SpawnDev.AI
Run and serve local LLMs and image generation, on the desktop and in the browser.
SpawnDev.AI runs and serves local LLMs and image generation on the SpawnDev.ILGPU.ML inference engine (GGUF models, KV-cache decode, SD-Turbo). SpawnDev.AI.Server is an Ollama-compatible model server as a library that can be hosted over HTTP on the desktop or over a MessagePort in a browser shared worker. SpawnDev.AI holds the core contracts: chat types, tool-calling types, memory store and speaker registry.
Runs on
WebGPU (browser)CUDAOpenCLCPU (via SpawnDev.ILGPU.ML)
Highlights
- Ollama native endpoints (/api/chat, /api/generate, /api/tags, /api/show), OpenAI /v1/chat/completions SSE and Anthropic Messages /v1/messages SSE.
- OpenAI image generation endpoint (/v1/images/generations, SD-Turbo) and an MCP surface (/mcp).
- One transport-free protocol router (AiApiRouter): host it over HTTP on desktop (drop-in on :11434) or over a MessagePort in a browser shared worker.
- Server-side agentic tool loop with built-in tools; tool calls are parsed server-side into structured calls on every protocol surface.
- Runs GGUF models such as Qwen3 in the browser on WebGPU and models such as Llama 3.1 on the desktop.
- Live demo: a room of characters with persona, model, voice and body; chat runs in the browser on WebGPU and can drive a Reachy Mini robot (Hugging Face Space build only).
using SpawnDev.AI.Server;
var store = new OllamaModelStore(); // ~/.ollama/models (or OLLAMA_MODELS)
var registry = new ModelRegistry(new OllamaCacheModelProvider(store), accelerator); // any SpawnDev.ILGPU accelerator
var engine = new AiChatEngine(registry);
var router = new AiApiRouter(engine);
// host it: map every request to router.TryHandleAsync(method, path, bodyJson, yourTransport)Install
dotnet add package SpawnDev.AIdotnet add package SpawnDev.AI.ServerBuilding on SpawnDev.AI?
Private tickets, response targets and a direct line to the engineer who wrote it.