Terrarium vs TUI-use

Terrarium is a multi-turn evaluation and optimization engine for LLM agents built by evolvent-ai. Unlike static benchmark suites that measure agents against fixed input-output pairs, Terrarium creates persistent, stateful "living environments" — simulated deployment contexts where agents operate over extended sessions, accumulate state, use tools, and interact with simulated external systems. You evaluate agents the way you'd test a car: by driving it, not by measuring its doors. The system supports configurable environment complexity, including simulated databases, APIs, file systems, and user personas. Agents are scored not just on final outputs but on trajectory quality — how efficiently they reached the answer, how often they hallucinated intermediate steps, and how well they recovered from dead ends. The engine also supports continuous optimization loops where poor-performing trajectories trigger automatic prompt refinement. With 17 stars and created April 14, Terrarium is extremely new. But it's addressing a genuine gap: the disconnect between how agents perform on static benchmarks versus how they behave in production. As enterprise AI deployments scale, the need for realistic pre-production evaluation is becoming critical.

TUI-use is an open-source library that gives AI agents the ability to interact with traditional interactive terminal (TUI) applications — think vim, htop, ssh sessions, database CLIs, and legacy text-based UIs that were never designed for programmatic control. Instead of requiring a GUI or a REST API, TUI-use interprets terminal output as structured state and sends synthetic keystrokes back, enabling agents to "see" and "drive" any TUI application as if they were a human at a keyboard. The project was born from a real pain point: AI coding agents can call bash commands and write files, but they fail badly the moment a tool opens an interactive prompt waiting for user input. TUI-use solves this by building a state machine layer over PTY (pseudo-terminal) interfaces, letting agents read the current screen buffer, detect interactive prompts, and respond intelligently. It ships with adapters for common TUI patterns and a clean API that works with any LLM tool-use framework. The Show HN post attracted genuine interest from the ops and DevOps community — many existing workflows depend on tools that expose only an interactive terminal interface. TUI-use fills a real gap in the "AI agents that control computers" space by handling the long tail of CLI programs that have no API, no GUI, and no intention of ever getting one.