Generative AI SDK for Lua

A developer-friendly Lua interface for working with various generative AI providers, abstracting away provider-specific payload structures and response parsing so that using multiple models is easy.

Features

⚠️ This is a work in progress so any help is highly appreciated!

• Easily switch between AI chat model providers
• Pass in prompts and get replies without the provider complexity
• Easily integrate new models and adjust settings
• Use the chat object for integrated message history
• For more granular control use the genai client directly if needed
• Stream output for real-time responses
• Structured JSON response abstraction layer
• Token usage tracking with cost calculation

Providers

• OpenAI

• Anthropic

Roadmap

1. Advanced error handling
2. Google Gemini integration
3. Audio models
4. Image models
5. Open-Source model integration
6. Video models

Installation

luarocks install lua-genai

Dependencies

• lua-cjson

• luasec

Usage

local genai = require("genai")

local client = genai.new("<YOUR_API_KEY>", "https://api.openai.com/v1/chat/completions")

local chat = client:chat("gpt-4o-mini")
print(chat:say("Hello, world!"))

System Prompt

local chat = client:chat("gpt-4o-mini", { system_prompt = "You are a fish." })
print(chat:say("What are you?"))

Streaming

local process_stream = function(text)
	io.write(text)
	io.flush()
end

local chat = client:chat("gpt-4o-mini", { settings = { stream = process_stream } })
chat:say("Tell me a very short story.")
print()

JSON Response

local npc_schema = {
	name = { type = "string" },
	class = { type = "string" },
	level = { type = "integer" },
}

local json_object = {
	title = "NPC",
	description = "A non-player character's attributes.",
	schema = npc_schema,
}

local chat = client:chat("gpt-4o-mini", { settings = { json = json_object } })
print(chat:say("Create a powerful wizard called Torben."))

See example.lua for a full-featured Anthropic implementation.