StellarGPT v1.0

Semantic Search over Soroban and Stellar Docs using Natural Language Queries. V1 includes just includes code to generate vector-database with Pinecone over Soroban Docs, and allowing Semantic Search.

Comment

Inspiration

My vision is to revolutionize the way developers and users interact with the Stellar ecosystem. I aim to achieve this through advanced natural language processing and semantic search capabilities.

What it does

StellarGPT v1 utilizes cutting-edge Natural Language Processing techniques to perform semantic searches over Soroban and Stellar Docs. By comprehending the context and intent behind user queries, StellarGPT provides the most relevant and accurate answers. This greatly enhances the user experience and diminishes the time developers invest in searching for information.

How I built it

The application is constructed atop Pinecone vector databases. I employed crawler functions to extract data from open-sourced Soroban Documentation and then stored the pertinent information in the PineCone Database. Subsequently, CustomGPT prompts search through this data using Semantic Natural Language queries, leveraging cosine similarity for results. As this is version 1, my focus has been solely on retrieving information using cosine similarity.

Challenges I ran into

During the development phase, I confronted multiple challenges, from determining the optimal database for pgvector to transitioning the existing frameworks into a web-based application. I also grappled with various design and UI challenges but managed to navigate them successfully.

Accomplishments that I'm proud of

I take immense pride in the final product with efficient and accurate multilingual answers. The combination of speed, user-friendly UI, and accuracy truly sets it apart.

What I learned

Throughout this journey, I grasped the importance of adaptability when faced with unforeseen challenges. I discovered the intricacies of integrating advanced Natural Language Processing techniques with real-world applications, realizing that a strong theoretical foundation often requires practical tweaks for effective implementation. Additionally, I gained a deeper appreciation for UI/UX design nuances, understanding that a positive end-user experience is crucial for any product's success. This project solidified my problem-solving capabilities and reaffirmed the significance of dedication and effort.

What's next for StellarGPT v1.0

In the pipeline is the release of StellarGPT v2.0, which will be centered around an index-based, multimodal search of the Soroban Docs. This means the system will be adept at deciphering user queries and ranking answers based on their relevance to the documentation's index.

Built With

Share this project:

Updates