aor 0.2.0

AI-on-Rails: A Python library for prototyping and publishing AI applications with guardrails.

AI-on-Rails

Rapidly create AI applications and publish to the AI Agent marketplace and AI-to-AI broker at https://aionrails.com.

Intro

AI-on-Rails helps bootstrapping AI agents that are triggered by external request and brokers such requests.

An AI-on-Rails application is one or more agents that accept external requests. Such agents can be new AI Agents deployed using AI-on-Rails (on your own infrastructure). But AI-on-Rails agents can also be AI agents that already exist somewhere but follow a known protocol convention (e.g. A2A).

Usage

Creating a New AI Agent

Create a new application with a basic conversation agent:

# Install the AI-on-Rails command line tools
pip install aionrails

# Create an application
aor new my-application
cd my-application

# Create a new AI Agent using a template
aor add my-conversation-agent --template langgraph/basic_conversation

# List available templates
aor templates list

# Get details about a specific template
aor templates info langgraph/basic_conversation

Deploying to Cloud Services

Deploy your AI agent to AWS Lambda:

# Add AWS Lambda deployment configuration
aor add my-agent --langgraph --aws-lambda

# Deploy to AWS using the default CLI profile
aor deploy

# Publish
aor publish --token "<your-personal-token-obtained-on-aionrails.com>"

Adding Existing Services

Integrate existing A2A-compatible services:

# Add an existing A2A service agent
aor add --existing --a2a "https://<my-a2a-domain>/<uri>" my-agent

# Publish to AI-on-Rails
aor publish --token "<your-personal-token-obtained-on-aionrails.com>"

Examples

While you would expect aor to generate the necessary service definition files automatically, several examples can be found in the examples folder to better understand the syntax for advanced users.

• X post using LangGraph
• Existing AI Agent via A2A

F.A.Q.

What happens when the peer agent changes its functionality or the protocol (e.g. inputs/outputs)?

Applications are versioned and more than one version can be published at the same time. When integrating with an agent published on AI-on-Rails, it is recommended to pin a specific version to continue using it even when newer revisions are getting published.