DVAIBridge
Embeds an OpenAI-compatible HTTP server inside your .NET app process (127.0.0.1:38883, /v1/chat/completions, /v1/embeddings, /v1/models, SSE streaming). Any OpenAI client speaks to it — the OpenAI .NET SDK, Semantic Kernel, LangChain.NET, anything pointing a `baseUrl` at an HTTP endpoint. Why not Ollama / LM Studio / llama-server? Those are servers your END USERS install. DVAIBridge ships INSIDE your app's binary — `dotnet add package DVAIBridge`, build, run. Zero setup for the user. Cross-platform: net10.0 (Desktop) / net10.0-ios26.5 / net10.0-android36.0. Mac Catalyst is temporarily off in v4.2.x while LiteRT-LM lands — coming back in v4.3.0. Backends: llama.cpp (Metal / CUDA / Vulkan / DirectML), Apple Foundation Models, CoreML / ANE, MLX, MediaPipe LLM, LiteRT, ONNX Runtime GenAI, ML.NET — engine selected at runtime by device capability. Same OpenAI wire as the dvai-bridge family on iOS (CocoaPods), Android (Maven Central), Flutter (pub.dev), React Native + Capacitor + browser + Node (npm). One library, every client platform, identical request/response shapes. Reactive progress via IAsyncEnumerable<ProgressEvent>. Optional LAN-mDNS + self-hostable WebSocket rendezvous for transparent peer offload. Docs + architecture: https://bridge.deepvoiceai.co
View on Github
Activity
- Latest release
- 2mo ago
- Total releases
- 5
- Cadence
- ~12 days
- Last 12 months
- 5
Reach
- Stars
- —
Details
- First release
- May 20, 2026
| Version | Released | |
|---|---|---|
4.2.1
minor
| ||
4.1.0
minor
| ||
4.0.2
patch
| ||
4.0.1
patch
| ||
4.0.0
initial
|