M4 Master ShopGPT - AI Shopping Assistant

Team

  • Ilham Muhammad
  • Thu Huong Vu
  • Kim Ngan Le Dang
  • Matthias Hamborg

Supervision

Christopher Lennan, Jakob Panten

New AI Technologies

Using cutting-edge AI technologies presented us with significant challenges. It required considerable time and effort to thoroughly research Prompt Engineering, GPT models, LangChain framework and grasp these concepts, ensuring a solid understanding for effective integration into our project.

Keeping Up with Rapid Technology Changes

As technology rapidly evolves, we encountered challenges in staying up-to-date with frequent updates in LangChain versions, OpenAI function calls, GPT models, and Vector Database. This necessitated continuous monitoring and adaptation to ensure compatibility.

Accessing idealo's Database

The process of gaining access to idealo’s database and API was not as fast as we expected, which did impact the early stages of development. Despite the delays, we persevered and established effective communication to overcome this obstacle.

Image Similarity Search with Vector Database

Developing an image similarity search with Vector DB posed a significant challenge for us due to limited time and various methods considered to retrieve similar images from the database. Creating custom scripts to retrieve extensive product metadata and embedding data effectively in our vector database introduced additional complexities.

Deployment Hurdles

Although we considered deploying our website, we encountered security concerns and limitations. Due to security concerns and the early version prototype status of Idealo, we couldn’t expose the OpenSearch API key for deployment. Additionally, we needed to ensure cost-effectiveness and prevent potential misuse by users. As a result, we made the decision not to proceed with deployment at this time.