Vane is a high-performance, multimodal-native data engine for AI workloads built on a DuckDB fork, offering Python and SQL interfaces and local-to-Ray distributed execution.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Vane Data is a high-performance, multimodal-native data engine for AI workloads. It is built on a fork of DuckDB and extends the core execution engine with native multimodal processing and a unified framework for local and distributed execution. It provides Python and SQL interfaces, and supports built-in AI operations such as invoking LLMs, generating embeddings, and running batch inference through OpenAI and Anthropic APIs or native vLLM integration.
How it works
Vane combines multimodal data processing with asynchronous scheduling to overlap CPU, GPU, I/O, and model inference workloads. It supports local execution and distributed execution across Ray clusters, enabling consistent pipelines from local to distributed environments. It offers a unified type system for modalities (images, video, audio, text, documents, events, sensor data, and tables) and dynamic batching with backpressure control.
Getting started
Installation
pip install vane-ai
Quick Start
Follow the Quickstart guide to build and run your first Vane pipeline.
More Resources
- Examples: https://vane.astrovela.ai/docs/data/examples
- Production deployment: https://vane.astrovela.ai/docs/data/deploy/deployment
Installation prerequisites
Vane supports Python 3.10, 3.11, and 3.12. Python 3.12 is recommended and is the primary development version.
Recent releases
Latest releases: none
Traction
Stars: 40 Forks: 8 Open issues: 7
Behind the repo
N/A
Caveats
License: Apache-2.0 Created: 2026-07-16 Last push: 2026-07-17