AIA is a powerful CLI console application that brings multi-model AI capabilities to your command line, supporting 20+ providers including OpenAI, Anthropic, and Google. Built on robot_lab for robust robot orchestration, AIA v2 provides a thin CLI shell over a rich execution engine. Run multiple AI models simultaneously for comparison, get consensus responses from collaborative AI teams, or compare individual outputs side-by-side. With dynamic prompt management, embedded directives, shell and Ruby integration, interactive chats, and comprehensive history tracking, AIA transforms how you interact with AI.

Exposes RobotLab robots and networks as A2A agents over HTTP+SSE. Implements the Agent2Agent Protocol v1.0 (Linux Foundation) via the simple_a2a gem. Supports sync and interactive execution modes, bridging RobotLab's AskUser tool to A2A's input_required/resume lifecycle for multi-turn flows without a terminal dependency.

Appends structured records for network runs, robot invocations, and tool calls to a SQLite database for post-mortem analysis.

Adds a human peer worker to RobotLab networks: same bus and memory, receiving and issuing tasks alongside robots.

Allows RobotLab robots to advertise themselves and discover each other on a local network using multicast DNS (mDNS/Bonjour/Avahi) via the zeroconf gem.

Provides RobotLab::DocumentStore — a thread-safe, in-memory semantic search store backed by fastembed (BAAI/bge-small-en-v1.5). Store text documents by key and retrieve the closest matches to a natural-language query using cosine similarity. Works standalone or as a drop-in extension for robot_lab agents and networks.

Provides RobotLab::Durable — an adapter between robot_lab and the HTM gem (Hierarchical Temporal Memory). Replaces the former YAML store with PostgreSQL-backed long-term memory featuring vector embeddings, hybrid semantic search, hierarchical tagging, and multi-robot federation. Includes RecallKnowledge and RecordKnowledge tools that integrate directly into Robot when robot_lab is present.

Provides RobotLab::RactorWorkerPool — a pool of Ruby Ractor workers for executing CPU-bound, Ractor-safe tools in parallel. Includes RactorBoundary (deep-freeze utility for crossing Ractor boundaries), RactorJob/RactorJobError/RobotSpec (frozen data carriers), RactorMemoryProxy (shared Memory access from inside Ractors), and RactorNetworkScheduler (dependency-ordered pipeline across Ractors).

Provides Rails Engine, Railtie, ActiveJob base class with Turbo Stream support, persistence models, and generators for using robot_lab in Rails applications.

Deny-by-default, capability-scoped execution for AgentSkills scripts. On macOS, wraps scripts in a generated sandbox-exec (Seatbelt) profile derived from the skill's declared capabilities intersected with a configured ceiling; robot_lab core runs unconfined until this gem is required.

robot_lab-to ("takeover") runs a robot_lab Robot in an autonomous loop, committing one focused change per iteration toward a stated objective. Each iteration the robot reads shared notes (notes.md), does work, then calls submit_iteration_result to report success or failure. The orchestrator commits good iterations, rolls back failures, and appends to the notes log for cross-iteration memory. Stop conditions include max iterations, max tokens, consecutive failure threshold, and a natural- language --stop-when condition. Runs overnight; review the branch in the morning.

A Rails-free Sinatra + HTMX + SSE web interface for robot_lab. Register robots, chat with them from the browser, and watch each lifecycle event (tool calls, results, errors) stream in real time via robot_lab's hook system.

A turn-based game built on the RobotLab framework: each robot occupies a square on a 2D grid, all robots move simultaneously each turn, conflicts are resolved between turns, and the last robot remaining wins.

under development