On-device data and AI for .NET MAUI and .NET 10
On-device data and AI for .NET MAUI and .NET 10: SQLite with sqlite-vec,
ONNX Runtime embeddings, a Microsoft.Extensions.VectorData store, document
ingestion, and local chat with retrieval-augmented generation. Everything runs
in-process on the device. There is no service to call, nothing leaves the
phone, and nothing is downloaded without the user's consent.
Install
dotnet add package Qavren.Edge --prerelease # hosting core + SQLite + the native library
dotnet add package Qavren.Edge.Maui --prerelease # MAUI lifecycle bridge and app paths
Then add what you use: Qavren.Edge.Embeddings.Onnx, Qavren.Edge.VectorData,
Qavren.Edge.Ingestion (with .Onnx, .Pdf, .OpenXml), Qavren.Edge.Chat.Onnx,
Qavren.Edge.Rag. Every package is MIT except that Qavren.Edge.Ingestion.Pdf
depends on PdfPig (Apache-2.0), which is why PDF support is a separate package.
Where to go
- Getting started: the whole stack composed once, then index, search and ask.
- Guides: hosting core and SQLite, embeddings and the vector store, ingestion, chat and RAG.
- Error codes: every
EdgeExceptioncarries a numbered code and a remediation. - API reference for every package, and the MAUI package.
- Decisions: the architecture decision records for each area, in the sidebar.
Supported platforms
| Adopting | Android | iOS / Mac Catalyst |
|---|---|---|
| SQLite and hosting only | 21 | 15.0 |
| + embeddings, vector store, ingestion | 24 | 15.1 |
| + chat | 24 | 15.4 |
Windows x64 and arm64, Linux x64 and arm64 and macOS arm64 are supported for plain .NET 10. Source, issues and releases: github.com/qavren-oss/qavren-edge.