flutter_edge_ai_qdrant 2.0.0 copy "flutter_edge_ai_qdrant: ^2.0.0" to clipboard
flutter_edge_ai_qdrant: ^2.0.0 copied to clipboard

qdrant-edge on-device vector-store provider for flutter_edge_ai_rag, via the official qdrant_edge UniFFI SDK. Native platforms only.

2.0.0 #

  • Breaking: now a flutter_edge_ai_rag storage provider; a location binds one embedding profile.

1.3.2 #

  • Renamed from flutter_gemma_rag_qdrant.
  • Use qdrant_edge 0.8.0-dev.4 to prevent native result and argument leaks.
  • README: how to install the agent skills that cover this package.

1.3.1 #

  • Add flush(); without it an index was lost when the process ended (#492).

1.3.0 #

  • Breaking: moved onto the official qdrant_edge UniFFI SDK.
  • Breaking: a 1.x store is not readable — remove its files, then re-index.
  • Breaking: clear() empties the shard in place; it never deletes files.
  • Breaking: clear() cannot reset the vector size; switching embedders needs a new path.
  • Android x86_64 is now provisioned alongside arm64 (emulators, Chromebooks).
  • A re-opened store now reports its contents without waiting for a write.
  • A shard that cannot be opened is reported instead of read as empty.
  • A shard held open elsewhere is reported as such, not as damaged.
  • Concurrent addDocument calls no longer race each other's shard open.
  • configure() rejects a field name that collides with a stored payload key.

1.2.0 #

  • Encode filters by the declared field type, so both backends answer alike.
  • Breaking: reject a field name containing . — qdrant reads it as a nested path.
  • Fix numeric equality never matching a float payload (4 vs 4.0).
  • Fix an unbounded FieldRange excluding every non-numeric document instead of matching all.

1.1.0 #

  • Fix metadata filtering: declared FilterSchema fields are promoted to top-level payload keys, so Filter predicates actually match (previously narrowed to zero).
  • Implement configure(FilterSchema); requires flutter_gemma ^1.1.0.
  • Correct the "~75×" search-speed claim to the re-measured ~5–11× vs the new in-SQLite vec0 store (was vs the deleted Dart brute-force path).

1.0.1 #

  • Point homepage to fluttergemma.dev. No code change.

1.0.0 #

  • Stable 1.0.0.
  • Rebuild Android .so with 16KB page alignment for Android 15 / Play target SDK 35+ (#319; native tag qdrant-edge-v0.7.3).

1.0.0-rc.1 #

  • Initial release: qdrant-edge on-device RAG vector store for flutter_gemma (native FFI; ~75× faster search than the legacy Dart brute-force sqlite store).
  • Provides QdrantVectorStore; implements VectorStoreRepository. Honors the payload-aware Filter DSL.
  • Native platforms only (Android, iOS, macOS, Linux, Windows); no web (use flutter_gemma_rag_sqlite).
0
likes
160
points
81
downloads

Documentation

API reference

Publisher

verified publishersashadenisov.dev

Weekly Downloads

qdrant-edge on-device vector-store provider for flutter_edge_ai_rag, via the official qdrant_edge UniFFI SDK. Native platforms only.

Homepage
Repository (GitHub)
View/report issues

Topics

#rag #qdrant #vector-search #embeddings #on-device

License

MIT (license)

Dependencies

flutter, flutter_edge_ai_rag, path, qdrant_edge, uuid

More

Packages that depend on flutter_edge_ai_qdrant