{{product.name}} is a set of open-source services for planning and approval procedures. Each service runs on its own, and the bundled UI is one way to use them.
An applicant's documents arrive. {{product.name}} parses them, reads scanned pages and indexes every section for search.
The UI calls the services, the services read the graph and the data stores, and the graph is built from official legislation. Select a layer to see what it holds.
So far {{product.name}} has been deployed to a development environment on Azure Kubernetes Service. Staging and production workflows are defined but have not run, and other hosting has not been tested.
They come from the code and the team's own engineering guidance. Where one is not fully met yet, we say so.
Each backend has its own project, Dockerfile and Helm chart, and can be used without the UI.
Services name models by gateway alias, so changing a model is a gateway change, not a code change. Two settings still bypass the gateway today.
Fetching and parsing legislation differ by country. The graph model, extraction and storage are shared, so a new country needs a parser and a registry entry, not a fork.
Every AI step is meant to be explainable and reviewable by a caseworker. Explainability and feedback services, in development, support that.
Most services that call an LLM pass inputs through a prompt-injection check, based on OWASP LLM01, before they reach the model.
Business events go through bizlog, a structured audit-logging library. Durable storage for the audit trail is in progress.
Book a walkthrough with the team. We show you the services that fit your procedures, and what is still on the roadmap.
