Anvil: High-Performance Object Storage for AI Workloads

Anvil is an AI-native, open-source object store built for modern machine learning and AI applications. Designed with performance and scalability in mind, it provides S3-compatible storage optimized for large AI models, datasets, and inference workloads. With features like model-aware indexing, tensor streaming via QUIC, erasure coding, and multi-region clustering, Anvil enables efficient storage and retrieval of ML artifacts such as safetensors, GGUF files, and ONNX models. It supports multi-tenancy with isolated namespaces and per-tenant policies, making it ideal for both individual developers and teams working on high-performance AI systems. Built in Rust and fully compatible with Docker Compose, Anvil offers a self-hosted, scalable solution for AI infrastructure.

Key Features:

• S3 Compatibility: Full support for S3 API for seamless integration with existing tools.
• Model-Aware Storage: Optimized indexing and organization of AI model files and datasets.
• High-Performance Tensor Streaming: Low-latency streaming using QUIC for fast model loading and inference.
• Efficient Storage: Erasure coding for reliable, space-efficient replication.
• Multi-Region Clustering: Deploy across regions for global access and redundancy.
• Native gRPC API: High-speed communication for AI workloads.
• Multi-Tenancy: Isolated namespaces and per-tenant policies for secure team collaboration.
• Self-Hosted & Open Source: Built in Rust, licensed under Apache 2.0, deployable via Docker Compose.

Pricing: Anvil is open source and free to use. It is available under the Apache 2.0 license, with no paid tiers or usage-based fees. Users can self-host the solution with no cost.

Conclusion: Anvil is a powerful, scalable, and open-source object store tailored specifically for AI workloads. Its combination of S3 compatibility, high-performance streaming, and AI-optimized features makes it an ideal infrastructure choice for developers and teams building modern machine learning applications.