SDK Overview
The Atelico AI Engine exposes its capabilities through platform-specific SDKs that follow each engine's native conventions. All SDKs share the same underlying Rust core and provide the same capabilities.
Architecture
atelico-sdk (Rust Core)
/ | \
gdext (Godot) C ABI (FFI) PyO3 (Python)
/ \
Unity C# Unreal C++
- Godot and Python bind directly to Rust (zero-overhead)
- Unity, Unreal, and raw C/C++ go through the C FFI layer
Capabilities
All SDKs expose the same capabilities:
| Capability | Description |
|---|---|
| LLM Chat | Streaming and non-streaming text generation |
| Image Generation | PixArt and Sana image generation |
| Embeddings | Text embeddings with BERT models |
| Classifiers | Embedding-based text classification |
| LoRA | Runtime LoRA adapter loading/unloading |
| Structured Generation | JSON Schema-constrained output |
| Prefix Cache | Reuse prompt KV state across requests; fork/release prefixes for dialogue branching |
| Answer Cache | Cache full prompt → completion results with TTL, LRU, and adapter/namespace isolation |
| Matcher | Pluggable embedding / LM-choice / cascading matchers for option selection |
| GPU Scheduling | CiG integration for GPU sharing with renderer |
Available SDKs
- Godot -- GDExtension, GDScript-native signals and async
- Unity -- C# UPM package with async/await and delegates
- Unreal Engine -- Plugin with Blueprints and C++ API
- Python -- Native module with async iterators
- C FFI -- Raw C ABI for custom integrations