Juno is maintained by a small team. Financial support allows maintainers to dedicate time to the core engine, fund GPU-enabled CI infrastructure, and sustain long-term development. All expenditure is publicly itemised on Open Collective.
Three options depending on your preferred accounting workflow. All contributions — individual or corporate — are acknowledged in the README and release notes.
The primary channel for individual developers and companies. One-time and recurring contributions. GitHub waives platform fees for open-source sponsored accounts — verify current terms before contributing.
All income and expenses are publicly visible to anyone. Suitable for teams and companies that require invoice documentation. One-time and recurring tiers available.
For companies preferring a direct transfer without platform fees. Contact us for wire transfer details. A formal invoice can be issued on request.
All expenditure via Open Collective is publicly itemised. Typical allocations:
Stipends for development time on core engine work — scheduler, generation loop, GPU backends.
GPU-enabled build runners for CUDA and ROCm test coverage. Benchmarks don't lie without hardware.
Periodic security audit engagements and legal review of licenses and contributor agreements.
Conference travel for project representation and ecosystem outreach.
Pull requests, benchmark reports, and reproduction cases all move the project forward. No CLA required for small patches; see docs/CLA.md for larger contributions.
Fork the project, pick up an open board issue, move it to QA and add a comment referencing your fork. Follow existing test coverage patterns — unit tests for business logic first.
Send your performance report to dev@ml.cab. Include: GPU card details, Juno startup command, conversation log, and the JFR Metrics Summary section from your run.
Tried Juno on your setup and found a rough edge? Open an issue on the board. Specific, reproducible reports with version and hardware details move things forward fastest.
Contributing fine-tuning datasets or LoRA adapters trained on Juno? Open a discussion on GitHub. Domain-specific adapters tested against our benchmark matrix are particularly valuable.
The Juno project actively pursues open-source and research grants. If you represent a research institution or NGO interested in a joint application, contact dev@ml.cab with subject "Grant Partnership Inquiry".
| Programme | Notes |
|---|---|
| NLnet Foundation (NGI Zero) | EU-funded; covers privacy-preserving and open infrastructure software |
| Sovereign Tech Fund | Germany; funds maintenance of open digital infrastructure critical to the ecosystem |
| Open Technology Fund | US-based; Internet freedom and open technology focus |
| Linux Foundation / LF AI & Data | Project affiliation grants; technical governance support |
| Apache Software Foundation | Infrastructure and release tooling support for Apache-licensed projects |
| EU Horizon Europe (MSCA/RIA) | Research and innovation actions; applicable via academic partner consortium |
| Mozilla Foundation Grants | Open internet and privacy-respecting AI infrastructure |
The project maintains the following to meet eligibility requirements across most programmes:
Financial support, code, benchmarks, or a bug report — every contribution is counted.