ai

AI toolkit for OpenComputers - build AI-powered automation in Minecraft.

Libraries

Library	Description
ai	Vercel-like Core SDK for text generation, streaming, tool calling, and structured output
oc-code	Interactive AI coding agent with file editing, shell access, and TUI

Quick Start: oc-code

oc-code is an AI coding assistant that runs directly in OpenComputers. It can read, write, and edit files, run shell commands, and help you with coding tasks.

Here are some examples of what oc-code can do:

3D_Spinning_Cube.mp4

3D_HilbertCurve.mp4

And here is the demo YouTube video: https://youtu.be/u-gIQOXLimQ

---

Installation

oppm register DonChong2000/oc-ai
oppm install oc-code

Setup

Set your API key. For example, you can get one for free from Google AI Studio:

It supports Vercel, OpenAI, Google for now.

set GOOGLE_GENERATIVE_AI_API_KEY=<your-key>

Run

oc-code

Features

• Interactive TUI - Color-coded interface with scrolling and command autocomplete

  

• Monochorme mode - Works on robots' GPU/screen

• Multi-provider support - Anthropic, OpenAI, Google (direct or via gateway)

• Built-in tools - read_file, write_file, edit_file, list_directory, glob, grep, shell

• Skills system - Extensible command handlers

Commands

Command	Description
/help	Show available commands
/model	View or switch AI model
/clear	Clear conversation history
/exit	Exit oc-code
!<cmd>	Execute shell command directly

Quick Start: AI SDK

The AI SDK provides programmatic access to AI models for your own scripts.

Basic Usage

local ai = require("ai")
local google = require("ai.google")

local result = ai.generateText({
  model = google("gemini-2.5-flash"),
  prompt = "Hello, world!",
})

print(result.text)

Tool Calling

local weatherTool = ai.tool({
  name = "get_weather",
  description = "Get current weather",
  parameters = {
    type = "object",
    properties = {
      location = { type = "string" },
    },
    required = { "location" },
  },
  execute = function(args)
    return { temp = 72, condition = "sunny" }
  end,
})

local result = ai.generateText({
  model = google("gemini-2.5-flash"),
  prompt = "What's the weather in Tokyo?",
  tools = { weatherTool },
  maxSteps = 3,
})

Structured Output

local result = ai.generateObject({
  model = google("gemini-2.5-flash"),
  schema = {
    type = "object",
    properties = {
      name = { type = "string" },
      age = { type = "number" },
    },
  },
  prompt = "Generate a random person.",
})

print(result.object.name, result.object.age)

Installation

OPPM (Recommended)

oppm register DonChong2000/oc-ai
oppm install ai

Manual

Copy lib/ai/, lib/oc-code/, and lib/cmn-utils/ to your /lib/ directory.

Copy bin/oc-code.lua to your /bin/ directory.

Environment Variables

Variable	Provider	Required For
GOOGLE_GENERATIVE_AI_API_KEY	Google	google("model") or /model google <model>
OPENAI_API_KEY	OpenAI	openai("model") or /model openai <model>
AI_GATEWAY_API_KEY	Vercel Gateway	"provider/model" string format

Requirements

• OpenComputers computer with internet card
• API key for your chosen provider
• For TUI: Tier 2+ screen (4-bit color) and GPU
• For terminal mode: Any computer or robot

License

MIT