zkai-cli 0.5.0

ZKai provider CLI — setup, register, and manage your verifiable AI inference node on 0G chain

zkai-cli

Provider node management CLI for ZKai — the verifiable AI inference marketplace on 0G chain.

This package installs the zkai command, which automates everything a provider needs: wallet setup, container lifecycle (Docker Compose), and on-chain registration against the ZKai contracts deployed on 0G mainnet.

Install

pip install zkai-cli

Quick start

Becoming a provider takes 4 commands:

# 1. Initial setup — prompts for an EVM private key, writes .env
zkai init

# 2. Start the provider (bridge + TDX-attested enclave)
zkai start

# 3. Verify it's running
zkai status

# 4. Register on-chain — your wallet becomes your provider ID
zkai register --model qwen2.5:1.5b --price 100

Your node is now discoverable by ZKai consumers. Every inference you serve pays you native 0G straight to your wallet.

Commands

Command	Purpose
zkai init	First-time setup wizard (wallet, .env, relay config).
zkai start	Bring up the provider containers via docker compose.
zkai stop	Stop and remove containers.
zkai restart [service]	Restart all or one of: bridge, enclave.
zkai status	Show container health, wallet sync, enclave state, dashboard URL.
zkai logs [service]	Tail logs from a service.
zkai register	Register the provider in the on-chain ProviderRegistry.
zkai deregister	Remove the provider from the on-chain registry.
zkai info	Print local provider metadata.
zkai keys	Print API key management info.

Run zkai --help for the full reference.

Requirements

• Linux machine with Docker Engine + Compose plugin (Docker Desktop on macOS works too).
• An EVM-compatible wallet with a small amount of native 0G for gas. Get testnet 0G from faucet.0g.ai.
• A model you want to serve (qwen2.5:1.5b is the default; any Ollama model works).

How it fits together

ZKai providers run two containers:

• bridge — Node.js sidecar that wraps ethers.js and talks to 0G chain on your behalf. Registers, attests, and accepts payment.
• enclave — Python/FastAPI service running inside a Gramine TDX-sealed runtime, serving an OpenAI-compatible inference endpoint.

The CLI hides all of this behind a few commands. Edit provider/.env if you need to override the model, gateway URL, or relay endpoint.

Links

• Repository — https://github.com/skyyycodes/zkai-eth
• Dashboard — https://zkai-ether-og.vercel.app
• 0G chain explorer — https://chainscan.0g.ai

License

MIT