dspy-ruby

Framework for building type-safe, composable LLM applications in Ruby. Enables predictable AI features via structured signatures, modules, and provider configuration.

• Build type-safe LLM signatures with input/output contracts
• Create composable, reusable modules and workflows
• Configure LLM providers (OpenAI, Anthropic, Gemini, Ollama)
• Implement agents with tools (ReAct, ChainOfThought, CodeAct)
• Support for vision and multimodal tasks

• Type-safe signatures with runtime type checking
• Composable module chains for complex workflows
• Multiple predictor types (Predict, ChainOfThought, ReAct, CodeAct)
• Support for multiple LLM providers (OpenAI, Anthropic, Gemini, Ollama)
• Vision support with image processing
• Testable, version-controlled, and deployable AI features
• Built-in error handling, retry, and caching
• Cost optimization strategies by provider and use case

1. 1.Define LLM tasks using type-safe signatures with clear inputs/outputs
2. 2.Chain modules to build reusable, composable workflows
3. 3.Configure LLM providers via environment or code configuration
4. 4.Choose appropriate predictor types (Predict, ChainOfThought, ReAct, CodeAct)
5. 5.Use vision support with DSPy::Image for image-text tasks
6. 6.Test and optimize with structured outputs and observability tools