meta-ai-sdk
v0.0.7
Published
A TypeScript SDK for interacting with the Meta AI, providing tools for AI agents, integrations, and utilities to enhance developer productivity.
Downloads
720
Maintainers
Readme
Overview
Meta AI SDK provides a unified TypeScript interface for building, integrating, and automating with Meta AI models. It includes:
- SDK for agent orchestration, tool integration, and context management
- Utilities for safe, context-aware automation
- CLI for advanced developer workflows (coming soon)
Features
- MetaAI Class: Simple interface for interacting with Meta AI models
- Agent Framework: Compose multi-agent systems with planning, tool selection, and feedback loops
- Safe Execution: Built-in guardrails to prevent destructive actions
- Streaming & Context: Stream model outputs and manage working memory
- Extensible Tools: Integrate your own tools, functions, and APIs
- Beautiful Terminal Output: Colorful, structured, and readable logs
Installation
Add to your project:
npm install npm i meta-ai-sdkRequires Node.js 18+ and TypeScript 5+ for SDK usage.
Usage
Basic Usage with MetaAI
import { MetaAI } from '@orello/meta-sdk';
const meta = new MetaAI({
apiKey: 'YOUR_API_KEY',
model: 'meta-llama/Llama-2-7b-chat-hf',
});
async function main() {
const response = await meta.generateContent({
prompt: 'Summarize the latest AI research trends.',
maxTokens: 256,
temperature: 0.7,
});
console.log(response.result);
}
main();Streaming Responses
import { MetaAI } from '@orello/meta-sdk';
const meta = new MetaAI({ apiKey: 'YOUR_API_KEY' });
async function streamExample() {
const stream = await meta.generateContentStream({
prompt: 'Write a poem about the ocean.',
maxTokens: 100,
});
for await (const chunk of stream) {
process.stdout.write(chunk.text);
}
}
streamExample();Agentic Workflow Example
The SDK supports multi-agent orchestration:
- Information Gathering: Collect files, logs, docs, and context
- Planning: Generate a markdown implementation plan and decompose into tasks
- Tool Detection: Select tools/functions and create execution plans
- Execution & Feedback: Run plans, detect errors, and retry with feedback loops
Contributing
Contributions are welcome! Please open an issue or submit a pull request.
License
MIT License
