@gonkalabs/blocks-cli

Command-line interface for Gonka Blocks — package any project as a block, run it on the Gonka inference network, and stream results back to your terminal.

gonkablocks magic           # interactive walkthrough
gonkablocks deploy ./mybot  # ship a block
gonkablocks run alice/translate text="hello"

Install

One-line installer (recommended):

curl -fsSL https://blocks.gonka.gg/install.sh | sh

Or with npm:

npm install -g @gonkalabs/blocks-cli

Then verify:

gonkablocks --version

Two binaries are installed: gonkablocks (full name) and gbk (short alias). They are identical.

Requires Node.js 20 or later.

Quickstart

# 1. Connect once. Opens your browser, mints a long-lived API key,
#    saves it to ~/.gonkablocks/config.json.
gonkablocks connect

# 2. Run any published block.
gonkablocks run admin/qwen-quick text="hello world"

# 3. Or wrap your own script as a Run, with the inference proxy
#    auto-injected into OPENAI_API_KEY / OPENAI_BASE_URL.
gonkablocks exec --title "my agent" -- python my_agent.py

To ship your own block:

gonkablocks init mybot          # scaffold manifest.yaml + Dockerfile + main.py
gonkablocks deploy ./mybot      # upload, build, publish
gonkablocks run me/mybot        # invoke it

Commands

| Command | What it does | | --- | --- | | gonkablocks magic | Interactive setup: connect + scaffold + run/deploy | | gonkablocks connect | Mint a long-lived API key (one-shot login) | | gonkablocks login | Classic browser cookie session | | gonkablocks env | Print shell exports for the active key (source <(gonkablocks env)) | | gonkablocks init [dir] | Scaffold a blank block | | gonkablocks deploy [dir] | Auto-scaffold (if needed) + upload + build + publish | | gonkablocks run a/b [k=v ...] | Run a published block by <author>/<slug> | | gonkablocks runs | List your recent runs | | gonkablocks exec -- <cmd> | Wrap a local subprocess as an observed cloud Run | | gonkablocks proxy [--port N] | Local OpenAI-compatible HTTP forwarder |

Run any command with --help for full options.

How runs are billed

Inference is metered against the live on-chain price the Gonka validator network charges, converted to USD via the GONKA/USD exchange rate. Today that's about $0.00044 per 1M tokens for Qwen3-235B and Kimi K2.6 — roughly 100× cheaper than the big clouds. Your active per-block spend cap (set in manifest.yaml) prevents runaway costs.

Anonymous users get 10 free runs per week without any login.

Documentation

• Guides: https://blocks.gonka.gg/guides
• Block manifest reference: https://blocks.gonka.gg/docs
• Discover blocks & models: https://blocks.gonka.gg/library

License

MIT — see LICENSE.