Active AgentBuild AI in Rails

Now Agents are Controllers Makes code tons of fun!

Agents

Controllers for AI. Define actions, use callbacks, render views. Rails conventions for LLM interactions.

AI calls Ruby methods to fetch data and make decisions. Works like RPC for agents.

Extract typed data with JSON schemas. Validated responses for forms and APIs.

OpenAI, Anthropic, Ollama, OpenRouter. Switch providers with one line of code.

Real-time response streaming with callbacks for dynamic UIs and live updates.

Generate vectors for semantic search, RAG, and clustering applications.

Test with fixtures and VCR cassettes. Mock providers for fast tests.

Check what your agents answer and which tools they call, as Minitest or RSpec tests, and publish each run to the dashboard.

Add the solid_agent gem for database-backed conversations, long-term memory, runs and cost — as models you own.

Add the actionagent gem and mount its engine for traces, metrics, the agent builder, conversations and evaluations.

Run evaluations and gate CI from a terminal; move runs, scenario catalogs and traces between dashboards as JSON.

Process generations async with Active Job. Scale AI operations in the background.

Automatic retries with exponential backoff. Graceful degradation for production.