Object store built for AI workloads
High performance, AI-native objectstore with S3 API included
What it does
Anvil is an open-source, AI-native object store designed for modern workloads. We built it after hitting the limits of Git LFS, Hugging Face repos, S3, and others when working with multi-GB model files. It is S3-compatible & gRPC-native, supports: * Model-aware indexing - so it understands safetensors, gguf, and ONNX. * Tensor-level streaming * Erasure-coded storage * Open source (Apache-2.0) If you’re storing large models, versioning fine-tunes, running local inference, we want your feedback.
Does a similar job
all alternatives →
- SASpice.ai – materialize, accelerate, and query SQL data from any source2024 · github.com · ▲177
Hi HN, We're Luke and Phillip, and we're building Spice.ai OSS - a lightweight, portable runtime, built in Rust and powered by Apache DataFusion to locally materialize, accelerate, and query data tables sourced from any database, data warehouse or data lake. Phillip and I first introduced Spice on Show HN in September 2021. Since then, we’ve been schooled and humbled in every way building 100TB+ data and ML systems for the https://spice.ai cloud platform. Along with our customers, we struggled with getting fast, low-latency, high-concurrency SQL query within a budget, accessing and…
- CSCallFS – S3-style object store in one Go binary (MIT)2025 · github.com · ▲66
We started CallFS after yet another late-night “why did the uploads vanish?” incident. Our small team had stitched together rsync, a fragile NFS mount, and an S3 bucket—none of it observable, all of it waiting to bite us. So we wrote a single-process file service in Go that: • Speaks the S3 API (so existing tooling works). • Stores hot data on local disks for speed; cold data can sit in any S3-compatible bucket. • Exposes Prometheus metrics and JSON logs by default, because “what happened?” shouldn’t be guesswork. • Ships as a ~25 MB static binary—no external deps, MIT license. Today it’s…
- HAHelixDB – A graph database built on object storageJun 2026 · github.com · ▲159
Hey HN, it’s been just over a year since we launched HelixDB (https://news.ycombinator.com/item?id=43975423), a project a friend and I started in college. It’s an OLTP graph database built on object-storage, with native vector search and full-text search (FTS). Why graph, vector and FTS? Graph databases provide a natural cognitive model for data, vectors allow for a semantic understanding of the entities and relationships in the graph, and FTS provides more specific filtering. Many AI-driven applications attempt to combine all of these functionalities by stitching together…
- OOObjectiv – Open-source product analytics infrastructure2022 · objectiv.io · ▲88
Hi! One of the creators here. Very proud to finally be able to show you what we've been working on for over a year now. Curious to hear your thoughts! Objectiv is open-source (APLv2) product analytics infrastructure. It's built around a generic but strict event taxonomy, open/common data- and infra tools (currently PG, snowplow, working on bigquery with more to come), and the analyses are done using our pandas-like, SQL speaking modeling library called Bach. As a result, we’re moving towards a vision wherein models can be shared openly, independent of product, platform[1] or data…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 18d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 19d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 28d ago · cactuscompute.com
Launched alongside, November 2025
the whole month →
- IB
Life & fun · Nov 2025 · bitsnpieces.dev
- BBoing▲782
Life & fun · Nov 2025 · boing.greg.technology