Offline AI sidecar for games: local tiny LLM (llama.cpp Vulkan, Qwen3-4B), streaming STT (sherpa-onnx Nemotron) and token-streaming TTS (Kyutai Pocket TTS) behind an OpenAI-compatible localhost API. Any engine, any GPU vendor, no cloud tokens.
rust text-to-speech offline-first vulkan game-development speech-to-text sidecar streaming-tts llama-cpp local-llm game-middleware sherpa-onnx openai-compatible qwen3 npc-ai
-
Updated
Jul 20, 2026 - Rust