Fully API driven - which allows integration with other systems
Manage concurrent traffic and auto-scale during traffic spikes
A central reusable repository of parsers, loaders, embedders and retrievers
Work with different kinds of data, like regular text files, PDFs, and even Markdown files
load the data from different sources like local directories, S3 buckets, databases, Truefoundry artifacts, etc
Support for various pre-trained models available to embed the data such as models from OpenAI, Cohere, etc
Support for SOTA reranker (as of April, 2024) from mixedbread-ai
Support for various available vector databases in the market, like Qdrant, Singlestore, Chroma, Weaviate, etc
Inference with best support for concurrent requests, autoscaling, etc