The Ruby AI framework. Build AI features the Ruby way, with conventions that feel at home in Rails: switch models without rewriting your code, then grow from a single call to production agents that run code in a sandbox, follow skills, and resume across jobs. Features chat (text, images, audio, PDFs), tools, MCP, structured output, evaluations, embeddings, image, audio, and video generation, Rails integration, and streaming. Works with OpenAI, Azure, Anthropic, Google Gemini, Vertex AI, AWS Bedrock, xAI, Cohere, DeepSeek, Mistral, Perplexity, OpenRouter, ElevenLabs, Deepgram, Ollama and GPUStack (local models), and any OpenAI-compatible API. Plain Ruby with a small set of runtime dependencies.

Forwards RubyLLM::Schema to Schematist::Schema. Depend on schematist instead.

Wraps RubyLLM::Chat with input/output contracts, business-rule validation, retry with model escalation on validation failure, pre-flight cost ceilings, and an evaluation framework. Sibling abstraction to RubyLLM::Agent — same niche (reusable class-based prompts), wider contract.

Define a safe component catalog, expose it as a RubyLLM tool schema, validate model-generated component trees, and render them with Rails partials, ViewComponent, or JSON.

A light agent framework on top of RubyLLM. An agent is an object: its fields are state, its methods are what the model can call, and the methods it declares without a body are written by the model at runtime — the method name and prompt are the specification, the schema is the contract. The model acts by writing Ruby that runs on the agent itself.

Phronomy is a Ruby AI agent framework that provides composable building blocks — Agents, Workflows, Tools, Filters, and Tracing — for building AI agents in Ruby. Powered by RubyLLM for LLM abstraction.

Declaratively define multiple modes for a RubyLLM chat and automatically route each user message to the appropriate mode. Each mode has its own instructions, tools, model, and reasoning effort, while all modes share the conversation history. Configure mode availability and fallbacks, and inspect or test routing decisions independently of the response.

Squishling makes Ruby classes elastic. Add a squishling to a class and it can send inputs straight to an LLM (OpenAI, Anthropic Claude, Google Gemini, or any provider RubyLLM supports, or any OpenAI-compatible endpoint, or your own adapter) and return strict-schema-validated, typed results: the same type your Ruby code returns. Use it to gracefully maintain fickle integrations, build interfaces for unknown data formats, and rescue production errors by handing the failing call to the LLM with the source class as context. Write Ruby only for the paths where token costs justify maintaining it. Per-input routing, validated retries, and fallbacks for your LLM calls included.