AI assistant
The AI assistant is an experimental feature and is disabled by default. Its behavior, configuration, and supported models can change in later releases.
The AI assistant helps you understand, check, test, and write ALFA policies. Questions are asked in plain language in a panel next to your work, and answers are based on the policies and attributes of the open project.
The AI assistant uses a large language model. Its answers can be incomplete or incorrect, and AI-generated content may contain bugs, vulnerabilities, or license restrictions. Every proposed change must be reviewed and tested before it is saved or published. The assistant does not replace your own review or your organization's approval process.
What the assistant can do
- Explain what a policy, policy set, or rule does, and why a request is permitted or denied.
- Locate the policy where a new rule belongs.
- Validate a policy and point to the line of each error.
- Test a policy against example requests, using the same ALFA compiler and decision engine as Authorization Hub.
- Change a policy: propose an edit and offer it to be applied in the editor.
- Look up the attributes and projects visible to you.
Only the data permitted by your project role is read by the assistant. The assistant never saves or publishes anything.
Who can use the assistant
The AI assistant can be opened by every signed-in user. To save a proposed change, the Project editor or Project admin role is required. Read the Projects section for the project roles.