Today's RAG & Vector Databases: Fastest-Growing Projects — September 05, 2026
Today's the RAG & Vector Databases space, we see a mix of emerging and established projects gaining traction on GitHub. While some repositories are showing significant growth scores, others have been maintaining steady progress with regular commits. Among these, deeplethe/utopia stands out as one of the most active and quickly growing projects.
deeplethe/utopia is a self-hosted knowledge platform built using Rust and PostgreSQL that leverages Retrieval-Augmented Generation (RAG) over documents on a knowledge graph. The platform remembers when each fact was true, offering a dynamic and contextually aware search experience. Its growth score of 95.48 indicates rapid community interest, likely due to its unique approach to managing temporal information in a knowledge base.
rostamlabs/rostam is an open-source vector database designed as a Go engine that can be embedded as a library or run standalone. It supports sub-microsecond key-value operations and various indexing methods like HNSW, IVF, and Vamana, along with quantization and hybrid dense+sparse search capabilities. With a growth score of 12.23 and consistent commits over the last month, rostam appears to be steadily gaining traction among developers looking for high-performance vector storage solutions.
makralabs/makra serves as a memory layer between web services and AI agents by providing real-time structured data from the open web at the cost of a vector search. This makes it particularly useful for applications requiring frequent updates from dynamic sources. Despite having a lower growth score of 4.62, its steady increase in stars suggests that developers are beginning to recognize its potential for integrating AI with live data feeds.
Shuo-Liang-0111/RA-Bench is a real-event-anchored benchmark designed to detect AI-generated videos in crisis settings. The project aims to provide reliable methods for distinguishing between genuine and synthetic media, which has become increasingly important as deepfake technology advances. With 104 stars but only a modest growth score of 2.78, RA-Bench seems to be attracting interest from researchers and practitioners concerned with the authenticity of visual content in critical situations.
These projects collectively illustrate the diversity and dynamism within the RAG & Vector Databases space, highlighting both innovative approaches to knowledge management and cutting-edge solutions for vector storage and search.
deeplethe/utopia is a self-hosted knowledge platform built using Rust and PostgreSQL that leverages Retrieval-Augmented Generation (RAG) over documents on a knowledge graph. The platform remembers when each fact was true, offering a dynamic and contextually aware search experience. Its growth score of 95.48 indicates rapid community interest, likely due to its unique approach to managing temporal information in a knowledge base.
rostamlabs/rostam is an open-source vector database designed as a Go engine that can be embedded as a library or run standalone. It supports sub-microsecond key-value operations and various indexing methods like HNSW, IVF, and Vamana, along with quantization and hybrid dense+sparse search capabilities. With a growth score of 12.23 and consistent commits over the last month, rostam appears to be steadily gaining traction among developers looking for high-performance vector storage solutions.
makralabs/makra serves as a memory layer between web services and AI agents by providing real-time structured data from the open web at the cost of a vector search. This makes it particularly useful for applications requiring frequent updates from dynamic sources. Despite having a lower growth score of 4.62, its steady increase in stars suggests that developers are beginning to recognize its potential for integrating AI with live data feeds.
Shuo-Liang-0111/RA-Bench is a real-event-anchored benchmark designed to detect AI-generated videos in crisis settings. The project aims to provide reliable methods for distinguishing between genuine and synthetic media, which has become increasingly important as deepfake technology advances. With 104 stars but only a modest growth score of 2.78, RA-Bench seems to be attracting interest from researchers and practitioners concerned with the authenticity of visual content in critical situations.
These projects collectively illustrate the diversity and dynamism within the RAG & Vector Databases space, highlighting both innovative approaches to knowledge management and cutting-edge solutions for vector storage and search.