Aeo Infra

Use when the user asks about MLX serving, mlx_lm.server, oMLX, Apple Silicon LLM serving, or running a local LLM on a Mac, and when troubleshooting a model that fails to load, OOM during load or inference, a server that hangs or crashes at batch>1, tool calls returned as plaintext content, a throughput regression, or the choice between mlx-lm and oMLX. Also covers oMLX feature-flag tuning (turboquant_kv, dflash, MTP, specprefill, thinking_budget, max-concurrent-requests, force_sampling), the OptiQ proxy for models exceeding RAM, Llama-4 ChunkedKVCache batch handling, Llama-3 tool-call JSON format, and bench-driven validation of serving configs. Apple Silicon M-series only, not cloud LLM hosting, non-MLX backends such as llama.cpp, Ollama and vLLM, or model training.

Use when the user asks about the GL-iNet Slate 7 or GL-BE3600 travel router (Wi-Fi 7) - first connection and setup, admin panel access, VPN client or server configuration with WireGuard or OpenVPN (NordVPN, Mullvad, Surfshark and 30+ providers), AdGuard Home, Tor, Tailscale or ZeroTier, DDNS, network modes, SSH and CLI access, lost admin access, factory reset, firmware update, or U-Boot recovery of a bricked device. Also use when scripting the device through its JSON-RPC admin API at /rpc, provisioning the built-in WireGuard server without the UI (wg-server, add_peer, generate_peer), or configuring a Linux WireGuard client with an overlay VPN stacked on top of it. Applies to most GL-iNet sdk4 firmware devices, not to generic router questions.

Use when the user asks about Azure CLI commands, writing Azure CLI scripts, managing Azure resources, az login, az account, service principals, managed identity, or the credential chain; when driving VMs, Storage, VNets, NSGs, Key Vault, App Service, AKS, or RBAC from the command line; or when the question is JMESPath --query syntax, az graph query, CLI pagination, az wait, output formatting, multi-subscription work, resource locks, safe destructive commands, az rest API calls, az bicep, CLI extensions, cost management, cost analysis and spending, or az resource for services without a dedicated command. Covers the Azure CLI only, not the Azure SDKs or portal walkthroughs, and not AWS, which belongs to aws-cli-operations.

Use when the user asks about AWS CLI commands, writing AWS CLI scripts, querying AWS resources, managing AWS credentials, aws login, aws configure sso, assume role, or the credential chain; when driving S3, EC2, IAM, Lambda, VPC, RDS, DynamoDB, CloudWatch, or SSM from the command line; or when the question is JMESPath query syntax, --query versus --filters, CLI pagination, an aws cli waiter, output formatting, cross-account access, safe destructive commands, an S3 sync dry run, the aws cli pager blocking a script, or migrating from AWS CLI v1 to v2. Covers the AWS CLI v2 command surface only, not the AWS SDKs, not CloudFormation or Terraform authoring, and not Azure, which belongs to az-cli-operations.

Use when the user asks about Lima, limactl, lima.yaml, lima start, or lima shell; wants to install Lima, configure or edit a Lima VM, create a Linux VM on a Mac, spin up an Ubuntu VM, run Docker on macOS, or run Linux on Apple Silicon. Also use for lima vmType vz, vz versus qemu, host.lima.internal, socket_vmnet, shared and bridged lima networking, virtiofs and 9p mounts, writable mounts, lima port forward, limactl edit, validate and template, Rosetta for running x86 binaries, and debugging a VM that will not start. Applies whenever Lima is mentioned even without an explicit request for help, because the vz-versus-qemu, mount-type, and network-mode choices are easy to get wrong. Apple Silicon Macs only.

Use when adding a macOS host (macOS 26 Tahoe and later) to a self-hosted headscale mesh as a Tailscale client, or when Tailscale on macOS misbehaves: "failed to connect to local tailscale service", Tailscale.app stuck on "Starting...", `tailscale up` hanging on "joining", a blank menu-bar icon after install, or the Homebrew cask and formula fighting over the daemon. Covers the Homebrew Cask install, the NetworkExtension permission grants, `tailscale up --login-server` with a headscale preauth key, the deep-link fallback when the CLI cannot reach the daemon, headscale's numeric `--user` id pitfall, and reach verification. Not for official Tailscale onboarding, Linux or Windows clients, or headscale server install.