Home

Agent-o-rama is a library for building scalable and stateful LLM agents in Java or Clojure. Agents are defined as simple graphs of pure Java or Clojure functions, and Agent-o-rama automatically captures detailed traces and provides facilities and a web UI for offline experimentation, online evaluation, time-series telemetry (e.g. latencies, token usage), and much more.

Documentation

• Quickstart - Get up and running in minutes with an agent with an agent running locally or on a Rama cluster
• Programming Agents Guide - Comprehensive guide to programming and deploying agents
• Client API - How to invoke and interact with agents from your applications
• Interfacing with LLMs - How to use the capabilities LLMs within agents
• Tools - Creating tools agents for AI model function calling
• Streaming - How to create and consume streams from agents
• Human-in-the-loop - Human-in-the-loop patterns for agent interactions
• Datasets, evaluators, and experiments - Evaluate performance of agents and nodes
• Human feedback – Complement automated feedback with structured human feedback
• Actions, online evaluation, and telemetry - Automate actions on production runs and view comprehensive analytics
• Integrating with regular Rama modules – How to add agents to regular Rama modules and use depots, PStates, and query topologies from agent nodes

API docs

• Javadoc
• Clojuredoc

Resources

• Mailing List: Rama mailing list
• Community Chat: Discord server
• Chatbot: https://chat.redplanetlabs.com/
• Examples: Explore the examples/ directory for lots of example agents
• Rama Documentation: redplanetlabs.com/docs