Shopping Assistant

Shopping Assistant built using Pydantic AI & Google Gemini LLM. Deployed on Huggingface

Demo Video

shop_agent_gui_demo_wmusic.mp4

Helps :

• Create the optimal search query
• Retrieve the available filters for the product on the website
• Filter out products based on User's preferences
• Recommend the most suitable product as well as the top 10 candidates

Table of Contents

• Introduction
• Demo Video
• Setup
• Workflow
• Results

Introduction:

It is an agentic system built using Pydantic AI & Pydantic for data validation . All data is consistently validated. Google Gemini 2.5 Flash lite is used as the base LLM & flipkart is chosen as the shopping site currently. The agent has access to various tools :
- get_pro_class: Extract the product category & type from the user's query. - prompt_user0: Prompt the user for further details about the product. - rephrase_query : Generate a very concise product search query - combining original query & user provided details - get_best_site: Find the best website to search for the product. - get_site_filters: Retrieve all the available filters on the best site for the product. - prompt_user1: Prompt the user for further details about the product using the available filters - get_candidates : Return the best candidates found for the product & chooses them depending on the workflow , user's demands & its own logic.

Ideal Workflow:

graph TD;
  A[User Input] --> B[Agent];
  B --> C[Product Class];
  B --> D[Prompt User for Product Details];
  B --> E[Optimize Search Query with All Details];
  B --> F[Get Best Site];
  B --> G[Get All Site Filters];
  B --> H[Prompt User for Which Filters to Use];
  B --> I[Search, Filter & Return Products];

Loading

Setup

To run the project locally:

1. Clone the repository: git clone https://github.com/eshan1347/shop_agent
2. Install Python dependencies: pip install -r requirements.txt
3. Run: python app_gui.py

Results

This agentic system can be further improved by increasing the number of tools available - so that even more freedom is afforded to the agent. Web scraping data from websites add too much latency to the system - will further explore if any API's are available . Various other sites will be added so that products from various sites can be retrieved & a more holistic recommendation can be returned.