ai-shopping-buddy

An e-commerce conversational chatbot that gathers customer preferences to provide personalized product recommendations. Developed for "Generative AI World Cup 2024: So you think you can hack" by Databricks.

Agent Graph

The graph below illustrates the implemented logic of the shopping buddy's agentic service.

%%{init: {'flowchart': {'curve': 'linear'}}}%%
graph TD;
	__start__([<p>__start__</p>]):::first
	gather_preference(gather_preference)
	manage_state(manage_state)
	greeting(greeting)
	parse_preference(parse_preference)
	match_products(match_products)
	recommend(recommend)
	find_related_products(find_related_products)
	recommend_related_products(recommend_related_products)
	__end__([<p>__end__</p>]):::last
	__start__ --> manage_state;
	find_related_products --> recommend_related_products;
	manage_state --> greeting;
	match_products --> recommend;
	parse_preference --> match_products;
	recommend_related_products --> __end__;
	greeting -.-> __end__;
	greeting -.-> gather_preference;
	gather_preference -.-> parse_preference;
	gather_preference -.-> __end__;
	recommend -.-> find_related_products;
	recommend -.-> __end__;
	gather_preference -.-> gather_preference;
	classDef default fill:#1e1e1e, stroke:#ffffff, color:#ffffff;
    classDef first fill:#4a4a4a, color:#ffffff;
    classDef last fill:#3b3b3b, color:#ffffff;
    classDef node fill:#2c2c2c, stroke:#ffffff, color:#ffffff;

Loading

☁️ Databricks Setup

1. Refer to "Accessing the Dataset" section to obtain access to the dataset
2. Get the Serving endpoints of databricks-meta-llama-3-1-70b-instruct on databricks
3. Run vectorstore/create_vectorstore.py
4. Execute the main.py notebook, which serves as the entry script for the chatbot, to start chatting.

🗂️ Dataset

The Best Buy Products Dataset is used. It is a sample dataset obtained from Databricks Marketplace.

Data Fields

Here are the key fields in the dataset, omitting unused columns:

Schema({
    'product_id': Int64,
    'title': String,
    'images': String,
    'final_price': String,
    'discount': String,
    'initial_price': String,
    'root_category': String,
    'features_summary': String,
    'features': String,
    'product_specifications': String,
})

Accessing the Dataset

To access the dataset on databricks, follow these steps:

1. Go to Best Buy Products Dataset
2. Click Get Instant Access
3. Set Catalog name as bright_data_best_buy_products_dataset

The data will be available as bright_data_best_buy_products_dataset.datasets.best_buy_products

Acknowledgments

This dataset is provided by Bright Data. All rights reserved.

We would like to thank Bright Data for making this data available for research and development purposes. For more information about Bright Data, please visit their official website.