feat: Add .NET Server AI SDK examples

.github/workflows/server-sdk-ai.yml

@@ -0,0 +1,43 @@
+name: server-sdk-ai
+on:
+  schedule:
+    # * is a special character in YAML so you have to quote this string
+    - cron: '0 9 * * *'
+  push:
+    branches: [ main, 'feat/**' ]
+    paths:
+      - 'server-sdk-ai/**'
+      - '.github/workflows/server-sdk-ai.yml'
+  pull_request:
+    branches: [ main, 'feat/**' ]
+    paths:
+      - 'server-sdk-ai/**'
+      - '.github/workflows/server-sdk-ai.yml'
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest]
+      fail-fast: false
+    runs-on: ${{ matrix.os }}
+
+    steps:
+      - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3
+
+      - name: Setup dotnet build tools
+        uses: actions/setup-dotnet@9a946fdbd5fb07b82b2f5a4466058b876ab72bb2 # v5.3.0
+        with:
+          dotnet-version: 8.0
+
+      # Build-only: these examples call external model providers and need
+      # provider credentials to run, so CI just compiles them to catch build errors.
+      - name: Build server-sdk-ai examples
+        shell: bash
+        run: |
+          set -e
+          for proj in $(find server-sdk-ai -name '*.csproj' | sort); do
+            echo "::group::Building $proj"
+            dotnet build "$proj"
+            echo "::endgroup::"
+          done

CODEOWNERS

@@ -1,2 +1,5 @@
 # Repository Maintainers
 * @launchdarkly/team-sdk-net
+
+# AI examples
+/server-sdk-ai/ @launchdarkly/team-ai-agents @launchdarkly/team-ai-evals

README.md

@@ -7,9 +7,10 @@ or the [.NET SDK reference guide](https://docs.launchdarkly.com/sdk/server-side/
 
 ## SDKs
 
-| SDK             | Package                  | Examples                       |
-|-----------------|--------------------------|--------------------------------|
-| .NET Server SDK | `LaunchDarkly.ServerSdk` | [`server-sdk/`](./server-sdk/) |
+| SDK                | Package                     | Examples                                 |
+|--------------------|-----------------------------|------------------------------------------|
+| .NET Server SDK    | `LaunchDarkly.ServerSdk`    | [`server-sdk/`](./server-sdk/)           |
+| .NET Server AI SDK | `LaunchDarkly.ServerSdk.Ai` | [`server-sdk-ai/`](./server-sdk-ai/)     |
 
 ## Requirements
 

server-sdk-ai/README.md

@@ -0,0 +1,26 @@
+# LaunchDarkly .NET Server AI SDK examples
+
+Examples for the [`LaunchDarkly.ServerSdk.Ai`](https://www.nuget.org/packages/LaunchDarkly.ServerSdk.Ai) package.
+
+For more comprehensive instructions, you can visit the [AI Configs Quickstart](https://docs.launchdarkly.com/home/ai-configs/quickstart) or the [.NET AI SDK reference guide](https://docs.launchdarkly.com/sdk/ai/dotnet).
+
+Each example is a self-contained .NET console application you can run independently.
+
+## Getting Started
+
+These examples show how to integrate LaunchDarkly AI with different providers.
+
+| Provider | Example                                                  | Description                                                       |
+|----------|----------------------------------------------------------|------------------------------------------------------------------|
+| OpenAI   | [Chat Completions](./frameworks/openai/chat-completions/) | `CompletionConfig` with OpenAI, automatic metrics tracking       |
+| Bedrock  | [Converse](./frameworks/bedrock/converse/)               | `CompletionConfig` with the AWS Bedrock Converse API, metrics tracking |
+
+## Features
+
+These examples focus on the LaunchDarkly AI SDK itself. They are **provider-agnostic** — they retrieve and resolve AI Configs, then exercise the tracker API with synthetic operations rather than calling any model provider.
+
+| Example                                              | Description                                                                              |
+|------------------------------------------------------|-----------------------------------------------------------------------------------------|
+| [Completion Config](./features/completion-config/)   | Default values, Mustache variables, model parameter extraction, tool enumeration, and `TrackMetricsOf` |
+| [Agent Config](./features/agent-config/)             | Agent instructions and tools; `TrackMetricsOf` and `TrackToolCall`                       |
+| [Judge Config](./features/judge-config/)             | Judge config retrieval and `TrackJudgeResult`                                            |

server-sdk-ai/features/agent-config/.env.example

@@ -0,0 +1,5 @@
+# Your LaunchDarkly server-side SDK key
+LAUNCHDARKLY_SDK_KEY=
+
+# Override to use a different AI Config key
+LAUNCHDARKLY_AGENT_KEY=sample-agent

server-sdk-ai/features/agent-config/AgentConfig.csproj

@@ -0,0 +1,16 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net8.0</TargetFramework>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="LaunchDarkly.ServerSdk" Version="8.*" />
+    <PackageReference Include="LaunchDarkly.ServerSdk.Ai" Version="0.*" />
+    <PackageReference Include="DotNetEnv" Version="3.*" />
+  </ItemGroup>
+
+</Project>

server-sdk-ai/features/agent-config/NOTICE

@@ -0,0 +1,4 @@
+LaunchDarkly .NET Server AI SDK Agent Config example
+Copyright 2026 Catamorphic, Co.
+
+This product includes software developed at LaunchDarkly (https://launchdarkly.com/).

server-sdk-ai/features/agent-config/Program.cs

@@ -0,0 +1,152 @@
+using DotNetEnv;
+using LaunchDarkly.Sdk;
+using LaunchDarkly.Sdk.Server;
+using LaunchDarkly.Sdk.Server.Ai;
+using LaunchDarkly.Sdk.Server.Ai.Adapters;
+using LaunchDarkly.Sdk.Server.Ai.Config;
+using LaunchDarkly.Sdk.Server.Ai.Tracking;
+
+Env.TraversePath().Load();
+
+var sdkKey = Environment.GetEnvironmentVariable("LAUNCHDARKLY_SDK_KEY");
+if (string.IsNullOrEmpty(sdkKey))
+{
+    Console.Error.WriteLine(
+        "LaunchDarkly SDK key is required: set the LAUNCHDARKLY_SDK_KEY environment variable and try again.");
+    return;
+}
+
+// Set agentKey to the AI config key you want to evaluate.
+var agentKey = Environment.GetEnvironmentVariable("LAUNCHDARKLY_AGENT_KEY")
+    ?? "sample-agent";
+
+var ldClient = new LdClient(Configuration.Builder(sdkKey).Build());
+if (!ldClient.Initialized)
+{
+    Console.Error.WriteLine(
+        "*** SDK failed to initialize. Please check your internet connection and SDK credential for any typo.");
+    ldClient.Dispose();
+    return;
+}
+Console.WriteLine("*** SDK successfully initialized!");
+
+var aiClient = new LdAiClient(new LdClientAdapter(ldClient));
+
+// Set up the evaluation context. This context should appear on your
+// LaunchDarkly contexts dashboard soon after you run the demo.
+var context = Context.Builder(ContextKind.Of("user"), "example-user-key")
+    .Name("Sandy")
+    .Build();
+
+// Default agent config used when the AI Config is unreachable or missing.
+var defaultValue = LdAiAgentConfigDefault.New()
+    .Enable()
+    .SetModelName("gpt-4")
+    .SetModelProviderName("openai")
+    .SetInstructions("You are a helpful research assistant for {{companyName}}.")
+    .Build();
+
+try
+{
+    var agentConfig = aiClient.AgentConfig(
+        agentKey,
+        context,
+        defaultValue,
+        variables: new Dictionary<string, object>
+        {
+            ["companyName"] = "LaunchDarkly"
+        });
+
+    if (!agentConfig.Enabled)
+    {
+        Console.WriteLine($"Agent config '{agentKey}' is disabled.");
+        return;
+    }
+
+    Console.WriteLine();
+    Console.WriteLine($"Resolved agent model: {agentConfig.Model.Name} (provider: {agentConfig.Provider.Name})");
+    Console.WriteLine();
+    Console.WriteLine("Agent instructions (Mustache variables interpolated):");
+    Console.WriteLine(agentConfig.Instructions);
+
+    if (agentConfig.Tools.Count > 0)
+    {
+        Console.WriteLine();
+        Console.WriteLine("Configured tools (map to your provider's function-calling API):");
+        foreach (var (name, tool) in agentConfig.Tools)
+        {
+            Console.WriteLine($"  {name}: {tool.Description} (type: {tool.Type})");
+        }
+    }
+
+    var tracker = agentConfig.CreateTracker();
+
+    // To keep this example provider-agnostic, the operation below is a
+    // synthetic stand-in for an actual model call — see the getting-started/*
+    // examples for end-to-end integrations with OpenAI or AWS Bedrock. The
+    // tracker API is exercised exactly the same way regardless of provider.
+    Console.WriteLine();
+    Console.WriteLine("Running a synthetic agent invocation wrapped in TrackMetricsOf...");
+
+    var result = await tracker.TrackMetricsOf(
+        SyntheticMetrics,
+        async () =>
+        {
+            await Task.Delay(150);
+            return new SyntheticResult(
+                Content: "[simulated agent response]",
+                InputTokens: 80,
+                OutputTokens: 120,
+                ToolsInvoked: new[] { "search", "calculator" });
+        });
+
+    // Record each tool invocation on the tracker so it appears in the
+    // metric summary. In a real integration these names would come from
+    // the provider's tool-use response.
+    foreach (var toolName in result.ToolsInvoked)
+    {
+        tracker.TrackToolCall(toolName);
+    }
+
+    Console.WriteLine();
+    Console.WriteLine("Simulated agent response:");
+    Console.WriteLine(result.Content);
+
+    PrintSummary(tracker.Summary);
+}
+catch (Exception ex)
+{
+    Console.Error.WriteLine($"Error: {ex.Message}");
+}
+finally
+{
+    ldClient.FlushAndWait(TimeSpan.FromSeconds(5));
+    ldClient.Dispose();
+}
+
+static AiMetrics SyntheticMetrics(SyntheticResult result) => new(
+    success: true,
+    tokens: new Usage(
+        Total: result.InputTokens + result.OutputTokens,
+        Input: result.InputTokens,
+        Output: result.OutputTokens));
+
+static void PrintSummary(MetricSummary summary)
+{
+    Console.WriteLine();
+    Console.WriteLine("Done! The tracker captured the following metrics:");
+    Console.WriteLine($"  Duration:      {summary.DurationMs}ms");
+    Console.WriteLine($"  Success:       {summary.Success}");
+    if (summary.Tokens is { } tokens)
+    {
+        Console.WriteLine($"  Input tokens:  {tokens.Input}");
+        Console.WriteLine($"  Output tokens: {tokens.Output}");
+        Console.WriteLine($"  Total tokens:  {tokens.Total}");
+    }
+}
+
+internal sealed record SyntheticResult(
+    string Content,
+    int InputTokens,
+    int OutputTokens,
+    IReadOnlyList<string> ToolsInvoked);

server-sdk-ai/features/agent-config/README.md

@@ -0,0 +1,34 @@
+# Agent Config
+
+This example focuses on the LaunchDarkly AI SDK itself — it is **provider-agnostic** and does not call OpenAI, Bedrock, Anthropic, or any other model. The actual agent invocation is replaced with a synthetic operation so you can see the SDK's tracking surface clearly. For end-to-end provider integrations see the [getting-started examples](../../frameworks).
+
+It demonstrates:
+
+- Retrieving agent instructions and tools from LaunchDarkly via `LdAiClient.AgentConfig`
+- Using the builder pattern (`LdAiAgentConfigDefault.New()`) to declare a fallback default
+- Interpolating Mustache placeholders in agent instructions
+- Enumerating the tools attached to the agent (so they can be mapped to your provider's function-calling API)
+- Tracking metrics via `tracker.TrackMetricsOf` and recording individual tool invocations via `tracker.TrackToolCall`
+
+## Prerequisites
+
+- .NET 8.0 SDK
+- A LaunchDarkly account and a server-side SDK key
+
+## Setup
+
+1. [Create an AI Config](https://launchdarkly.com/docs/home/ai-configs/create) of type **Agent** in LaunchDarkly with the key `sample-agent`. Set the instructions (you may include Mustache placeholders such as `{{companyName}}`), choose a provider and model.
+2. Copy `.env.example` to `.env` and fill in your SDK key:
+   ```sh
+   cp .env.example .env
+   ```
+3. Restore dependencies:
+   ```sh
+   dotnet restore
+   ```
+
+## Run
+
+```sh
+dotnet run
+```

server-sdk-ai/features/completion-config/.env.example

@@ -0,0 +1,5 @@
+# Your LaunchDarkly server-side SDK key
+LAUNCHDARKLY_SDK_KEY=
+
+# Override to use a different AI Config key
+LAUNCHDARKLY_COMPLETION_KEY=sample-completion

server-sdk-ai/features/completion-config/CompletionConfig.csproj

@@ -0,0 +1,16 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net8.0</TargetFramework>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="LaunchDarkly.ServerSdk" Version="8.*" />
+    <PackageReference Include="LaunchDarkly.ServerSdk.Ai" Version="0.*" />
+    <PackageReference Include="DotNetEnv" Version="3.*" />
+  </ItemGroup>
+
+</Project>

server-sdk-ai/features/completion-config/NOTICE

@@ -0,0 +1,4 @@
+LaunchDarkly .NET Server AI SDK Completion Config example
+Copyright 2026 Catamorphic, Co.
+
+This product includes software developed at LaunchDarkly (https://launchdarkly.com/).