/// documentation
Docs
How Wide Area Intelligence works, and how to run it.
Plain-English definitions and pronunciation guides for local-AI terms like llama.cpp, Ollama, LM Studio, GGUF, LoRA, RAG, and KV cache.
Run Claude Code through WAI's Anthropic Messages-compatible gateway: local GPU routing, Claude-shaped aliases, model discovery, token counting, and cloud failover.
Run reusable agent workflows with external-user isolation, persistent local browser profiles, login handoff, schedules, and exports.
How the gateway decides where a request runs — edge cache, your own GPUs, then capability-aware cloud failover. Substitution mode, vision models, cross-provider backups.
Mix a fast, cheap model and a strong one per step — classify-then-generate, extract-then-write — over one endpoint. Run the cheap step on your own GPUs for free.
Generate text-to-video on your own GPU (Wan 2.2, LTX-2.3): deploy a video model, smoke-test it with wai test-video, and use it from Chat, the playground, and the node API.
Every command for the node CLI: status, logs, start/stop/restart, update, models-dir, uninstall, plus config and auto-update.