memlocal 0.1.0

Local-first, on-device cognitive memory for AI apps — a Flutter binding to the Rust memlocal engine (CozoDB) with pluggable OpenAI + Jina providers.

Changelog

0.1.0

Initial release of the Flutter/Dart binding to the memlocal cognitive memory engine.

• FFI binding to the Rust engine via flutter_rust_bridge, built from source with cargokit:
  • Open a persistent engine (Memlocal.open) backed by a database file, or an in-memory engine (Memlocal.openInMemory).
  • Store a memory with a typed classification and a caller-supplied embedding (addMemory).
  • Semantic (HNSW) vector search over stored memories (searchSemantic).
  • Count stored memories (memoryCount).
• Bring-your-own provider model with abstractions for embeddings, LLM completion, and reranking, plus shipped implementations:
  • OpenAIEmbeddingProvider and OpenAILlmProvider (OpenAI HTTP APIs).
  • JinaReranker (Jina AI rerank API).
• Interactive memory-chat example that stores typed memories, recalls them with optional Jina reranking, and answers grounded in recalled memory.
• Platforms: Android and iOS, built from source via flutter_rust_bridge + cargokit (a Rust toolchain with the relevant targets is required).

This is an early release: the FFI surface is a focused subset of the engine and the API will evolve in subsequent versions.