Grafial started as an internal tool at Iridae.
We needed a way to reason about graphs where the structure itself is uncertain—where nodes, edges, and attributes represent beliefs that should evolve as new evidence arrives. Existing graph tools and probabilistic systems each covered part of the problem, but not the whole thing in a way that fit how we wanted to work.
So we built Grafial: a domain-specific language and runtime for Bayesian reasoning over graphs.
What makes it unusual is that uncertainty is built into the graph model itself. You can define probabilistic relationships, accumulate evidence over time, and run rules and inference flows against the resulting posterior graph.
It is undeniably niche. It is also one of those tools that feels obvious once you need it, and we have not seen much else quite like it.
Grafial is now open source, and the browser-based Composer is available to try: