Struct Model 

pub struct Model<W, T = DynHttpClient> {
    pub wire: W,
    pub transport: T,
}

An endpoint of one provider: a wire bound to a transport.

The pair holds no invariant, so both halves are public. To share one transport across models, clone it; to send a model’s requests another way, replace its transport. The transport defaults to the erased HTTP client, so a model on the default transport is Model<W>.

Fields

wire: W

What to send and how to read the reply.

transport: T

How the payload travels.

Implementations

impl<T> Model<CachedContents, T>

where T: Transport<CachedContents>,

Explicit context-cache operations. Requests targeting an existing handle map HTTP 403 and 404 to ProviderError::CacheExpired.

pub async fn create( &self, request: NewCachedContent, ) -> Result<CachedContent, ProviderError>

Creates cached content and returns its handle and storage usage metadata.

pub async fn get(&self, name: &str) -> Result<CachedContent, ProviderError>

Fetch one cached content by handle.

pub async fn list(&self) -> Result<Vec<CachedContent>, ProviderError>

Every cached content this API key can see, following pagination at the wire’s page size.

pub async fn update_expiry( &self, name: &str, expiry: CacheExpiry, ) -> Result<CachedContent, ProviderError>

Changes cache expiry without modifying its immutable content. Returns the updated resource or an error, including Expired for HTTP 403/404.

pub async fn delete(&self, name: &str) -> Result<(), ProviderError>

Deletes cached content before its expiry. Callers should also clean up task-scoped caches on failure to avoid continued storage charges. Handles other than cachedContents/<id> or bare <id> return ProviderError::Request before dispatch.

impl<T> Model<GenerateContent, T>

pub fn caching(self, book: &CacheBook) -> Model<GenerateContent, Caching<T>>

This model, reading and creating explicit caches as book decides. See crate::providers::gemini::caching.

impl<W, T> Model<W, T>

where W: Wire, T: Transport<W>,

pub fn erase(self) -> DynModel<W::Op>

Erase this model to its operation.

impl<W, T> Model<W, T>

pub fn new(wire: W, transport: T) -> Self

Pair wire with the transport that sends it.

impl<W, T> Model<W, T>

where W: Wire, T: Transport<W>,

pub fn name(&self) -> &str

The wire’s provider descriptor name ("anthropic").

pub fn id(&self) -> Option<&str>

The model id the wire addresses, when the operation addresses one.

pub fn capabilities(&self) -> Capabilities

What a runtime accounts for about this model, such as an embedding model’s width.

pub fn call( &self, request: impl Into<Request<W>>, ) -> impl Future<Output = Result<Response<W>, ProviderError>> + WasmCompatSend + 'static

Send request and fold the whole reply into the operation’s response: Self::stream of a unary reply, finished. A completion takes a prompt, a conversation or a CompletionRequest.

pub fn call_observed( &self, request: impl Into<Request<W>>, observation: AdapterContext, ) -> impl Future<Output = Result<Response<W>, ProviderError>> + WasmCompatSend + 'static

Self::call, with the attempt observed under observation.

pub fn stream( &self, request: impl Into<Request<W>>, ) -> Result<Streamed<W::Op>, ProviderError>

Open a streamed reply. Encoding errors return here; every later failure arrives in-band, as the stream’s last item. Nothing is sent until the stream is first polled.

pub fn stream_observed( &self, request: impl Into<Request<W>>, observation: AdapterContext, ) -> Result<Streamed<W::Op>, ProviderError>

Self::stream, with the attempt observed under observation.

impl<W, T> Model<W, T>

where W: Wire<Op = Embedding>, T: Transport<W>,

pub async fn embed_text(&self, text: &str) -> Result<Embedding, ProviderError>

Embed one text, returning the last vector or an error if none is returned.

impl<W, T> Model<W, T>

where W: Wire<Op = ModelListing>, T: Transport<W>,

pub async fn list(&self) -> Result<ModelList, ProviderError>

Every model the provider lists, every page followed in order.

A failed page names the provider and its request path. Paging stops at a cursor that repeats the one just answered, and after a bounded number of pages.

impl<W, T> Model<W, T>

where W: Wire<Op = Verify>, T: Transport<W>,

pub async fn verify(&self) -> Result<(), ProviderError>

Check that the provider accepts the configured credentials. A 401 or 403 reply is ProviderError::InvalidAuthentication.

impl<T> Model<GenerateContent, T>

pub fn thought_replay(self, replay: ThoughtReplay) -> Self

This model, re-sending thought signatures as replay says. See ThoughtReplay. Call it before caching, which reads it.

impl<T> Model<Responses, T>

pub fn responses_websocket(&self) -> ResponsesWebSocketSessionBuilder

Available on crate feature websocket only.

Start configuring a Responses WebSocket session for this model. Open it with connect_with and a backend, or, under the tungstenite feature, with connect.

impl Model<MockScript, MockRuntime>

pub fn text(text: impl Into<String>) -> Self

Available on crate feature test-utils only.

Create a mock model that returns one text completion.

pub fn from_turns(turns: impl IntoIterator<Item = MockTurn>) -> Self

Available on crate feature test-utils only.

Create a mock model from scripted non-streaming turns.

pub fn from_stream_turns( stream_turns: impl IntoIterator<Item = impl IntoIterator<Item = MockStreamEvent>>, ) -> Self

Available on crate feature test-utils only.

Create a mock model from scripted streaming turns.

pub fn requests(&self) -> Vec<CompletionRequest>

Available on crate feature test-utils only.

Return cloned requests received by this model.

pub fn contexts(&self) -> Vec<Option<AdapterContext>>

Available on crate feature test-utils only.

Return invocation contexts in the same order as the captured requests.

pub fn request_count(&self) -> usize

Available on crate feature test-utils only.

Return the number of requests received by this model.

pub fn script(&self) -> Vec<MockTurn>

Available on crate feature test-utils only.

The non-streaming turns not yet consumed, in order — the read-back half of the script, so a script is serde in and serde out.

pub fn stream_script(&self) -> Vec<Vec<MockStreamEvent>>

Available on crate feature test-utils only.

The streaming turns not yet consumed, in order.

Trait Implementations

impl<W: Clone, T: Clone> Clone for Model<W, T>

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<W: Debug, T: Debug> Debug for Model<W, T>

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

Formats the value using the given formatter.

impl<W: Default, T: Default> Default for Model<W, T>

fn default() -> Self

Returns the “default value” for a type.

impl<W, T> From<Model<W, T>> for DynModel<W::Op>

where W: Wire, T: Transport<W>,

fn from(model: Model<W, T>) -> Self

Converts to this type from the input type.

impl<W: PartialEq, T: PartialEq> PartialEq for Model<W, T>

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl<W: PartialEq, T: PartialEq> StructuralPartialEq for Model<W, T>