MeshWeaver.AI.OpenAI
3.0.0-rc3
dotnet add package MeshWeaver.AI.OpenAI --version 3.0.0-rc3
NuGet\Install-Package MeshWeaver.AI.OpenAI -Version 3.0.0-rc3
<PackageReference Include="MeshWeaver.AI.OpenAI" Version="3.0.0-rc3" />
<PackageVersion Include="MeshWeaver.AI.OpenAI" Version="3.0.0-rc3" />
<PackageReference Include="MeshWeaver.AI.OpenAI" />
paket add MeshWeaver.AI.OpenAI --version 3.0.0-rc3
#r "nuget: MeshWeaver.AI.OpenAI, 3.0.0-rc3"
#:package MeshWeaver.AI.OpenAI@3.0.0-rc3
#addin nuget:?package=MeshWeaver.AI.OpenAI&version=3.0.0-rc3&prerelease
#tool nuget:?package=MeshWeaver.AI.OpenAI&version=3.0.0-rc3&prerelease
MeshWeaver.AI.OpenAI
OpenAI-wire-protocol model providers for the MeshWeaver AI framework. One assembly serves every provider that speaks the OpenAI chat-completions protocol — they differ only by base URL and credentials:
Provider (AddXxx) |
ProviderName |
Endpoint |
|---|---|---|
AddOpenAI |
OpenAI |
https://api.openai.com (SDK default) |
AddAzureOpenAI |
AzureOpenAI |
your *.openai.azure.com deployment |
AddOpenAICompatible |
OpenAICompatible |
any base URL you supply (OpenRouter, Groq, Together, vLLM, …) |
Each AddXxx self-registers a LanguageModelCatalogSource (so the provider appears
in the chat model picker and the Settings → Language Models tab) plus its
IChatClientFactory. There is no central registry — a host opts into each provider it
needs, gated by the Features:Ai:Providers:* flags.
Registration
meshBuilder
.AddOpenAI() // direct api.openai.com
.AddAzureOpenAI() // Azure-hosted OpenAI deployment
.AddOpenAICompatible(); // generic custom-URL provider (OpenRouter, …)
Credentials
Credentials are not read from appsettings per user. Each saved provider is a
nodeType:ModelProvider mesh node carrying its endpoint + (encrypted) API key; each
LanguageModel child points back at it via ModelDefinition.ProviderRef. At chat time
ChatClientCredentialResolver follows that reference, so the factory builds an
OpenAIClient aimed at the right endpoint with the right key — and several distinct
OpenAI-compatible gateways coexist without collision.
A system-default OpenAI: / AzureOpenAI: config section is still honoured as a
fallback for the built-in catalog. OpenAICompatible has no system default — the user
supplies the URL + key and fetches the model list live in the Settings tab.
Key components
OpenAIChatClientAgentFactory— servesOpenAIandOpenAICompatible(plainOpenAIClientat the resolved endpoint).AzureOpenAIChatClientAgentFactory— servesAzureOpenAI(AzureOpenAIClient).OpenAIExtensions/AzureOpenAIExtensions— theAddXxxbuilder extensions.OpenAIConfiguration/AzureOpenAIConfiguration— system-default config shapes.
Related
- MeshWeaver.AI — core AI abstractions, the provider/credential resolver, the catalog node types.
- MeshWeaver.AI.AzureFoundry — Azure AI Foundry + direct Anthropic.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- Azure.AI.OpenAI (>= 2.9.0-beta.1)
- MeshWeaver.AI (>= 3.0.0-rc3)
- MeshWeaver.Messaging.Hub (>= 3.0.0-rc3)
- Microsoft.Agents.AI.OpenAI (>= 1.17.0)
- Microsoft.Extensions.AI.OpenAI (>= 10.8.3)
- Microsoft.Extensions.DependencyInjection.Abstractions (>= 10.0.10)
- Microsoft.Extensions.Options (>= 10.0.10)
- OpenAI (>= 2.12.0)
NuGet packages
This package is not used by any NuGet packages.
GitHub repositories
This package is not used by any popular GitHub repositories.