Struct VertexAiRagMemoryService 

pub struct VertexAiRagMemoryService { /* private fields */ }

Memory service backed by Vertex AI RAG.

Stores memory as RAG files in a corpus (via ragFiles:upload) and uses semantic search (:retrieveContexts) for retrieval.

Quick start

let svc = VertexAiRagMemoryService::new(
    VertexAiRagMemoryConfig::from_corpus(
        "projects/my-proj/locations/us-central1/ragCorpora/my-corpus",
    ),
)
.with_token("ya29.my-access-token");

let hits = svc.search("session-1", "what did the user order?").await?;

Implementations

impl VertexAiRagMemoryService

pub fn new(config: VertexAiRagMemoryConfig) -> Self

Create a new Vertex AI RAG memory service.

No auth token is configured yet — call with_token or with_token_refresher before issuing requests.

pub fn with_token(self, token: impl Into<String>) -> Self

Set a static bearer token for all requests.

pub fn with_token_refresher( self, f: impl Fn() -> String + Send + Sync + 'static, ) -> Self

Set a dynamic token refresher closure (invoked before every request).

pub fn corpus(&self) -> &str

Returns the configured corpus resource name.

pub async fn add_session_to_memory( &self, app_name: &str, user_id: &str, session_id: &str, entries: &[MemoryEntry], ) -> Result<(), MemoryError>

Ingest a batch of memory entries for a session, mirroring ADK’s add_session_to_memory. Each entry is serialised as one JSON line and the whole batch is uploaded as a single RAG file tagged with an encoded display name.

app_name / user_id / session_id are encoded into the file display name so that search can later filter by scope.

pub async fn search_memory( &self, app_name: &str, user_id: &str, query: &str, ) -> Result<Vec<MemoryEntry>, MemoryError>

Search memory scoped to a specific app_name / user_id, mirroring ADK’s search_memory. Filters retrieved contexts by the encoded display name and parses the per-event JSON lines back into MemoryEntrys.

Trait Implementations

impl MemoryService for VertexAiRagMemoryService

fn store<'life0, 'life1, 'async_trait>( &'life0 self, session_id: &'life1 str, entry: MemoryEntry, ) -> Pin<Box<dyn Future<Output = Result<(), MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Ingest a single entry into the corpus. The session_id scopes the uploaded file’s display name (app/user default to the entry key / session id when not separately tracked).

fn get<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, _session_id: &'life1 str, _key: &'life2 str, ) -> Pin<Box<dyn Future<Output = Result<Option<MemoryEntry>, MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Vertex AI RAG doesn’t support direct key-based retrieval — use search instead.

fn list<'life0, 'life1, 'async_trait>( &'life0 self, _session_id: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<Vec<MemoryEntry>, MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Listing is not supported by the RAG retrieval API (semantic only).

fn search<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, _session_id: &'life1 str, query: &'life2 str, ) -> Pin<Box<dyn Future<Output = Result<Vec<MemoryEntry>, MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Semantic search against the corpus via :retrieveContexts. Results are not scope-filtered here (use search_memory for app/user scoping).

fn delete<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, _session_id: &'life1 str, _key: &'life2 str, ) -> Pin<Box<dyn Future<Output = Result<(), MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Deleting an individual RAG file by key is not supported by the retrieval API — returns MemoryError::Unsupported rather than silently succeeding (which would mask data that was never removed).

fn clear<'life0, 'life1, 'async_trait>( &'life0 self, _session_id: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<(), MemoryError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Clearing a corpus is a destructive admin operation performed out of band — returns MemoryError::Unsupported rather than silently succeeding.