agentsop-llm-engine-selection

What problem does it solve?

This Skill helps coder-agents make robust, hardware-aware decisions when selecting or migrating an LLM inference serving stack, avoiding “fastest” myths and benchmark misuse.

Core Features & Use Cases

• Cross-engine decision rubric: Chooses between vLLM, SGLang, TensorRT-LLM, TGI, llama.cpp, Ollama, and MLX using the constraint vector (hardware × workload × constraint).
• Activation for real production tradeoffs: Supports picking a runtime for new projects, defending an existing stack, planning migrations, designing mixed multi-tier deployments, and auditing questionable benchmark-based recommendations.
• Guardrails and evidence mapping: Provides hard filters, a structured decision workflow, dilemma cases, anti-patterns, and a cited-source evidence approach for performance and topology claims.

Quick Start

Ask an AI coder-agent to use the llm-engine-selection skill to recommend the best serving engine for your specific hardware and workload, and have it justify the choice using the (hardware × workload × constraint) model.