July 3, 2026·5 min read·The Reliafy Team

Reliafy is now open source

Today we're making Reliafy open source. The complete core product — life-data modelling, reliability block diagrams, maintenance strategy, and dataset management — is now available under the AGPL-3.0 license, and you can run the whole thing on your own hardware with a single command:

git clone https://github.com/Reliafy/reliafy
cd reliafy
docker compose up --build

Open http://localhost:8000 and you're in. No account, no sign-up, no telemetry — a persistent database in a Docker volume, sample datasets and a worked reliability block diagram already seeded, and every feature of the core toolkit ready to use.

Why open source?

Reliability engineering happens in places the cloud can't always reach.

If you've worked in this field, you know the pattern: the failure data that matters most lives inside defense contractors, mining operations, utilities, rail networks, and manufacturing plants — organisations with strict rules about where data goes. Sometimes it's policy. Sometimes it's an air-gapped network. Sometimes it's simply that the data describes critical infrastructure and nobody wants it on someone else's servers, however good the encryption story.

A cloud-only tool tells those engineers: sorry, not for you. We think that's the wrong answer. The people with the strictest environments are often the ones who need good reliability tooling the most.

There's a second reason, and it matters just as much: you should be able to check the statistics. When a tool tells you the optimal replacement interval for a component, or that a Weibull shape parameter is 1.8 with a given confidence bound, that number feeds real maintenance decisions with real costs. With Reliafy you can now read exactly how every figure is computed — the fitting is built on SurPyval and RePyability, both open source, and now the application around them is too. No black boxes between your data and your decisions.

What's in the open-source version

Everything that makes Reliafy useful, with no feature gates:

Life-distribution modelling. Fit Weibull, Lognormal, Exponential, Gamma, Normal, and proportional-hazards models to exact, censored, and truncated data. Probability plots on distribution-specific paper, confidence bounds, goodness-of-fit, and a survival calculator.
Reliability block diagrams. A drag-and-drop canvas for series, parallel, k-out-of-n, and standby structures, computing system reliability, MTTF, importance measures, and minimal path and cut sets.
Maintenance strategy. Rank candidate distributions against your data, compare two designs head-to-head, and find the cost-optimal preventive replacement interval.
Datasets. Upload a CSV once, reuse it across models, and keep everything organised.

There are no artificial limits in the self-hosted version — no capped model counts, no locked features, no nag screens. It runs in single-user mode: your instance, your data, your machine.

What the cloud version is for

Reliafy Cloud is the same product, hosted and operated by us, for engineers and teams who want zero setup. It adds the things that only make sense hosted: user accounts, managed persistence, and the AI assistant — a chat panel that can fit models, save datasets, and build reliability block diagrams for you, with no API key or configuration on your side.

AI usage on the cloud runs on credits, included in the plans. Every new account gets starter credits to try the assistant. The Pro plan (US$19/month) includes 1,000 AI credits every month on top of unlimited saved datasets, models, and RBDs; if you need more, top-up packs start at $5 and credits never expire. You'll never see a surprise bill — usage simply draws down a balance you can see at all times.

We want to be straightforward about the model, because open-core projects sometimes aren't: it's one repository. The open-source version isn't a stripped-down community edition maintained on the side; it's the same code we deploy to production. When we fix a bug in the cloud, the fix lands in the public repository, because they are the same place. Team workspaces are next on the cloud roadmap.

Why AGPL?

We chose the GNU Affero General Public License deliberately. For you as a user, it means the four freedoms with no fine print: use it, study it, modify it, share it — free forever, for any purpose, commercial included.

The "Affero" part adds one condition that protects the project: if someone offers a modified Reliafy as a network service, they must publish their modifications. You can self-host for your team, your plant, or your entire company without sharing anything. What the license discourages is a third party taking the code, improving it privately, and selling it back as a closed competing service. Plausible, Metabase, and Cal.com made the same choice for the same reason, and we think it's the honest middle ground: genuinely open for users, defensible as a project.

Getting started, contributing, and what's next

The README covers the details: the docker-compose quickstart, the full configuration reference for self-hosting, and the development setup if you want to hack on it — the backend is FastAPI and the frontend is React, and the test suite runs with plain pytest.

Bug reports and pull requests are welcome. If you find a security issue, please use the private reporting path in SECURITY.md rather than a public issue. And if Reliafy is useful to you, starring the repository genuinely helps other engineers find it.

Next on the roadmap: team workspaces for the cloud, more distribution and strategy tooling in the core, and continued polish on the RBD builder. The best way to influence what comes after that is to open an issue and tell us what your reliability workflow actually needs.

Run it, break it, tell us what's missing. It's yours now too.