Infinity is an AI-native database designed for LLM applications, offering hybrid search capabilities across dense vectors, sparse vectors, tensors, and full-text data. It targets developers building RAG applications, conversational AI, and recommendation systems, providing fast and flexible data retrieval.

How It Works

Infinity employs a hybrid search approach, combining multiple indexing and retrieval strategies. It supports dense vector search with low latency (0.1ms) and high throughput (15K+ QPS), alongside efficient full-text search (1ms, 12K+ QPS on 33M documents). The system handles rich data types and offers various reranking methods like RRF and weighted sum, enabling sophisticated query combinations.

Quick Start & Requirements

• Install via pip: pip install infinity-embedded-sdk==0.6.0.dev3
• Requires Python.
• Supports embedded and client-server modes.
• Documentation: Document, Python API, HTTP API References

Highlighted Details

• Achieves 0.1ms query latency and 15K+ QPS on million-scale vector datasets.
• Supports hybrid search (dense, sparse, tensor, full-text) with filtering.
• Offers a single-binary, dependency-free architecture for easy deployment.
• Intuitive Python API for seamless integration into AI workflows.

Maintenance & Community

• Active development with a public roadmap.
• Community channels available via Discord and Twitter.

Licensing & Compatibility

• License details are not explicitly stated in the provided README snippet.

Limitations & Caveats

• The SDK version 0.6.0.dev3 indicates a pre-release or development version, potentially subject to instability or breaking changes.
• Full licensing information is not immediately available in the README.