AI Potluck
Contribute
Back to Gap Map Infrastructure / Model hubs & registries

MatrixHub

DaoCloud
open source / Overall score: n/a

Self-hosted model registry that works as a drop-in private replacement for Hugging Face Hub in inference infrastructure. Pointing HF_ENDPOINT at it lets the hf CLI, vLLM, SGLang and Dynamo download and upload models, and repositories are reachable over Git, Git LFS and HTTP with project-scoped permissions and version management. It can cache public Hugging Face models on first request, on local or NFS storage. Most of its maintainers work at DaoCloud, and the project describes itself as community-led.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(github.com/matrixhub-ai/matrixhub)
core features withheld
no — governance puts commercial support and enterprise packaging outside the core repository

The whole registry is Apache-2.0 and deploys with Docker Compose or Helm. The project's governance document states that any commercial support, hosted service or enterprise packaging sits outside the core repository, and the repository has no enterprise code.

Adoption

not assessed

MatrixHub runs only as private installations and publishes no install or user count; its adopters page names one organization. GitHub stars would count interest rather than use, so no reading was recorded.

Capability

4 high confidence
4.0

Users create model repositories, version and release them, and push or pull them with the same clients they would use against Hugging Face Hub, over Git or HTTP. It holds models only, with dataset support still on the roadmap and no app hosting, so it sits a step below Hugging Face Hub.

Verified 2026-09-27