What is Dera?
Dera is a user-friendly tool for managing chunks and embeddings, enabling rapid iteration on chunking strategies for better Retrieval-Augmented Generation (RAG) apps. It eliminates the need for complex vector databases during the development phase and offers a structured approach to evaluate chunking strategies as a team.
🗂️ Schema Definition: Define embedding schemas as Postgres tables, providing a structured approach to manage chunks and embeddings.
📥 Upload and Match: Seamlessly upload text and embeddings to defined schemas, and perform embedding matches via intuitive REST APIs.
📊 Results Evaluation: Store and review match queries with their results in the web application, facilitating collaborative analysis and evaluation.
🤖 NLP Research: Dera empowers researchers to swiftly test various chunking strategies, accelerating the development of effective RAG models.
💻 Software Development: Developers can leverage Dera to integrate chunking strategies into their RAG applications, enhancing retrieval accuracy.
🎓 Educational Institutions: Dera serves as a valuable tool for teaching and learning about chunking strategies and their impact on RAG performance.
Dera's user-friendly interface, coupled with its efficient Postgres-based embedding schema management, makes it an invaluable asset for developers, researchers, and educators working on Retrieval-Augmented Generation. Its ability to streamline chunking strategy iteration and evaluation empowers users to build high-performing RAG systems with ease.