diff --git a/www/src/content/api-reference/csharp.md b/www/src/content/api-reference/csharp.md
new file mode 100644
index 00000000..d4445d42
--- /dev/null
+++ b/www/src/content/api-reference/csharp.md
@@ -0,0 +1,681 @@
+---
+title: "C# / .NET API Reference"
+description: Complete API reference for the Foundry Local C# SDK. Covers FoundryLocalManager, ICatalog, Model, OpenAIChatClient, OpenAIAudioClient, and configuration.
+---
+
+# C# / .NET API Reference
+
+Package: `Microsoft.AI.Foundry.Local`  
+Target Framework: .NET 8.0+
+
+```bash
+dotnet add package Microsoft.AI.Foundry.Local
+```
+
+For WinML hardware acceleration on Windows:
+
+```bash
+dotnet add package Microsoft.AI.Foundry.Local.WinML
+```
+
+Namespace for all types:
+
+```csharp
+using Microsoft.AI.Foundry.Local;
+```
+
+---
+
+## FoundryLocalManager
+
+The singleton entry point for the Foundry Local SDK. Manages initialization, provides access to the model catalog, and controls the optional web service. Implements `IDisposable`.
+
+```csharp
+public class FoundryLocalManager : IDisposable
+```
+
+### Static Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `IsInitialized` | `bool` | Whether the singleton has been created. |
+| `Instance` | `FoundryLocalManager` | The singleton instance. Only valid after `CreateAsync` completes. |
+| `Urls` | `string[]?` | Bound URLs after calling `StartWebServiceAsync`. Null if web service is not running. |
+
+### Static Methods
+
+#### `CreateAsync(configuration, logger, ct?)`
+
+Creates the singleton `FoundryLocalManager` instance. Must be called before accessing `Instance`.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `configuration` | `Configuration` | Configuration settings. `AppName` is required. |
+| `logger` | `ILogger` | Application logger. Use `NullLogger.Instance` to suppress SDK log output. |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+**Throws:** `FoundryLocalException` if initialization fails.
+
+```csharp
+var config = new Configuration { AppName = "MyApp", LogLevel = LogLevel.Information };
+await FoundryLocalManager.CreateAsync(config, logger);
+var manager = FoundryLocalManager.Instance;
+```
+
+### Instance Methods
+
+#### `GetCatalogAsync(ct?)`
+
+Gets the model catalog instance. The catalog is populated on first use.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<ICatalog>`
+
+> **Note:** If using a WinML build, this call may trigger a one-off execution provider download. Call `EnsureEpsDownloadedAsync` first to separate the two operations.
+
+#### `StartWebServiceAsync(ct?)`
+
+Starts the optional OpenAI-compatible web service. After this call, `Urls` is populated with the bound addresses. The web service exposes:
+
+- `POST /v1/chat/completions` — Chat completions
+- `GET /v1/models` — List downloaded models
+- `GET /v1/models/{model_id}` — Get model details
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+#### `StopWebServiceAsync(ct?)`
+
+Stops the web service if it is running.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+#### `EnsureEpsDownloadedAsync(ct?)`
+
+Downloads and registers execution providers. Only relevant when using the WinML package. Once downloaded, execution providers are cached and subsequent calls are fast.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+#### `Dispose()`
+
+Releases all resources held by the manager.
+
+---
+
+## Configuration
+
+Settings passed to `FoundryLocalManager.CreateAsync()`.
+
+```csharp
+public class Configuration
+```
+
+| Property | Type | Required | Default | Description |
+|----------|------|----------|---------|-------------|
+| `AppName` | `string` | **Yes** | — | Your application name. Used to derive default paths. |
+| `AppDataDir` | `string?` | No | `~/.{AppName}` | Application data directory. |
+| `ModelCacheDir` | `string?` | No | `{AppDataDir}/cache/models` | Directory for downloaded model files. |
+| `LogsDir` | `string?` | No | `{AppDataDir}/logs` | Directory for log files. |
+| `LogLevel` | `LogLevel` | No | `LogLevel.Warning` | Logging level. |
+| `Web` | `WebService?` | No | — | Web service configuration. |
+| `AdditionalSettings` | `IDictionary<string, string>?` | No | — | Additional key-value settings for the native core. |
+
+### WebService
+
+Optional configuration for the built-in web service.
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Urls` | `string?` | Semicolon-separated listener URLs (e.g. `"http://127.0.0.1:8080"`). |
+| `ExternalUrl` | `Uri?` | URL of an external Foundry Local service to connect to. |
+
+```csharp
+var config = new Configuration
+{
+    AppName = "MyApp",
+    LogLevel = LogLevel.Information,
+    Web = new WebService { Urls = "http://127.0.0.1:8080" }
+};
+```
+
+### LogLevel
+
+```csharp
+public enum LogLevel
+{
+    Verbose,
+    Debug,
+    Information,
+    Warning,
+    Error,
+    Fatal
+}
+```
+
+---
+
+## ICatalog
+
+Interface for discovering and managing models. Obtained via `manager.GetCatalogAsync()`.
+
+```csharp
+public interface ICatalog
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Name` | `string` | The catalog name. |
+
+### Methods
+
+#### `ListModelsAsync(ct?)`
+
+Lists all available models in the catalog.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<List<Model>>`
+
+```csharp
+var catalog = await manager.GetCatalogAsync();
+var models = await catalog.ListModelsAsync();
+foreach (var model in models)
+{
+    Console.WriteLine($"{model.Alias} ({model.Variants.Count} variants)");
+}
+```
+
+#### `GetModelAsync(modelAlias, ct?)`
+
+Looks up a model by its alias (e.g. `"qwen2.5-0.5b"`).
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `modelAlias` | `string` | The model alias to look up. |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<Model?>` — The matching model, or `null` if not found.
+
+#### `GetModelVariantAsync(modelId, ct?)`
+
+Looks up a specific model variant by its unique ID.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `modelId` | `string` | The unique model variant identifier. |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<ModelVariant?>` — The matching variant, or `null` if not found.
+
+#### `GetCachedModelsAsync(ct?)`
+
+Returns all model variants that have been downloaded to the local cache.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<List<ModelVariant>>`
+
+#### `GetLoadedModelsAsync(ct?)`
+
+Returns all model variants currently loaded into memory.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<List<ModelVariant>>`
+
+---
+
+## Model
+
+Represents a logical AI model with one or more variants. Operations are performed on the currently selected variant. Implements `IModel`.
+
+```csharp
+public class Model : IModel
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Alias` | `string` | The model alias (e.g. `"qwen2.5-0.5b"`). |
+| `Id` | `string` | The ID of the currently selected variant. |
+| `Variants` | `List<ModelVariant>` | All available variants for this model. |
+| `SelectedVariant` | `ModelVariant` | The currently selected variant. |
+
+### Methods
+
+#### `SelectVariant(variant)`
+
+Selects a specific variant for subsequent operations.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `variant` | `ModelVariant` | The variant to select. Must be one of `Variants`. |
+
+**Returns:** `void`
+
+**Throws:** `FoundryLocalException` if the variant is not valid for this model.
+
+```csharp
+// Select a specific variant (e.g. CPU variant)
+var cpuVariant = model.Variants.First(v => v.Info.Runtime.DeviceType == DeviceType.CPU);
+model.SelectVariant(cpuVariant);
+```
+
+#### `GetLatestVersion(variant)`
+
+Returns the latest version of the specified variant.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `variant` | `ModelVariant` | The variant to check. |
+
+**Returns:** `ModelVariant` — The latest version. Same as input if already latest.
+
+**Throws:** `FoundryLocalException` if the variant is not valid for this model.
+
+#### `DownloadAsync(downloadProgress?, ct?)`
+
+Downloads the selected variant to the local cache.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `downloadProgress` | `Action<float>?` | Optional callback receiving progress (0–100). |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+```csharp
+await model.DownloadAsync(
+    progress => Console.Write($"\rDownloading: {progress:F1}%"),
+    cancellationToken);
+```
+
+#### `LoadAsync(ct?)`
+
+Loads the selected variant into memory for inference.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+#### `UnloadAsync(ct?)`
+
+Unloads the selected variant from memory.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task`
+
+#### `IsCachedAsync(ct?)`
+
+Checks whether the selected variant is downloaded to the local cache.
+
+**Returns:** `Task<bool>`
+
+#### `IsLoadedAsync(ct?)`
+
+Checks whether the selected variant is loaded in memory.
+
+**Returns:** `Task<bool>`
+
+#### `GetPathAsync(ct?)`
+
+Gets the local file path of the selected variant.
+
+**Returns:** `Task<string>`
+
+#### `RemoveFromCacheAsync(ct?)`
+
+Deletes the selected variant from the local cache.
+
+**Returns:** `Task`
+
+#### `GetChatClientAsync(ct?)`
+
+Creates an OpenAIChatClient for performing chat completions with this model.
+
+**Returns:** `Task<OpenAIChatClient>`
+
+#### `GetAudioClientAsync(ct?)`
+
+Creates an OpenAIAudioClient for performing audio transcription with this model.
+
+**Returns:** `Task<OpenAIAudioClient>`
+
+---
+
+## ModelVariant
+
+A specific variant of a model (e.g. a particular quantization or device target). Contains detailed metadata and implements the same lifecycle interface as `Model`. Implements `IModel`.
+
+```csharp
+public class ModelVariant : IModel
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Alias` | `string` | The model alias. |
+| `Id` | `string` | Unique identifier of this variant. |
+| `Version` | `int` | Version number. |
+| `Info` | `ModelInfo` | Detailed metadata about this variant. |
+
+### Methods
+
+Same lifecycle methods as `Model`: `DownloadAsync()`, `LoadAsync()`, `UnloadAsync()`, `IsCachedAsync()`, `IsLoadedAsync()`, `GetPathAsync()`, `RemoveFromCacheAsync()`, `GetChatClientAsync()`, `GetAudioClientAsync()`.
+
+---
+
+## ModelInfo
+
+Detailed metadata for a model variant. Accessed via `ModelVariant.Info`.
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Id` | `string` | Unique model variant identifier. |
+| `Name` | `string` | Display name. |
+| `Alias` | `string` | Model alias. |
+| `Version` | `int` | Version number. |
+| `Uri` | `string` | Download URI. |
+| `FileSizeBytes` | `long` | File size in bytes. |
+| `PromptTemplate` | `PromptTemplate` | Template for structuring prompts. |
+| `Runtime` | `Runtime` | Runtime configuration (device type, execution provider). |
+| `IsCached` | `bool` | Whether this variant is downloaded. |
+| `ModelSettings` | `ModelSettings` | Model-specific parameters. |
+
+---
+
+## OpenAIChatClient
+
+Client for performing chat completions. Uses OpenAI-compatible request/response types (from the Betalgo.Ranul.OpenAI SDK).
+
+```csharp
+public class OpenAIChatClient
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Settings` | `ChatSettings` | Configuration settings for chat completions. |
+
+### ChatSettings
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `Temperature` | `float?` | — | Sampling temperature (0–2). |
+| `MaxTokens` | `int?` | — | Maximum tokens to generate. |
+| `TopP` | `float?` | — | Nucleus sampling threshold (0–1). |
+| `FrequencyPenalty` | `float?` | — | Frequency penalty (-2 to 2). |
+| `PresencePenalty` | `float?` | — | Presence penalty (-2 to 2). |
+
+### Methods
+
+#### `CompleteChatAsync(messages, ct?)`
+
+Executes a non-streaming chat completion.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `messages` | `IEnumerable<ChatMessage>` | Chat messages. The system message is automatically added. |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<ChatCompletionCreateResponse>`
+
+```csharp
+var chat = await model.GetChatClientAsync();
+chat.Settings.Temperature = 0.7f;
+chat.Settings.MaxTokens = 256;
+
+var response = await chat.CompleteChatAsync(new[]
+{
+    ChatMessage.FromSystem("You are a helpful assistant."),
+    ChatMessage.FromUser("What is 2+2?")
+});
+Console.WriteLine(response.Choices.First().Message.Content);
+```
+
+#### `CompleteChatStreamingAsync(messages, ct)`
+
+Executes a streaming chat completion. Returns an `IAsyncEnumerable` so you can process tokens as they arrive.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `messages` | `IEnumerable<ChatMessage>` | Chat messages. |
+| `ct` | `CancellationToken` | Cancellation token. |
+
+**Returns:** `IAsyncEnumerable<ChatCompletionCreateResponse>`
+
+```csharp
+await foreach (var chunk in chat.CompleteChatStreamingAsync(
+    new[] { ChatMessage.FromUser("Tell me a story") },
+    cancellationToken))
+{
+    var text = chunk.Choices?.FirstOrDefault()?.Delta?.Content;
+    if (text != null) Console.Write(text);
+}
+```
+
+---
+
+## OpenAIAudioClient
+
+Client for performing audio transcription. Uses OpenAI-compatible request/response types.
+
+```csharp
+public class OpenAIAudioClient
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `Settings` | `AudioSettings` | Configuration settings for audio transcription. |
+
+### AudioSettings
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `Language` | `string?` | — | Language of the audio (ISO 639-1 code). |
+| `Temperature` | `float?` | — | Sampling temperature. |
+
+### Methods
+
+#### `TranscribeAudioAsync(audioFilePath, ct?)`
+
+Transcribes an audio file.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `audioFilePath` | `string` | Path to the audio file. Supported format: mp3. |
+| `ct` | `CancellationToken?` | Optional cancellation token. |
+
+**Returns:** `Task<AudioCreateTranscriptionResponse>`
+
+```csharp
+var audio = await model.GetAudioClientAsync();
+var result = await audio.TranscribeAudioAsync("recording.mp3");
+Console.WriteLine(result.Text);
+```
+
+#### `TranscribeAudioStreamingAsync(audioFilePath, ct)`
+
+Transcribes an audio file with streaming output.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `audioFilePath` | `string` | Path to the audio file. Supported format: mp3. |
+| `ct` | `CancellationToken` | Cancellation token. |
+
+**Returns:** `IAsyncEnumerable<AudioCreateTranscriptionResponse>`
+
+```csharp
+await foreach (var chunk in audio.TranscribeAudioStreamingAsync("recording.mp3", ct))
+{
+    if (chunk.Text != null) Console.Write(chunk.Text);
+}
+```
+
+---
+
+## Enums and Types
+
+### DeviceType
+
+```csharp
+public enum DeviceType
+{
+    Invalid = 0,
+    CPU = 1,
+    GPU = 2,
+    NPU = 3
+}
+```
+
+### Runtime
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `DeviceType` | `DeviceType` | The target hardware device. |
+| `ExecutionProvider` | `string` | The execution provider name (e.g. `"CUDAExecutionProvider"`). |
+
+### PromptTemplate
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `System` | `string` | Template for system messages. |
+| `User` | `string` | Template for user messages. |
+| `Assistant` | `string` | Template for assistant messages. |
+| `Prompt` | `string` | Overall prompt template. |
+
+### FoundryLocalException
+
+```csharp
+public class FoundryLocalException : Exception
+```
+
+Base exception type for all errors thrown by the SDK. Includes logging context.
+
+---
+
+## Complete Example
+
+```csharp
+using Microsoft.AI.Foundry.Local;
+using Microsoft.Extensions.Logging;
+
+// 1. Initialize
+var logger = LoggerFactory.Create(b => b.AddConsole()).CreateLogger("MyApp");
+var config = new Configuration
+{
+    AppName = "MyApp",
+    LogLevel = LogLevel.Information
+};
+
+await FoundryLocalManager.CreateAsync(config, logger);
+var manager = FoundryLocalManager.Instance;
+
+// Optional: download execution providers (WinML only)
+await manager.EnsureEpsDownloadedAsync();
+
+// 2. Discover and select a model
+var catalog = await manager.GetCatalogAsync();
+var model = await catalog.GetModelAsync("qwen2.5-0.5b");
+
+// 3. Download with progress
+await model.DownloadAsync(progress =>
+    Console.Write($"\rDownloading: {progress:F1}%"));
+Console.WriteLine("\nDownload complete.");
+
+// 4. Load into memory
+await model.LoadAsync();
+
+// 5. Chat completion
+var chat = await model.GetChatClientAsync();
+chat.Settings.Temperature = 0.7f;
+chat.Settings.MaxTokens = 256;
+
+// Non-streaming
+var response = await chat.CompleteChatAsync(new[]
+{
+    ChatMessage.FromUser("What is Foundry Local?")
+});
+Console.WriteLine(response.Choices.First().Message.Content);
+
+// Streaming
+await foreach (var chunk in chat.CompleteChatStreamingAsync(
+    new[] { ChatMessage.FromUser("Tell me a joke") },
+    CancellationToken.None))
+{
+    var text = chunk.Choices?.FirstOrDefault()?.Delta?.Content;
+    if (text != null) Console.Write(text);
+}
+
+// 6. Cleanup
+await model.UnloadAsync();
+manager.Dispose();
+```
+
+## Web Service Example
+
+Start an OpenAI-compatible REST endpoint alongside the SDK:
+
+```csharp
+var config = new Configuration
+{
+    AppName = "MyApp",
+    Web = new WebService { Urls = "http://127.0.0.1:8080" }
+};
+
+await FoundryLocalManager.CreateAsync(config, logger);
+var manager = FoundryLocalManager.Instance;
+
+var catalog = await manager.GetCatalogAsync();
+var model = await catalog.GetModelAsync("qwen2.5-0.5b");
+await model.DownloadAsync();
+await model.LoadAsync();
+
+// Start the web service
+await manager.StartWebServiceAsync();
+Console.WriteLine($"Web service running at: {string.Join(", ", manager.Urls!)}");
+
+// The service exposes OpenAI-compatible endpoints:
+//   POST /v1/chat/completions
+//   GET  /v1/models
+//   GET  /v1/models/{model_id}
+
+// Stop when done
+await manager.StopWebServiceAsync();
+await model.UnloadAsync();
+manager.Dispose();
+```
diff --git a/www/src/content/api-reference/index.md b/www/src/content/api-reference/index.md
new file mode 100644
index 00000000..297c3b42
--- /dev/null
+++ b/www/src/content/api-reference/index.md
@@ -0,0 +1,104 @@
+---
+title: API Reference
+description: Complete API reference for the Foundry Local SDKs. Build AI-powered applications with JavaScript/TypeScript or C#/.NET.
+---
+
+# API Reference
+
+Foundry Local provides native SDKs for building AI-powered applications that run entirely on-device. Both SDKs share the same architecture and concepts, so code patterns transfer across languages.
+
+## Available SDKs
+
+| SDK | Package | Install |
+|-----|---------|---------|
+| **JavaScript / TypeScript** | `foundry-local-sdk` | `npm install foundry-local-sdk` |
+| **C# / .NET** | `Microsoft.AI.Foundry.Local` | `dotnet add package Microsoft.AI.Foundry.Local` |
+
+## Core Concepts
+
+Every Foundry Local application follows the same workflow regardless of language:
+
+1. **Initialize** — Create a `FoundryLocalManager` with your app configuration
+2. **Discover** — Use the `Catalog` to browse available models
+3. **Download** — Pull a model to your local cache (with progress tracking)
+4. **Load** — Load the model into memory for inference
+5. **Infer** — Create a `ChatClient` or `AudioClient` and run completions or transcriptions
+6. **Cleanup** — Unload models when done to free resources
+
+## Architecture Overview
+
+Both SDKs expose the same class hierarchy:
+
+- **FoundryLocalManager** — Singleton entry point. Initializes the native core, provides access to the catalog and optional web service.
+- **Catalog** — Discovers and lists models from the model registry. Returns `Model` objects grouped by alias.
+- **Model** — A logical model (e.g. "qwen2.5-0.5b") that may have multiple variants (quantizations, device targets). Delegates lifecycle operations to its selected variant.
+- **ModelVariant** — A specific model binary (e.g. "qwen2.5-0.5b-cpu-int4"). Handles download, load, unload, and cache operations.
+- **ChatClient** — Performs chat completions (synchronous and streaming) against a loaded model. Follows the OpenAI Chat Completions API structure.
+- **AudioClient** — Performs audio transcription (synchronous and streaming) against a loaded model. Follows the OpenAI Audio API structure.
+- **Configuration** — Specifies app name, directories, log level, and optional web service settings.
+
+## Quick Comparison
+
+### Initialization
+
+**JavaScript:**
+
+```typescript
+import { FoundryLocalManager } from 'foundry-local-sdk';
+
+const manager = FoundryLocalManager.create({
+  appName: 'MyApp',
+  logLevel: 'info'
+});
+```
+
+**C#:**
+
+```csharp
+using Microsoft.AI.Foundry.Local;
+
+var config = new Configuration { AppName = "MyApp", LogLevel = LogLevel.Information };
+await FoundryLocalManager.CreateAsync(config, logger);
+var manager = FoundryLocalManager.Instance;
+```
+
+### Chat Completion
+
+**JavaScript:**
+
+```typescript
+const model = await manager.catalog.getModel('qwen2.5-0.5b');
+await model.download();
+await model.load();
+
+const chat = model.createChatClient();
+const response = await chat.completeChat([
+  { role: 'user', content: 'Hello!' }
+]);
+```
+
+**C#:**
+
+```csharp
+var catalog = await manager.GetCatalogAsync();
+var model = await catalog.GetModelAsync("qwen2.5-0.5b");
+await model.DownloadAsync();
+await model.LoadAsync();
+
+var chat = await model.GetChatClientAsync();
+var response = await chat.CompleteChatAsync(new[] {
+    ChatMessage.FromUser("Hello!")
+});
+```
+
+## Detailed References
+
+- [JavaScript / TypeScript API Reference](/docs/api-reference/javascript) — Full class and method documentation for the JS SDK
+- [C# / .NET API Reference](/docs/api-reference/csharp) — Full class and method documentation for the C# SDK
+
+## Source Code
+
+- [JavaScript SDK on GitHub](https://github.com/microsoft/Foundry-Local/tree/main/sdk_v2/js)
+- [C# SDK on GitHub](https://github.com/microsoft/Foundry-Local/tree/main/sdk_v2/cs)
+- [JavaScript Samples](https://github.com/microsoft/Foundry-Local/tree/main/samples/js)
+- [C# Samples](https://github.com/microsoft/Foundry-Local/tree/main/samples/cs)
diff --git a/www/src/content/api-reference/javascript.md b/www/src/content/api-reference/javascript.md
new file mode 100644
index 00000000..c487fb7f
--- /dev/null
+++ b/www/src/content/api-reference/javascript.md
@@ -0,0 +1,545 @@
+---
+title: JavaScript / TypeScript API Reference
+description: Complete API reference for the Foundry Local JavaScript SDK. Covers FoundryLocalManager, Catalog, Model, ChatClient, AudioClient, and configuration.
+---
+
+# JavaScript / TypeScript API Reference
+
+Package: `foundry-local-sdk`
+
+```bash
+npm install foundry-local-sdk
+```
+
+All classes are exported from the package root:
+
+```typescript
+import {
+  FoundryLocalManager,
+  Catalog,
+  Model,
+  ModelVariant,
+  ChatClient,
+  AudioClient
+} from 'foundry-local-sdk';
+```
+
+---
+
+## FoundryLocalManager
+
+The singleton entry point for the Foundry Local SDK. Manages initialization of the native core and provides access to the Catalog and web service.
+
+### Static Methods
+
+#### `FoundryLocalManager.create(config)`
+
+Creates and returns the singleton `FoundryLocalManager` instance.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `config` | `FoundryLocalConfig` | Configuration object. `appName` is required. |
+
+**Returns:** `FoundryLocalManager`
+
+```typescript
+const manager = FoundryLocalManager.create({
+  appName: 'MyApp',
+  logLevel: 'info'
+});
+```
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `catalog` | `Catalog` | The Catalog instance for discovering and managing models. |
+| `urls` | `string[]` | URLs where the web service is listening. Empty if web service is not running. |
+
+### Methods
+
+#### `startWebService()`
+
+Starts the local OpenAI-compatible web service. After calling this, use the `urls` property to get the bound addresses. If no listener address is configured, defaults to `127.0.0.1:0` (random ephemeral port).
+
+**Returns:** `void`
+
+**Throws:** `Error` if starting the service fails.
+
+#### `stopWebService()`
+
+Stops the local web service.
+
+**Returns:** `void`
+
+**Throws:** `Error` if stopping the service fails.
+
+---
+
+## FoundryLocalConfig
+
+Configuration object passed to `FoundryLocalManager.create()`.
+
+| Property | Type | Required | Default | Description |
+|----------|------|----------|---------|-------------|
+| `appName` | `string` | **Yes** | — | Your application name. Used to derive default paths. |
+| `logLevel` | `string` | No | `'warning'` | Log level: `'trace'`, `'debug'`, `'info'`, `'warn'`, `'error'`, or `'fatal'`. |
+| `appDataDir` | `string` | No | `~/.{appName}` | Application data directory. |
+| `modelCacheDir` | `string` | No | `{appDataDir}/cache/models` | Directory for downloaded model files. |
+| `logsDir` | `string` | No | `{appDataDir}/logs` | Directory for log files. |
+| `webServiceUrls` | `string` | No | — | Semicolon-separated listener URLs for the web service (e.g. `"http://127.0.0.1:8080"`). |
+| `serviceEndpoint` | `string` | No | — | URL of an external Foundry Local service to connect to instead of using the embedded core. |
+| `libraryPath` | `string` | No | — | Path to the native library. Normally auto-detected. |
+| `additionalSettings` | `Record<string, string>` | No | — | Additional key-value settings passed to the native core. |
+
+---
+
+## Catalog
+
+Discovers and lists AI models available in the system. Accessed via `manager.catalog`.
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `name` | `string` | The name of the catalog. |
+
+### Methods
+
+#### `getModels()`
+
+Lists all available models in the catalog.
+
+**Returns:** `Promise<Model[]>`
+
+```typescript
+const models = await manager.catalog.getModels();
+for (const model of models) {
+  console.log(model.alias, model.variants.length, 'variants');
+}
+```
+
+#### `getModel(alias)`
+
+Retrieves a model by its alias (e.g. `"qwen2.5-0.5b"`).
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `alias` | `string` | The alias of the model to retrieve. |
+
+**Returns:** `Promise<Model>`
+
+**Throws:** `Error` if alias is empty or model is not found.
+
+#### `getModelVariant(modelId)`
+
+Retrieves a specific model variant by its unique ID.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `modelId` | `string` | The unique identifier of the model variant. |
+
+**Returns:** `Promise<ModelVariant>`
+
+**Throws:** `Error` if modelId is empty or variant is not found.
+
+#### `getCachedModels()`
+
+Returns all model variants that have been downloaded to the local cache.
+
+**Returns:** `Promise<ModelVariant[]>`
+
+#### `getLoadedModels()`
+
+Returns all model variants currently loaded into memory.
+
+**Returns:** `Promise<ModelVariant[]>`
+
+---
+
+## Model
+
+Represents a logical AI model (e.g. "qwen2.5-0.5b") that may have multiple variants. Operations are performed on the currently selected variant. Implements `IModel`.
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `alias` | `string` | The model alias (e.g. `"qwen2.5-0.5b"`). |
+| `id` | `string` | The ID of the currently selected variant. |
+| `isCached` | `boolean` | Whether the selected variant is downloaded to the local cache. |
+| `path` | `string` | Local file path of the selected variant. |
+| `variants` | `ModelVariant[]` | All available variants for this model. |
+
+### Methods
+
+#### `selectVariant(modelId)`
+
+Selects a specific variant by its ID.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `modelId` | `string` | The ID of the variant to select. |
+
+**Returns:** `void`
+
+**Throws:** `Error` if no variant with the given ID exists.
+
+#### `download(progressCallback?)`
+
+Downloads the selected variant to the local cache.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `progressCallback` | `(progress: number) => void` | Optional callback receiving download progress (0–100). |
+
+**Returns:** `Promise<void>`
+
+```typescript
+await model.download((progress) => {
+  console.log(`Download: ${progress.toFixed(1)}%`);
+});
+```
+
+#### `load()`
+
+Loads the selected variant into memory for inference.
+
+**Returns:** `Promise<void>`
+
+#### `unload()`
+
+Unloads the selected variant from memory.
+
+**Returns:** `Promise<void>`
+
+#### `isLoaded()`
+
+Checks whether the selected variant is currently loaded in memory.
+
+**Returns:** `Promise<boolean>`
+
+#### `removeFromCache()`
+
+Deletes the selected variant from the local cache.
+
+**Returns:** `void`
+
+#### `createChatClient()`
+
+Creates a ChatClient for performing chat completions with this model.
+
+**Returns:** `ChatClient`
+
+#### `createAudioClient()`
+
+Creates an AudioClient for performing audio transcription with this model.
+
+**Returns:** `AudioClient`
+
+#### `addVariant(variant)`
+
+Adds a variant to this model. Automatically selects it if it is cached and the current one is not.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `variant` | `ModelVariant` | The variant to add. Must share the same alias. |
+
+**Returns:** `void`
+
+**Throws:** `Error` if the variant's alias does not match.
+
+---
+
+## ModelVariant
+
+A specific variant of a model (e.g. a particular quantization or device target). Contains detailed model metadata and implements the same lifecycle interface as `Model`. Implements `IModel`.
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `alias` | `string` | The model alias. |
+| `id` | `string` | The unique identifier of this variant. |
+| `isCached` | `boolean` | Whether this variant is downloaded locally. |
+| `path` | `string` | Local file path of this variant. |
+| `modelInfo` | `ModelInfo` | Detailed metadata about this variant. |
+
+### Methods
+
+Same lifecycle methods as `Model`: `download()`, `load()`, `unload()`, `isLoaded()`, `removeFromCache()`, `createChatClient()`, `createAudioClient()`.
+
+---
+
+## ModelInfo
+
+Detailed metadata for a model variant. Returned by `ModelVariant.modelInfo`.
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `id` | `string` | Unique model variant identifier. |
+| `name` | `string` | Display name. |
+| `alias` | `string` | Model alias. |
+| `version` | `number` | Version number. |
+| `uri` | `string` | Download URI. |
+| `fileSizeBytes` | `number` | File size in bytes. |
+| `promptTemplate` | `PromptTemplate` | Template for structuring prompts. |
+| `runtime` | `Runtime` | Runtime configuration (device type, execution provider). |
+| `isCached` | `boolean` | Whether this variant is downloaded. |
+| `modelSettings` | `ModelSettings` | Model-specific parameters. |
+
+---
+
+## ChatClient
+
+Client for performing chat completions with a loaded model. Follows the OpenAI Chat Completions API structure.
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `settings` | `ChatClientSettings` | Configuration settings for chat completions. |
+
+### Methods
+
+#### `completeChat(messages)`
+
+Performs a non-streaming chat completion.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `messages` | `any[]` | Array of message objects with `role` and `content` fields. |
+
+**Returns:** `Promise<any>` — The chat completion response.
+
+**Throws:** `Error` if messages are invalid or completion fails.
+
+```typescript
+const response = await chat.completeChat([
+  { role: 'system', content: 'You are a helpful assistant.' },
+  { role: 'user', content: 'What is 2+2?' }
+]);
+console.log(response.choices[0].message.content);
+```
+
+#### `completeStreamingChat(messages, callback)`
+
+Performs a streaming chat completion. The callback receives each response chunk as it arrives.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `messages` | `any[]` | Array of message objects. |
+| `callback` | `(chunk: any) => void` | Callback invoked for each streaming chunk. |
+
+**Returns:** `Promise<void>` — Resolves when the stream completes.
+
+**Throws:** `Error` if messages or callback are invalid, or streaming fails.
+
+```typescript
+await chat.completeStreamingChat(
+  [{ role: 'user', content: 'Tell me a story' }],
+  (chunk) => {
+    const delta = chunk.choices?.[0]?.delta?.content;
+    if (delta) process.stdout.write(delta);
+  }
+);
+```
+
+---
+
+## ChatClientSettings
+
+Configuration for chat completion requests. Assign to `chatClient.settings` before calling completion methods.
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `temperature` | `number?` | — | Sampling temperature (0–2). Higher values produce more random output. |
+| `maxTokens` | `number?` | — | Maximum number of tokens to generate. |
+| `topP` | `number?` | — | Nucleus sampling threshold (0–1). |
+| `topK` | `number?` | — | Top-K sampling parameter. |
+| `frequencyPenalty` | `number?` | — | Penalizes tokens based on frequency in the output so far (-2 to 2). |
+| `presencePenalty` | `number?` | — | Penalizes tokens based on presence in the output so far (-2 to 2). |
+| `n` | `number?` | — | Number of completions to generate. |
+| `randomSeed` | `number?` | — | Random seed for reproducible outputs. |
+
+```typescript
+const chat = model.createChatClient();
+chat.settings.temperature = 0.7;
+chat.settings.maxTokens = 512;
+```
+
+---
+
+## AudioClient
+
+Client for performing audio transcription with a loaded model. Follows the OpenAI Audio API structure.
+
+### Properties
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `settings` | `AudioClientSettings` | Configuration settings for audio operations. |
+
+### Methods
+
+#### `transcribe(audioFilePath)`
+
+Transcribes an audio file.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `audioFilePath` | `string` | Path to the audio file (supported format: mp3). |
+
+**Returns:** `Promise<any>` — The transcription result.
+
+**Throws:** `Error` if the file path is invalid or transcription fails.
+
+```typescript
+const result = await audio.transcribe('./recording.mp3');
+console.log(result.text);
+```
+
+#### `transcribeStreaming(audioFilePath, callback)`
+
+Transcribes an audio file with streaming output.
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `audioFilePath` | `string` | Path to the audio file. |
+| `callback` | `(chunk: any) => void` | Callback invoked for each streaming chunk. |
+
+**Returns:** `Promise<void>` — Resolves when the stream completes.
+
+**Throws:** `Error` if the file path or callback is invalid, or streaming fails.
+
+```typescript
+await audio.transcribeStreaming('./recording.mp3', (chunk) => {
+  if (chunk.text) process.stdout.write(chunk.text);
+});
+```
+
+---
+
+## AudioClientSettings
+
+Configuration for audio transcription requests. Assign to `audioClient.settings` before calling transcription methods.
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `language` | `string?` | — | Language of the audio (ISO 639-1 code, e.g. `"en"`). |
+| `temperature` | `number?` | — | Sampling temperature for transcription. |
+
+---
+
+## Enums and Types
+
+### DeviceType
+
+```typescript
+enum DeviceType {
+  Invalid = 0,
+  CPU = 1,
+  GPU = 2,
+  NPU = 3
+}
+```
+
+### Runtime
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `deviceType` | `DeviceType` | The target hardware device. |
+| `executionProvider` | `string` | The execution provider name (e.g. `"CUDAExecutionProvider"`). |
+
+### PromptTemplate
+
+| Property | Type | Description |
+|----------|------|-------------|
+| `system` | `string` | Template for system messages. |
+| `user` | `string` | Template for user messages. |
+| `assistant` | `string` | Template for assistant messages. |
+| `prompt` | `string` | Overall prompt template. |
+
+---
+
+## Complete Example
+
+```typescript
+import { FoundryLocalManager } from 'foundry-local-sdk';
+
+async function main() {
+  // 1. Initialize
+  const manager = FoundryLocalManager.create({
+    appName: 'MyApp',
+    logLevel: 'info'
+  });
+
+  // 2. Discover and select a model
+  const model = await manager.catalog.getModel('qwen2.5-0.5b');
+  console.log(`Model: ${model.alias}, Variants: ${model.variants.length}`);
+
+  // 3. Download with progress
+  await model.download((progress) => {
+    process.stdout.write(`\rDownloading: ${progress.toFixed(1)}%`);
+  });
+  console.log('\nDownload complete.');
+
+  // 4. Load into memory
+  await model.load();
+
+  // 5. Chat completion
+  const chat = model.createChatClient();
+  chat.settings.temperature = 0.7;
+  chat.settings.maxTokens = 256;
+
+  // Non-streaming
+  const response = await chat.completeChat([
+    { role: 'user', content: 'What is Foundry Local?' }
+  ]);
+  console.log(response.choices[0].message.content);
+
+  // Streaming
+  await chat.completeStreamingChat(
+    [{ role: 'user', content: 'Tell me a joke' }],
+    (chunk) => {
+      const text = chunk.choices?.[0]?.delta?.content;
+      if (text) process.stdout.write(text);
+    }
+  );
+
+  // 6. Cleanup
+  await model.unload();
+}
+
+main().catch(console.error);
+```
+
+## Web Service Example
+
+Start an OpenAI-compatible REST endpoint alongside the SDK:
+
+```typescript
+import { FoundryLocalManager } from 'foundry-local-sdk';
+
+const manager = FoundryLocalManager.create({
+  appName: 'MyApp',
+  webServiceUrls: 'http://127.0.0.1:8080'
+});
+
+const model = await manager.catalog.getModel('qwen2.5-0.5b');
+await model.download();
+await model.load();
+
+// Start the web service
+manager.startWebService();
+console.log('Web service running at:', manager.urls);
+
+// The service exposes OpenAI-compatible endpoints:
+//   POST /v1/chat/completions
+//   GET  /v1/models
+//   GET  /v1/models/{model_id}
+
+// Stop when done
+manager.stopWebService();
+await model.unload();
+```
diff --git a/www/src/lib/components/document/doc-renderer.svelte b/www/src/lib/components/document/doc-renderer.svelte
index fba20a64..4c0c5ab0 100644
--- a/www/src/lib/components/document/doc-renderer.svelte
+++ b/www/src/lib/components/document/doc-renderer.svelte
@@ -27,7 +27,9 @@
 				'css',
 				'svelte',
 				'shell',
-				'tsx'
+				'tsx',
+				'csharp',
+				'python'
 			]
 		});
 	});
diff --git a/www/src/lib/config.ts b/www/src/lib/config.ts
index 4489d3a0..312f742a 100644
--- a/www/src/lib/config.ts
+++ b/www/src/lib/config.ts
@@ -44,6 +44,11 @@ export let navItems: NavItem[] = [
 		title: 'Docs',
 		href: 'https://learn.microsoft.com/en-us/azure/ai-foundry/foundry-local/get-started',
 		icon: BookOpen
+	},
+	{
+		title: 'API Reference',
+		href: '/docs/api-reference',
+		icon: Code
 	}
 ];