Ananke.AspNetCore 0.7.2

Ananke.AspNetCore

ASP.NET Core integration helpers for Ananke — SSE streaming, ChatSessionEvent-to-SSE bridging, state-machine endpoint utilities, provider configuration, and session management.

Install

dotnet add package Ananke.AspNetCore

Quick start

SSE streaming

using Ananke.AspNetCore.Sse;

app.MapPost("/api/chat", async (ChatRequest request, HttpContext context, CancellationToken ct) =>
{
    // One call sets Content-Type + Cache-Control headers
    context.Response.EnableSse();

    // Write named SSE events with JSON data
    await context.Response.WriteSseAsync("session", new { sessionId = "abc" });
    await context.Response.WriteSseAsync("delta", new { text = "Hello!" });
    await context.Response.WriteSseAsync("done", new { text = "Hello!" });
});

ChatSessionEvent → SSE bridge

Pipe a StreamingChatWorkflow event stream directly to an HTTP response:

using Ananke.AspNetCore.Sse;

var events = StreamingChatWorkflow.Create("chat", model)
    .WithSystemPrompt("You are helpful.")
    .BuildStream(messages, ct);

await events.WriteSseAsync(context.Response);

Or via a delegate (useful for session-based scenarios where the writer is re-bound per request):

await events.WriteSseAsync(
    writeSse: myWriteDelegate,
    onError: msg => logger.LogError("Error: {Msg}", msg));

Event mapping:

State-machine SSE loop

Await a state machine's CurrentWork through phase transitions, surviving interrupt-induced cancellation races:

using Ananke.AspNetCore.Sse;

await machine.FireAsync(MyAction.Start);
var reachedDone = await machine.RunSseLoopAsync(MyPhase.Done);

if (reachedDone)
    await context.Response.WriteSseAsync("done", new { });

Provider configuration

Register LLM providers at startup and read settings from IConfiguration:

using Ananke.AspNetCore.Configuration;

// Register providers (typically in Program.cs)
AgentModelFactory.RegisterProvider("OpenAI",
    defaultModel: "gpt-4.1-mini",
    agentFactory: (key, model) => OpenAIChatAgentModel.Create(key, model),
    embeddingFactory: (key, model) => OpenAIEmbeddingModel.Create(key, model));

AgentModelFactory.RegisterProvider("Google",
    defaultModel: "gemini-2.5-flash",
    agentFactory: (key, model) => GeminiAgentModel.Create(key, model));

// Read from configuration (secrets.json / appsettings.json)
// { "Provider": "OpenAI", "OpenAI": { "ApiKey": "sk-...", "Model": "gpt-4.1-mini" } }
var profile = AgentModelFactory.FromConfiguration(builder.Configuration);
var agentModel = profile.CreateAgentModel();
var embedder = profile.CreateEmbeddingModel(); // null if not configured

Session store

Thread-safe in-memory session management:

using Ananke.AspNetCore.Sessions;

var sessions = new InMemorySessionStore<MySession>();

// Atomic get-or-create
var session = sessions.GetOrCreate(sessionId, () => new MySession());

// Lookup
var existing = sessions.Get(sessionId);          // null if missing
sessions.TryGet(sessionId, out var s);           // bool pattern

// Cleanup
sessions.Remove(sessionId);
var removed = sessions.RemoveAndGet(sessionId);  // remove + return

Features

• EnableSse() — one-liner SSE response header setup
• WriteSseAsync(name, data) — JSON-serialized SSE event writing with flush
• ChatSessionEvent → SSE bridge — maps all workflow event types to named SSE events
• RunSseLoopAsync(terminalState) — reusable state-machine await loop with interrupt-race handling
• AgentModelFactory — registry-based provider configuration from IConfiguration
• ProviderProfile — immutable settings record with CreateAgentModel() / CreateEmbeddingModel()
• InMemorySessionStore<T> — ConcurrentDictionary-backed thread-safe session store

Related packages

Documentation

Full docs, demos, and architecture: github.com/sevensamurai/Ananke

License

Apache 2.0