Smart Car Buyer AI Agent

Developed a Smart Car Buyer AI Agent during the AgentCraft Hackathon, leveraging LangGraph and Python to streamline car purchasing decisions.

AI/ML Solutions

Date: 17/11/2024

Tags:AI/ML Solutions

Client: AgentCraft Hackathon

Team: Aurore Pistono, Clément Florval, Louis Gauthier

Smart Car Buyer AI Agent for the AgentCraft Hackathon

Overview

Our team at Digiwave, consisting of Aurore Pistono, Clément Florval, and Louis Gauthier, participated in the AgentCraft Hackathon organized by DiamantAI in conjunction with LangChain. The hackathon attracted more than 100 participants, fostering a competitive and innovative environment. During this event, we developed a Smart Car Buyer AI Agent. This agent is designed to assist users in making informed car purchasing decisions by leveraging LangGraph and Python to create an interactive and efficient car-buying experience.

Conversation Screenshot 1

While the current implementation focuses on assisting users in buying cars, the agent is built to be easily extendable to support additional products and services in the future, such as buying computers or renting houses.

Project Details

Objective

The primary goal was to streamline the car purchasing process by developing an AI agent capable of:

  • Understanding user needs and preferences through natural language interaction.
  • Refining and applying complex filters across multiple platforms.
  • Providing actionable insights and recommendations to the user.

Solution

  1. User Input Processing: Utilizing LLM-powered interactions to dynamically understand and summarize user requirements.

  2. Filter Refinement: Translating user preferences into specific search filters that can be applied across different platforms.

  3. Web Scraping and Integration: Implementing web scrapers to interface with platforms like AutoTrader, enabling the agent to fetch and present relevant car listings.

  4. Summarization and Insights: Providing concise summaries and insights into listings, including general market reliability and additional information retrieved from the web.

  5. Detailed Listing Analysis: After scraping websites and identifying pertinent listings, users can inquire for more details about specific listings. The agent is equipped to:

    • Answer detailed questions about a listing.
    • Assess potential scams by verifying the authenticity of the listing.
    • Scrutinize information for any suspicious details, such as hidden accident reports or damage descriptions.
    • Perform online lookups using DuckDuckGo to validate prices and check for known issues with the car model.
    • Ensure that the listed price aligns with market standards and highlight any discrepancies.

Conversation Screenshot 2

Agent Architecture

The agent follows a structured workflow:

  1. User Need Assessment: Gathers and summarizes user preferences.

  2. Filter Building: Constructs and applies search filters based on user input.

  3. Listing Retrieval: Collects data from integrated platforms.

  4. Insight Delivery: Provides additional information and recommendations to assist in the decision-making process.

Smart Car Buyer Agent Architecture

Technologies Used

  • LangGraph: For building state-based workflows and managing the agent's conversational logic.
  • Python: The primary programming language for developing the agent and web scrapers.
  • OpenAI GPT Models: For natural language understanding and generation.
  • Playwright: For browser automation and web scraping.
  • LXML: For parsing and extracting data from HTML content.

Results

  • Efficiency: Reduced the time users spend searching and comparing cars by automating the filtering and retrieval process.

  • Clarity: Summarized complex data into actionable insights, making it easier for users to make informed decisions.

  • Flexibility: Designed the agent to be adaptable to various product categories beyond cars, allowing for future expansion into areas like buying computers or renting houses.

Demo on Hugging Face Spaces

We have deployed a demo of the Smart Car Buyer AI Agent on Hugging Face Spaces, featuring a Gradio interface:

Smart Car Buyer AI Agent Demo

You can duplicate the space and set your OPENAI_API_KEY environment variable in the space's parameters to try it out yourself.

Pitch Video

For a visual overview of the project, watch our pitch video.

Repository and Notebook

About Digiwave

Digiwave is committed to leveraging AI and innovative technologies to create impactful solutions. Our participation in the AgentCraft Hackathon organized by DiamantAI and LangChain reflects our dedication to pushing the boundaries of AI agent capabilities.


This project was developed by Aurore Pistono, Clément Florval, and Louis Gauthier at Digiwave. For more information about our services, visit Digiwave's Portfolio.

Ready to Work with Us?

Let's collaborate to bring your ideas to life. Get in touch with us to discuss your project requirements.