Module batches 

The Batches API: submit a batch of message requests, poll for completion, stream per-request results.

Anthropic’s batch endpoint is the cheapest way to run large fan-out workloads (50% off vs. per-request pricing) at the cost of higher latency. This module wraps the full surface:

• Batches::create – submit
• Batches::get – status (polling-friendly)
• Batches::list / Batches::list_all – enumerate
• Batches::cancel, Batches::delete
• Batches::wait_for – poller that returns once ended_at is set
• Batches::results / Batches::results_stream – decode the JSONL results body, eagerly into a Vec or lazily as a Stream

The batch ID is the only state you need to durably persist; reattach later by calling Batches::get(id) or Batches::wait_for(id, _).

Quick start

use claude_api::{Client, batches::{BatchRequest, BatchResultPayload, WaitOptions},
    messages::CreateMessageRequest, types::ModelId};
let client = Client::new(std::env::var("ANTHROPIC_API_KEY").unwrap());
let requests = vec![
    BatchRequest::new("q1",
        CreateMessageRequest::builder()
            .model(ModelId::HAIKU_4_5).max_tokens(32).user("2 + 2?").build()?),
];
let batch = client.batches().create(requests).await?;
let finished = client.batches()
    .wait_for(&batch.id, WaitOptions::default()).await?;
let items = client.batches().results(&finished.id).await?;
for item in &items {
    if let BatchResultPayload::Succeeded { message } = &item.result {
        println!("{}: {} tokens", item.custom_id, message.usage.output_tokens);
    }
}

Re-exports

pub use types::BatchDeleted;

pub use types::BatchRequest;

pub use types::BatchResultItem;

pub use types::BatchResultPayload;

pub use types::ListBatchesParams;

pub use types::MessageBatch;

pub use types::ProcessingStatus;

pub use types::RequestCounts;

pub use types::WaitOptions;

pub use api::Batches;async

Modules

apiasync

The async Batches<'a> namespace.

types

Wire types for the Batches API.