Struct ContentSource 

pub struct ContentSource {
    pub kind: ContentSourceKind,
    pub trust_level: ContentTrustLevel,
    pub identifier: Option<String>,
    pub memory_hint: Option<MemorySourceHint>,
}

Provenance metadata attached to a piece of untrusted content.

Created at the call-site (tool executor, MCP adapter, A2A handler, etc.) to describe where content came from. Passed into ContentSanitizer::sanitize] alongside the raw content so the pipeline can choose the appropriate spotlight wrapper and injection detection sensitivity.

Examples

use zeph_sanitizer::{ContentSource, ContentSourceKind, ContentTrustLevel, MemorySourceHint};

// Basic source for a shell tool result.
let source = ContentSource::new(ContentSourceKind::ToolResult)
    .with_identifier("shell");
assert_eq!(source.trust_level, ContentTrustLevel::LocalUntrusted);
assert_eq!(source.identifier.as_deref(), Some("shell"));

// Memory retrieval with a hint to skip injection detection for conversation turns.
let mem_source = ContentSource::new(ContentSourceKind::MemoryRetrieval)
    .with_memory_hint(MemorySourceHint::ConversationHistory);
assert!(mem_source.memory_hint.is_some());

Fields

kind: ContentSourceKind

The category of this content source.

trust_level: ContentTrustLevel

Trust tier that drives the spotlight wrapper choice.

identifier: Option<String>

Optional identifier: tool name, URL, agent ID, etc. Used in spotlight attributes.

memory_hint: Option<MemorySourceHint>

Optional hint for memory retrieval sub-sources. When Some, modulates injection detection sensitivity in ContentSanitizer::sanitize]. Non-memory sources leave this as None — full detection applies.

Implementations

impl ContentSource

pub fn new(kind: ContentSourceKind) -> ContentSource

Create a new source with the default trust level for the given kind.

Examples

use zeph_sanitizer::{ContentSource, ContentSourceKind, ContentTrustLevel};

let source = ContentSource::new(ContentSourceKind::WebScrape);
assert_eq!(source.trust_level, ContentTrustLevel::ExternalUntrusted);
assert!(source.identifier.is_none());

pub fn with_identifier(self, id: impl Into<String>) -> ContentSource

Set the identifier for this source (tool name, URL, agent ID, etc.).

The identifier appears in the spotlight wrapper’s XML attributes so the LLM can see where the content came from (e.g. name="shell", ref="https://example.com").

Examples

use zeph_sanitizer::{ContentSource, ContentSourceKind};

let source = ContentSource::new(ContentSourceKind::ToolResult)
    .with_identifier("shell");
assert_eq!(source.identifier.as_deref(), Some("shell"));

pub fn with_trust_level(self, level: ContentTrustLevel) -> ContentSource

Override the trust level for this source.

Use when the call-site has more context about the actual origin of the content than the default derived from the source kind.

Examples

use zeph_sanitizer::{ContentSource, ContentSourceKind, ContentTrustLevel};

// Elevate trust for a verified internal source.
let source = ContentSource::new(ContentSourceKind::McpResponse)
    .with_trust_level(ContentTrustLevel::LocalUntrusted);
assert_eq!(source.trust_level, ContentTrustLevel::LocalUntrusted);

pub fn with_memory_hint(self, hint: MemorySourceHint) -> ContentSource

Attach a memory source hint to modulate injection detection sensitivity.

Only meaningful for ContentSourceKind::MemoryRetrieval sources.

Trait Implementations

impl Clone for ContentSource

fn clone(&self) -> ContentSource

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ContentSource

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.