llama.cpp hits 1.0 with stable API

Adopting SemVer transforms llama.cpp from a fast-moving experimental library into an enterprise-grade inference runtime. Downstream runtimes and package maintainers finally get a predictable foundation without breaking changes on every git pull.

• –Breaking changes are now strictly quarantined behind major version bumps, bringing stability to dependent ecosystems like Ollama, LocalAI, and system package managers.
• –The release purges 35 legacy C APIs and promotes NextN speculative decoding directly into the primary header.
• –Production-hardening llama-server against SSRF, CORS, and DNS rebinding elevates it from a local hobbyist tool to a viable edge deployment target.
• –Supporting over 150 architectures alongside host-memory MoE caching cements llama.cpp as the de facto engine for local AI.