Endee is a high-performance, open-source vector database engineered for AI search, RAG, semantic search, and recommendation systems. It targets teams requiring production-grade performance and control, capable of handling up to 1 billion vectors on a single node. The database offers significant performance gains through optimized indexing and execution, combined with advanced features like hybrid retrieval and metadata filtering.

How It Works

Implemented in C++ and optimized for modern CPUs (AVX2, AVX512, NEON, SVE2), Endee provides a dedicated solution for retrieval-heavy AI workloads. Its core approach integrates dense vector search with sparse vector capabilities for hybrid search, alongside robust payload filtering for metadata-aware retrieval. This design aims for low-latency, high-throughput performance essential for RAG pipelines, AI agents, and semantic search applications.

Quick Start & Requirements

The fastest local setup involves downloading and running provided scripts: chmod +x ./install.sh ./run.sh, followed by ./install.sh --release --avx2 and ./run.sh. The server defaults to port 8080. CPU optimization flags (e.g., --avx2) are recommended for performance. Detailed installation, build, Docker, and runtime instructions are available in the hosted documentation at docs.endee.io.

Highlighted Details

• Scales to handle up to 1 billion vectors on a single node.
• Supports hybrid retrieval by combining dense vector search with sparse vector capabilities.
• Features payload filtering for structured, metadata-aware retrieval logic.
• Includes operational features such as backup APIs and runtime observability.
• Optimized builds target specific CPU instruction sets (AVX2, AVX512, NEON, SVE2).

Maintenance & Community

Endee welcomes community contributions via GitHub issues and pull requests. For support and discussion, the project maintains a Discord community. General inquiries and enterprise contact are available via enterprise@endee.io.

Licensing & Compatibility

Endee is licensed under the Apache License 2.0. However, the license explicitly prohibits using the "Endee" name, logos, or branding in a way that implies endorsement or affiliation. Users offering hosted services must use their own branding.

Limitations & Caveats

The primary caveat relates to licensing: while Apache 2.0 permits commercial use, the trademark restrictions necessitate distinct branding for any derived or hosted services, preventing implication of official Endee affiliation. Specific platform support beyond CPU optimizations and detailed performance benchmarks are not extensively detailed in the README.