cortexkit-model-batch 0.2.0

Batch rules shared by the CortexKit model-batch roles: error vocabulary and Workers AI code table, batch identity, item images, conservative cost estimate.
Documentation

cortexkit-model-batch

The batch rules every CortexKit model-batch role shares (classifier/v1, extractor/v1): the error vocabulary with the Cloudflare Workers AI code table, batch identity and its canonical hashing, the item image shape and its validation, the conservative cost estimate, and the absent-is-not-zero and number-text helpers.

RULES.md is normative: it states each shared rule once, and each role contract points to it. A role crate re-exports what it uses, so a module depends on its role crate and rarely on this one directly.

[dependencies]
cortexkit-model-batch = "0.2.0"

To include the vector bytes in your build, enable test-vectors:

[dev-dependencies]
cortexkit-model-batch = { version = "0.2.0", features = ["test-vectors"] }

The feature exposes cortexkit_model_batch::vectors::FILES and a constant for each file. The vectors live in test-vectors/model-batch-v1/ and ship inside the crate package, pinned by SHA256SUMS. The crate's own vector tests need the feature: cargo test -p cortexkit-model-batch --features test-vectors.