Welcome to the Healthsearch Demo, an open-source project aimed at showcasing the potential of leveraging user-written reviews and queries to retrieve supplement products based on specific health effects.
The search functionality in this demo accepts natural language queries that are translated into GraphQL queries using LLMs. These GraphQL queries are then utilized to retrieve supplements from a Weaviate database. The demo also exhibits an example of generative search by providing product summaries generated based on the retrieved objects.
⚠️ Disclaimer: Healthsearch is a technical demonstration, and the results shown should not be treated as health advice. The results and generated summaries are purely based on user-written reviews.
We use Large Language Models (LLM), like GPT4, to translate natural language queries into a structured query format, called a GraphQL query.
The demo extracts information about filters, sorting, and limits directly from the context of the query. Whether the query is the top 10 products for glowing skin, products for sleep from a specific brand, or best-rated products for toothache, the demo can interpret these queries and generate an appropriate GraphQL query in return.
Healthsearch relies on the power of semantic search in user reviews. When seeking products that are good for joint pain, for instance, Healthsearch scans user reviews for discussions on products that have alleviated joint pain or similar conditions. The results are then aggregated and grouped according to their respective products.
After the translation of the query to GraphQL and the retrieval of the most semantically relevant product, we enhance our demo with a feature called Generative Search. Essentially, we examine the top five results and employ an LLM to generate a product summary. This concise summary offers a brief overview of the products, highlighting their pros and cons and providing valuable insights. Each summary is crafted around the query, ensuring every search is unique and interesting.
We embed the generated results and queries to Weaviate, and use it as a Semantic Cache.
This method is advantageous as it enables the demo to return results from queries that are semantically equal to the new query. For example good for joint pain and helpful for joint pain are semantically very similar and should return the same results, whereas bad for joint pain should have its own generated result. This method allows us to gain much more from generated results than traditional string matching would permit. It's a simple yet potent solution that enhances the efficiency of the search process.
This repository is designed to serve as a template - a starting point for your own projects with Weaviate. Take inspiration from how we've implemented certain features and feel free to enhance it in your own project. We welcome comments, ideas, and feedback. Embrace the open-source spirit!
The Healthsearch Demo is structured in three main components:
- A Weaviate database (either cluster hosted on WCS or local).
- A FastAPI endpoint facilitating communication between the LLM provider and database.
- An interactive React frontend for displaying the information.
Make sure you have Python (>=3.8.0) and Node (>=18.16.0) installed. We also support Docker and provide Dockerfiles for the setup.
You can use Docker to setup the demo in one line of code! If you're not familiar with Docker you can read more about it here (https://docker-curriculum.com/)
- Set environment variables:
- The following environment variables need to be set
OPENAI_API_KEY=your-openai-api-key
Use the
.envfile inside the backend folder to set the variable (https://github.com/theskumar/python-dotenv) Note that if you're using the GPT-4 model (by default), ensure your OpenAI key has access.
- Use docker compose
docker-compose up
- Access the frontend on:
localhost:3000
To kick-start with the Healthsearch Demo, please refer to the READMEs in the Frontend and Backend folders:
Follow these steps to use the Healthsearch Demo:
- Set up the Weaviate database, FastAPI backend, and the React frontend by following the instructions in their respective READMEs.
- Launch the database, backend server, and the frontend application.
- Use the interactive frontend to input your natural language query related to a health condition or benefit.
- The frontend sends the query to the backend, which transforms the natural language query into a GraphQL query using the LLM.
- The backend sends the GraphQL query to the Weaviate database to fetch relevant reviews based on the user query.
- The frontend displays the results, allowing you to explore the most semantic-related supplements to your specific health-related query.
Your contributions are always welcome! Feel free to contribute ideas, feedback, or create issues and bug reports if you find any! Please adhere to the code guidelines that include formatting, linting, and testing.