Ragswift

Ragswift, a scalable centralized embeddings platform, built to effortlessly handle document ingestion, storage and retrieval tasks at scale. Accelerate the development of RAG applications with ease. It eliminates the concerns associated with embeddings management within your RAG pipeline, you can just self host the solution and manage embeddings across multiple apps from a single place.

It harnesses the power of distributed computing through Ray, empowering users to effortlessly process vast document sets in parallel across multiple CPU and GPU nodes. The incorporation of Qdrant disk-based indexing and storage guarantees robust support for the scale of billions of vectors, positioning Ragswift as a formidable choice for large-scale applications.

Moreover, Ragswift will soon feature compute autoscaling capabilities using kubernetes, ensuring that you only pay for the compute resources you use. This cost-efficient model enhances the platform's flexibility, allowing users to scale their infrastructure dynamically in response to varying workloads, optimizing both performance and cost-effectiveness.

Although the default configuration of ragswift is good enough for most use cases and has features like rerankers and hybrid search for retrieval, but we don't restrict you to use the default settings. Everything in ragswift is configurable and you can experiment to find the best configuration for your use case.

Key Features

1. Distributed Computing with Ray - We leverage the power of Ray for distributed computing, enabling parallel document ingestion across multiple CPU and GPU nodes. This ensures optimal resource utilization for efficient and scalable processing.

2. Qdrant Disk-Based Indexing - To handle billions of vectors, we've integrated Qdrant disk-based indexing and storage. This ensures high-performance indexing for rapid and precise retrieval of relevant information.

3. REST APIs for Seamless Integration - RAG Framework provides REST APIs for convenient asset ingestion from sources like S3 and GitHub. Retrieval is also a breeze, seamlessly integrating into your existing workflows.

4. Configurability at Your Fingertips - Ragswift is highly configurable, allowing users to tailor the system to their specific needs. Configure the number of CPUs/GPUs, choose embedding models, set chunk size, and more.

Upcoming Features

We have exciting plans for Ragswift's future, including autoscaled deployment on Kubernetes, an admin UI for document management, configurable projects with the flexibility to experiment with different embedding dimension, chunk size, embedding models etc within a single deployment, an observability tool to compare the performance of embeddings across various parameters to improve the quality of embeddings backed by experiments, and access management features.