Dria

Dria is a multi-region, serverless collective Knowledge Hub. It consists of embedding bases, each referred to as a Knowledge. Each Knowledge in Dria is a smart contract, allowing Dria to serve vast repositories of public knowledge, such as Wikipedia or Common Crawl, as shared RAG models.

Dria simplifies the creation and management of indexes, offering two primary ways to interact with the system: users can either upload documents directly through our platform or use our client tools for manual vector insertion. When files are uploaded via the platform, Dria handles document parsing, inference, storage, and indexing seamlessly.

Every index on Dria is publicly available, free, and forever for other developers to use locally through the public gateways of Arweave.

Dria Glossary

  • Knowledge: Once an index is created via the platform or through native clients, it appears in the Knowledge Hub as "Knowledge."

  • Knowledge Hub: A centralized repository on Dria where all created Knowledge is listed. This feature enables users to discover and utilize Knowledge crafted by others, integrating it into their Retrieval-Augmented Generation (RAG) workflows as a knowledge base.

  • Contract ID: Given that each index also functions as a smart contract, every piece of Knowledge (index) is assigned a unique contract ID. This ID can be used to locate the specific contract on Sonar Warp, which is essential for importing a contract from Dria to your local environment.

  • Category: Users have the ability to tag their Knowledge with categories, facilitating streamlined navigation and more efficient exploration of Knowledge.

  • File Type: Identifies the format of files uploaded to create Knowledge via the platform, indicating the source material from which the Knowledge was derived.

  • Embed Column: For CSV uploads, specifying a column will direct the system to generate embeddings from the data within that column, termed the Embed Column. In the absence of a specified column, embeddings are produced from the entirety of the dataset for each record.

  • Entry Count: Represents the total number of vectors contained within a piece of Knowledge.

  • Index Type: The Dria Platform supports the creation of two types of indexes: HNWS and ANNOY. "Index Type" denotes the algorithm used for a particular piece of Knowledge.

Last updated