysis & Visualization:

  • Plotly - Interactive charts and graphs
  • Matplotlib - Static visualizations
  • Seaborn - Statistical data visualization
  • Jupyter - Interactive notebooks for data exploration

• Data Processing:

  • Kaggle API - Dataset downloading and management
  • Custom data normalization and cleaning pipelines

Architecture

The application follows a modern data analysis architecture:

1. Data Layer:

   • MongoDB for storing normalized procurement data and chat history and user sessions

2. Processing Layer:

   • Data loading and normalization from Kaggle datasets
   • Field standardization and type conversion
   • Indexing for efficient querying

3. Analysis Layer:

   • Pandas-based data manipulation and analysis
   • Statistical computations and aggregations
   • Time-series analysis for procurement trends

4. AI Layer:

   • LangChain-powered RAG system with intelligent agent
   • Google Gemini for natural language understanding
   • MongoDB query tools for data retrieval and analysis
   • Custom prompts for procurement-specific queries

5. Visualization Layer:

   • Plotly for interactive web-based charts
   • Matplotlib/Seaborn for static analysis
   • Jupyter notebooks for exploratory analysis

6. API Layer:

   • FastAPI for RESTful endpoints
   • Chat management and data querying APIs

7. Frontend Layer:

   • React-based chat interface with text-based AI responses
   • Data visualization available through Jupyter notebooks
   • Responsive design for multiple devices

---

Pipeline

The procurement data analysis pipeline includes:

1. Data Acquisition: Download and load California procurement datasets from Kaggle
2. Data Normalization: Standardize field names, convert data types, handle missing values
3. Database Storage: Store processed data in MongoDB for efficient querying
4. Analysis Engine: Generate insights, statistics, and visualizations
5. AI Integration: Enable natural language queries about procurement data
6. User Interface: Provide chat interface and data exploration tools

---

✅ Requirements

Before starting, ensure you have the following installed:

• Python 3.11+
• Node.js 16+ and npm
• MongoDB (local or cloud instance)
• Google Gemini API key
• Kaggle API credentials (optional, for data updates)

---

🏁 Starting

# Clone this project
$ git clone https://github.com/romanyn36/california-procurement-agent.git

# Navigate to the project directory
$ cd california-procurement-agent

# Create a virtual environment i use uv package manger
$ uv sync

# Activate the virtual environment
$ source .venv/bin/activate  # For Linux/Mac
$ .\.venv\Scripts\activate    # For Windows

# Set up environment variables
$ cp .env.example .env
# Edit .env with your API keys:
# MONGODB_URI=mongodb://localhost:27017/
# OPENAI_API_KEY=your_openai_api_key
# GEMINI_API_KEY=your_gemini_api_key
# KAGGLE_USERNAME=your_kaggle_username
# KAGGLE_KEY=your_kaggle_key

# Load the procurement data
# and populate the MongoDB database
$ python -m database.data_loader

# Start the backend server
$ python -m uvicorn app:app --reload --host 0.0.0.0 --port 8000

# In a separate terminal, navigate to the frontend directory
$ cd agent-frontend

# Install frontend dependencies
$ npm install

# Start the development server
$ npm run dev

# Access the application:
# Frontend: http://localhost:5173
# Backend API: http://localhost:8000
# Data Exploration: Open data_exploring.ipynb in Jupyter

Data Exploration

For detailed data analysis, visit my Kaggle notebook: California State Procurement EDA

The notebook includes:

• Data loading and preprocessing
• Statistical analysis
• Interactive visualizations
• Procurement insights and trends

💬 Example Queries

• tell me about the order with purchase order number REQ0011118
• Show me all orders for laptops or computers
• Who are the top 10 suppliers by total spending
• How many purchase orders were created in 2013
• What is the total spending across all departments?
• What are the top 5 departments by number of orders?
• Which suppliers have DVBE certification?

Configuration

Key configuration files:

1. .env: API keys and database connections
2. prompt_template.py: AI model prompts and field mappings
3. database/mongodb_tools.py: Database query functions
4. agent.py: LangChain agent configuration with MongoDB query tools

🚧 What's Next?

Future enhancements planned:

• Chat Visualizations: Add interactive charts and graphs directly in the chat interface to visualize procurement data insights alongside text responses
• Advanced Analytics: Machine learning models for spending prediction

📝 License

This project is licensed under the MIT License. For more details, see the LICENSE file.

❤️ Contact Me

• Made by Romani – an AI Engineer and Backend Developer. Feel free to reach out for collaborations, questions, or new projects! You can contact me via email: contact@romaninasrat.com

• You can also find me on:

Back to top