ext-infer: PHP 8.3+ extension for native LLM inference and embeddings via llama.cpp

ext-infer is a PHP 8.3+ extension written in Rust that loads a GGUF model and runs LLM inference inside the PHP process via llama.cpp, enabling semantic search, RAG pipelines, and worker inference without calling Python or remote APIs.

The extension provides a fluent Prompt builder, a Response that separates reasoning from answer, and an Embedding class that handles normalization and cosine similarity—designed to feel native to PHP like the intl or pdo extensions.

In-process inference reduces latency (bounded only by decode time versus milliseconds or tens of milliseconds for subprocess or HTTP calls) and eliminates the need for a separate Python sidecar, daemon, or inference server to manage alongside PHP-FPM.