SayGM

Install Phala CLI and authenticate with a funded Phala Cloud account (create at https://cloud.phala.network)

npm i -g phala && phala auth login

Install gmcli binary from release (placed in ~/.cargo/bin)

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/taostat/gm-miner/releases/latest/download/gmcli-installer.sh | sh

Register serving hotkey with gm registry (provide ss58 directly or use btcli assisted flow with --wallet and --hotkey)

gmcli register-hotkey --hotkey-ss58 <ss58_address>

Authenticate with Taostats via device-code OAuth (credentials saved to ~/.gmcli/config.json)

gmcli login

Set provider API keys for upstreams you intend to serve (Anthropic, OpenAI, Google, Chutes, Z.ai, Moonshot, DeepInfra, KubeTEE, Engy, Moonmath, NEAR). Keys are encrypted in TEE, never seen by gm. Supports up to 8 semicolon-separated keys per provider for slot-based routing.

gmcli set-api-keys --anthropic sk-ant-... --openai sk-... --google AIza...

Run preflight checklist to verify network, login, keys, Phala CLI, and hotkey before deploying(optional)

gmcli doctor

Deploy miner: creates Phala CVM, verifies image hashes, registers worker. Takes a few minutes. Prints worker_id and app_id on success.

gmcli deploy

Declare which models you serve and at what discount off retail. Discount determines payout (10% discount = keep 90% of revenue). Can filter by provider with --provider flag.

gmcli declare-products --discount-pct 5

Check registration state and per-product eligibility and rates(optional)

gmcli status

Monitor lifetime served earnings and recent finalized epochs from registry(optional)

gmcli earnings