Struct ContentProcessor 

pub struct ContentProcessor { /* private fields */ }

Orchestrates processContent evaluation over a batch of messages. See the module docs for how this differs in scope from Python’s ScopedContentProcessor.

Implementations

impl ContentProcessor

pub fn new(client: PurviewClient) -> Self

pub async fn evaluate( &self, messages: &[Message], settings: &PurviewSettings, provided_user_id: Option<&str>, ) -> Result<(bool, Option<String>)>

Evaluate messages for policy violations, resolving the user id per resolve_user_id (provided_user_id lets a response-phase evaluation reuse the id resolved during the prompt phase, matching Python’s process_messages(..., user_id=resolved_user_id)).

Returns (should_block, resolved_user_id). One processContent call is made per message, in order, stopping at (and including) the first one that returns a block verdict — mirrors ScopedContentProcessor.process_messages’s for req in pc_requests: ...; if should_block: break.

Fails (rather than silently allowing) when tenant_id or purview_app_location aren’t set on settings, mirroring Python’s _map_messages raising ValueError in the same situation — an error a caller with ignore_exceptions = true treats as fail-open (see crate::middleware), same as Python.