LocalAI — локальный сервер моделей с OpenAI-совместимым API

LocalAI is a self-hosted engine for running models on your own hardware — LLMs plus voice, vision, and image models — with no data sent outside. Reach for it as a private replacement for cloud APIs: stand up text generation, speech recognition and synthesis, image generation, and embeddings locally, and call them through an OpenAI-compatible API so existing apps switch over with almost no changes. Its small-core architecture plugs in backends — llama.cpp, vLLM, whisper.cpp, stable-diffusion, MLX, and others — as modules for the model and hardware you need, and it works without a GPU too. Its focus is a single local server for many model types behind an OpenAI-compatible API, not a one-click desktop chat (LM Studio/Ollama fill that) or a cloud service: it is a server component you deploy and integrate into your own services.