Cisharpai

Cisharpai is a unified .NET client library for interacting with multiple LLM providers. It exposes shared interfaces so you can switch providers (OpenAI, Azure OpenAI, Azure AI Inference, Anthropic, Cohere) with minimal code changes.

Why Cisharpai?

• One shared IChatCompletionClient and IEmbeddingClient interface
• Unified request/response models across all providers
• Provider-specific packages: Cisharpai.OpenAi, Cisharpai.Azure, Cisharpai.Anthropic, Cisharpai.Cohere
• Built-in HTTP resilience for retries and timeouts
• Feature Collection pattern for optional capabilities: JSON output, tool calling, grounded chat (RAG), image embeddings, multimodal embeddings
• No exceptions for API errors -- consistent IsSuccess/ErrorMessage error handling
• Full debug support with RawRequestJson/RawResponseJson

Quick start

1. Add references to the core library and a provider package:

• Cisharpai
• Cisharpai.OpenAi or Cisharpai.Azure or Cisharpai.Anthropic or Cisharpai.Cohere

2. Register and call the client:

using Cisharpai;
using Cisharpai.Models;
using Cisharpai.OpenAi;
using Microsoft.Extensions.DependencyInjection;

var services = new ServiceCollection();
services.AddOpenAiClient(options =>
{
    options.ApiKey = "YOUR_API_KEY";
});

var provider = services.BuildServiceProvider();
var client = provider.GetRequiredService<IChatCompletionClient>();

var request = new ChatCompletionRequest(
    Messages: [new LlmMessage(LlmRole.User, "Say hello in one sentence.")],
    Model: "gpt-4.1-nano",
    Temperature: 0.2,
    MaxTokens: 100);

var response = await client.GetChatCompletionAsync(request);
Console.WriteLine(response.Content);

Documentation

Start here:

• Getting Started
• Provider Feature Matrix -- see what each provider supports
• OpenAI Quickstart
• Embeddings -- text, image, and multimodal embeddings across providers
• JSON Output -- JSON Mode and Structured Outputs
• Tool Calling -- function calling across providers
• Grounded Chat (RAG) -- document grounding with citations
• Feature Extensions -- Feature Collection pattern

Samples

• Interactive console demo: src/Cisharp.Console/ -- covers all providers and features

Building Locally

The project includes a PowerShell build script that handles versioning, building, testing, and NuGet packaging.

Prerequisites

• .NET 8 SDK and .NET 10 SDK
• PowerShell 7+ (pwsh)

Running the build

pwsh scripts/build.ps1

This will:

1. Restore dotnet tools (including GitVersion)
2. Calculate the version using GitVersion (ContinuousDeployment mode)
3. Restore NuGet packages
4. Build the solution in Release configuration
5. Run unit tests on both net8.0 and net10.0
6. Pack NuGet packages into artifacts/NuGet/

Build options

# Skip tests
pwsh scripts/build.ps1 -skiptest

# Build and publish to nuget.org
pwsh scripts/build.ps1 -nugetApiKey "YOUR_KEY" -nugetPublish $true

Build artifacts

Output	Location
NuGet packages (.nupkg + .snupkg)	artifacts/NuGet/
Test results (.trx)	artifacts/TestResults/

Running Integration Tests

Integration tests require environment variables to be set. Create a .env file in the repository root or set them in your environment.

Environment Variables Format

Variable	Format	Example
OPENAI_TEST_API_KEY	OpenAI API key	sk-proj-...
ANTHROPIC_TEST_API_KEY	Anthropic API key	sk-ant-...
AZURE_OPENAI_TEST_ENDPOINT	Azure OpenAI endpoint URL (no trailing slash)	https://myresource.openai.azure.com
AZURE_OPENAI_TEST_API_KEY	Azure OpenAI API key	abc123...
AZURE_OPENAI_TEST_DEPLOYMENTS	Comma-separated deployment names	gpt-4o,gpt-4o-mini
AZURE_OPENAI_TEST_EMBEDDING_DEPLOYMENT	Single embedding deployment name	text-embedding-ada-002
AZURE_INFERENCE_TEST_ENDPOINT	Azure AI Inference endpoint URL	https://mymodel.eastus.models.ai.azure.com
AZURE_INFERENCE_TEST_API_KEY	Azure AI Inference API key	abc123...
AZURE_INFERENCE_TEST_MODELS	Comma-separated model IDs for Azure AI Inference	Phi-3-mini-4k-instruct
AZURE_INFERENCE_TEST_EMBEDDING_ENDPOINT	Azure AI Inference embedding endpoint URL	https://myembedding.eastus.models.ai.azure.com
AZURE_INFERENCE_TEST_EMBEDDING_KEY	Azure AI Inference embedding API key	abc123...
AZURE_INFERENCE_TEST_EMBEDDING_MODEL	Model ID for Azure AI Inference embedding	Cohere-embed-v3-english
COHERE_TEST_API_KEY	Cohere API key	...

Example .env file

# OpenAI
OPENAI_TEST_API_KEY=sk-proj-your-key-here

# Azure OpenAI
AZURE_OPENAI_TEST_ENDPOINT=https://myresource.openai.azure.com
AZURE_OPENAI_TEST_API_KEY=your-azure-openai-key
AZURE_OPENAI_TEST_DEPLOYMENTS=gpt-4o,gpt-4o-mini
AZURE_OPENAI_TEST_EMBEDDING_DEPLOYMENT=text-embedding-ada-002

# Azure AI Inference
AZURE_INFERENCE_TEST_ENDPOINT=https://mymodel.eastus.models.ai.azure.com
AZURE_INFERENCE_TEST_API_KEY=your-azure-inference-key
AZURE_INFERENCE_TEST_MODELS=Phi-3-mini-4k-instruct
AZURE_INFERENCE_TEST_EMBEDDING_ENDPOINT=https://myembedding.eastus.models.ai.azure.com
AZURE_INFERENCE_TEST_EMBEDDING_KEY=your-azure-inference-embedding-key
AZURE_INFERENCE_TEST_EMBEDDING_MODEL=Cohere-embed-v3-english

# Anthropic
ANTHROPIC_TEST_API_KEY=sk-ant-your-key-here

# Cohere
COHERE_TEST_API_KEY=your-cohere-key

Note: The .env file parser supports both quoted and unquoted values, and lines starting with # are treated as comments.

License

See the repository license file for terms.