Iceberg Vector Database

Iceberg provides a high-performance, serverless vector search platform specifically engineered for applications requiring efficient data retrieval and persistent agent memory. Key features include: • Sub-20ms latency for rapid searches • Persistent memory for digital agents • Hybrid RAG engine for enhanced recall • India-hosted for data sovereignty and speed • Broad SDK support for various languages and frameworks This platform is designed to streamline the deployment of advanced search and agent capabilities without the need for extensive operational management. It combines dense vector search with BM25 keyword matching through a Reciprocal Rank Fusion (RRF) algorithm, offering superior recall compared to pure vector methods. Developers can configure the alpha parameter per query for tailored search results. Iceberg is hosted natively in Mumbai, ensuring compliance with data protection regulations and delivering ultra-low latency for users within India. This localized infrastructure means no data leaves India, which is critical for many organizations. The platform supports four types of agent memory—short-term, long-term, episodic, and semantic—to power sophisticated digital agent behaviors. Ideal for engineering teams, product developers, and researchers building applications that require fast, reliable semantic search, context-aware digital agents, or robust knowledge retrieval systems. It integrates seamlessly with popular developer tools and frameworks, allowing quick integration into existing stacks.