Citation policy capability

CitationPolicyCapability requires every answer to declare what grounds it. The RAG capability's instructions ask the model to cite and record what it cites, but nothing enforces the declaration without this capability.

Register it alongside the RAG capability, with the host carrying capability state between runs, as in Compose an agent. Without that state a follow-up about evidence cited earlier goes unenforced. An agent takes at most one policy capability. Pydantic AI rejects a second.

Declaring nothing is a valid answer

A model that finds nothing relevant calls the cite tool with an empty list. That records the answer as ungrounded, which is distinct from an answer that declared nothing at all (missing). A model can always satisfy the policy without inventing grounding.

What happens when a question ends undeclared

The model is asked once to record what grounded the answer it already gave. It is not asked to change the answer. If the cite tool is no longer available by then, or the question finishes undeclared anyway, the question is recorded in CitationPolicyState.violations under the "citation_policy" state key, and the model is not asked again.

Which answers are enforced

Every answer in a conversation that has something to declare: either this question retrieved evidence, or the conversation has already cited something, which stays available to later answers. A follow-up about evidence cited earlier is enforced even though it searched nothing.

Once anything has been cited, later turns are enforced too, a greeting included. The model satisfies the policy by citing an empty list, at the cost of one extra request. A conversation with neither a current-question evidence outcome nor any earlier citation is not enforced.