Roast is a convention-oriented framework for building structured AI workflows, designed for developers and researchers who need to orchestrate complex AI tasks. It simplifies the creation of AI-driven applications by providing a declarative, YAML-based approach to define workflows, manage prompts, and integrate custom logic, ultimately accelerating AI development cycles.

How It Works

Roast utilizes a convention-over-configuration philosophy, allowing users to define workflows and steps using YAML files and Markdown-formatted prompts with ERB support. It maintains a shared conversation context across steps, enabling natural data flow. For more explicit data handling, an output hash stores step results, and interpolation syntax ({{ }}) injects values into prompts, commands, or step names. Custom logic can be implemented in Ruby, extending the framework's capabilities.

Quick Start & Requirements

• Install via RubyGems: gem install roast-ai or add gem 'roast-ai' to your Gemfile.
• Run workflows using: roast execute workflow.yml
• Requires Ruby.

Highlighted Details

• Declarative YAML-based workflow definition.
• Supports parallel execution of steps.
• Session replay for debugging and development iteration.
• Built-in tools for file operations, search, and command execution.
• Extensible with custom Ruby tools and initializers.

Maintenance & Community

Maintained by the Augmented Engineering team at Shopify.

Licensing & Compatibility

MIT License. Permissive for commercial use and integration with closed-source projects.

Limitations & Caveats

Session data is not automatically cleaned up, requiring manual management of old session directories. The framework relies on Ruby, which may be a barrier for non-Ruby developers.