Usage

(require '[embeddings.core :as emb]
         '[embeddings.math :as emb.math])

;; load-model returns an AutoCloseable-style handle; with-model scopes it
(emb/with-model [model "models/all-MiniLM-L6-v2" {:pooling :mean :normalize? true}]
  (emb/dimension model)                    ;; => 384
  (let [a (emb/embed model "A cat sits on the mat")
        b (emb/embed model "A kitten rests on the rug")
        c (emb/embed model "The stock market crashed today")]
    (emb.math/cosine-similarity a b)       ;; => ~0.6+ (similar)
    (emb.math/cosine-similarity a c)))     ;; => ~0.1  (unrelated)

;; batch (padding is attention-mask aware; results match single embeds)
(emb/with-model [model "models/all-MiniLM-L6-v2"]
  (emb/embed-batch model ["first text" "second text" "third text"]))
;; => [float[384] float[384] float[384]]

Options for load-model with default values:

option	default	meaning
:pooling	:mean	:mean (mask-weighted), :mean-sqrt-len (sum / sqrt of token count), :cls, or :max over token embeddings
:normalize?	true	L2-normalize output vectors (unit length, ready for cosine)
:max-length	512	truncate inputs to this many tokens
:execution-providers	none (CPU)	ONNX Runtime execution providers to try, e.g. [:coreml] or [{:provider :cuda :device-id 0}]; also :rocm, :tensorrt, :directml, :xnnpack

Execution providers need an onnxruntime build that includes them. The default com.microsoft.onnxruntime/onnxruntime artifact is CPU-only, and CUDA needs onnxruntime_gpu. If you request a provider that the runtime does not supply, the library throws ex-info with {:embeddings/error :execution-provider-unavailable}.

Some ONNX graphs already output a pooled [batch, hidden] sentence embedding. The library detects these models automatically and ignores :pooling.

embeddings.math supplies the small vector functions: dot, norm, l2-normalize, and cosine-similarity. All of them operate on primitive float[].

Errors are ex-info maps keyed :embeddings/error (:model-not-found, :tokenizer-not-found, :model-closed, :unsupported-input, :unsupported-output, :dim-mismatch).