Struct Messages 

pub struct Messages<'a> { /* private fields */ }

Available on crate feature async only.

Namespace handle for the Messages API.

Obtained via Client::messages; cheap to construct (just borrows the client).

Implementations

impl<'a> Messages<'a>

pub async fn create(&self, request: CreateMessageRequest) -> Result<Message>

Send a request to POST /v1/messages and return the full response.

Retries are governed by the RetryPolicy configured on the client.

pub async fn create_with_beta( &self, request: CreateMessageRequest, betas: &[&str], ) -> Result<Message>

Like Self::create but with additional per-request beta headers merged into anthropic-beta.

pub async fn count_tokens( &self, request: CountTokensRequest, ) -> Result<CountTokensResponse>

Count the input tokens that would be billed for the given request, without invoking the model.

pub async fn count_tokens_with_beta( &self, request: CountTokensRequest, betas: &[&str], ) -> Result<CountTokensResponse>

Like Self::count_tokens but with per-request beta headers.

pub fn dry_run(&self, request: &CreateMessageRequest) -> Result<DryRun>

Render the HTTP request that Self::create would send, without firing it. Returns a DryRun containing the rendered method, URL, headers, and JSON body. Useful for tests, debugging, and reproducing requests as curl commands.

pub fn dry_run_with_beta( &self, request: &CreateMessageRequest, betas: &[&str], ) -> Result<DryRun>

Like Self::dry_run but with per-request beta headers merged in, matching Self::create_with_beta.

pub async fn cost_preview( &self, request: &CreateMessageRequest, pricing: &PricingTable, ) -> Result<CostPreview>

Available on crate feature pricing only.

Estimate the USD cost of a request before sending it.

Hits /v1/messages/count_tokens to get the exact server-counted input-token total, then uses request.max_tokens as the output upper bound. The returned CostPreview exposes input / max-output / max-total cost components.

Pre-flight charge from this call: one count-tokens request (significantly cheaper than a full generation, but not free).

pub async fn cost_preview_cached( &self, request: &CreateMessageRequest, pricing: &PricingTable, cache: &CountTokensCache, ) -> Result<CostPreview>

Available on crate feature pricing only.

Like Self::cost_preview but consults a CountTokensCache before issuing the network call. Cache hits avoid the count_tokens round-trip entirely. Useful for IDE integrations and long-running agents that preview the same prompt many times.

pub fn dry_run_count_tokens( &self, request: &CountTokensRequest, ) -> Result<DryRun>

Render the HTTP request that Self::count_tokens would send.

pub fn dry_run_count_tokens_with_beta( &self, request: &CountTokensRequest, betas: &[&str], ) -> Result<DryRun>

Like Self::dry_run_count_tokens but with per-request beta headers.

pub async fn create_stream( &self, request: CreateMessageRequest, ) -> Result<EventStream>

Available on crate feature streaming only.

Open a streaming connection to POST /v1/messages and return an EventStream.

The returned stream yields StreamEvents as they arrive; call EventStream::aggregate to drive it to completion and reconstruct the final Message.

Streaming requests are not retried.

pub async fn create_stream_with_beta( &self, request: CreateMessageRequest, betas: &[&str], ) -> Result<EventStream>

Available on crate feature streaming only.

Like Self::create_stream but with per-request beta headers.