DEV Community

Cover image for Enhancing Search Experience with AI: LLM Open Search on Cloudflare Workers
Jackson Kasi
Jackson Kasi

Posted on

Enhancing Search Experience with AI: LLM Open Search on Cloudflare Workers

This is a submission for the Cloudflare AI Challenge.

What I Built

LLM Open Search is a serverless AI app deployed on Cloudflare Workers, designed to elevate the search experience by providing intelligent responses, relevant questions, and chat completions. It leverages Cloudflare's Workers AI and a couple of models from the Cloudflare catalog to analyze and respond to user queries effectively.

Demo

Explore the capabilities of LLM Open Search through this live demonstration: LLM Open Search

LLM Open Search in action

Open Search - Powering insights with AI - 13 April 2024 - Watch Video

My Code

Dive into the code and setup for LLM Open Search on GitHub: GitHub Repository - LLM Open Search

You'll find a comprehensive README file that details installation and running the project, ensuring easy replication and local setup.

Journey

Throughout the development of LLM Open Search, I focused on utilizing Cloudflare’s AI models to enhance text-based search functionalities:

  • Text Generation with @cf/meta/llama-2-7b-chat-int8: Used for generating relevant follow-up questions based on the initial search results.
  • Chat Completion with @cf/meta/llama-2-7b-chat-fp16: Applied to refine search results into coherent and contextually relevant responses.

This application solely utilizes AI for text manipulation and generation, demonstrating a focused application of AI technology in enhancing search experiences.

Innovation and Creativity

LLM Open Search integrates Cloudflare’s AI models in a unique way, focusing on enhancing the depth and relevance of search engine responses. The system effectively parses user queries to deliver not only direct answers but also thoughtful follow-up questions, enriching user interaction.

What’s Next? 🚀

Moving forward, I plan to introduce features like personalized search histories and the ability to group searches into collections. These features aim to provide users with a more tailored and efficient search experience.

Submitting for Multiple Models and/or Triple Task Types

I want to mention that this project does not qualify for the additional prize categories, as it strategically focuses on utilizing two AI models to perform sophisticated text generation tasks efficiently. However, if I implement my upcoming features in this project with any other AI model, I will update it here.

Thank you for considering LLM Open Search in this challenge.😊

Top comments (0)