flutter_gemma_embeddings 2.1.1
flutter_gemma_embeddings: ^2.1.1 copied to clipboard
Runtime-agnostic on-device text embedding pipeline for flutter_gemma: tokenization, isolate worker, pooling and normalization. Pair with an engine package for a backend.
2.1.1 #
- Lift the
dart_sentencepiece_tokenizercap — 1.4.1 is now required, not excluded.
2.1.0 #
- Add a SigLIP2 text-embedding profile (no BOS, single trailing EOS, lowercased, fixed 64-token width).
- No longer pulls
dart:iointo the web graph;fromPaththrowsUnsupportedErrorthere. - Cap
dart_sentencepiece_tokenizerbelow 1.4.0, which rejects EmbeddingGemma's and SigLIP2'stokenizer.json.
2.0.0 #
- BREAKING: runtime-agnostic embedder;
LiteRtEmbeddingBackendmoved to flutter_gemma_litertlm. - Add tokenizer-factory seam (
ForwardPassDescriptor.tokenizerFactory) so engines can bring WordPiece too. - Add
EmbeddingForwardPass.outputContractoverride +ForwardResult.attentionMaskfor the mask thread. - Add
WordPieceEmbeddingTokenizer(BERT-style, e.g. MiniLM) alongside the existing Gemma SentencePiece adapter.
1.0.4 #
- internal: run on the flutter_gemma_litertlm LiteRT engine (no API change).
1.0.3 #
- Migrate to LiteRT v0.14.0 3-arg
LiteRtCreateModelFromFile(addsLiteRtEnvironment); native-v0.14.0.
1.0.2 #
- Realign the shared LiteRT-LM native bundle to native-v0.13.1-b (#364). No API change.
1.0.1 #
- Rebuild on native-v0.13.1-a (shares the LiteRT-LM native bundle; restores NPU dispatch libs, #155). No embeddings API change.
- Point
homepageto fluttergemma.dev. No code change.
1.0.0 #
- Stable 1.0.0; spec imports redirected off the
dart:iomobile lib for a wasm-clean web graph.
1.0.0-rc.1 #
- Initial release: on-device text embeddings (Gecko / EmbeddingGemma
.tflite) via the LiteRT C API + dart:ffi. - Provides
LiteRtEmbeddingBackend(EmbeddingBackendProvider); forward pass runs on a background isolate. - Autonomous (no dependency on flutter_gemma_litertlm); shares the LiteRT-LM native library when both are present.
- Android, iOS, macOS, Linux, Windows + web (LiteRT.js).