Summary

Libretto is an AI-powered toolkit for building and maintaining robust browser automations, targeting complex web applications like healthcare software. It equips coding agents with a live browser and a token-efficient CLI for inspecting pages, capturing network traffic to reverse-engineer APIs, and interactively debugging automation scripts. This approach significantly enhances the reliability and maintainability of web integrations.

How It Works

Libretto integrates a live browser environment with LLM-driven "snapshot analysis." When needed, it captures page snapshots (PNG + HTML) and uses an LLM to interpret states, extract selectors, or diagnose failures, minimizing context window usage. It also records network traffic, enabling the generation of direct API call scripts—a faster, more reliable alternative to UI automation.

Quick Start & Requirements

Installation is via npm install libretto. Initial setup, including skill installation, Chromium download, and LLM model configuration, is handled by npx libretto setup, which detects provider credentials (e.g., OPENAI_API_KEY). Workspace readiness can be checked with npx libretto status. Documentation is at libretto.sh/docs; community support is on Discord (discord.gg/NYrG56hVDt).

Highlighted Details

• LLM Snapshot Analysis: Uses LLMs to interpret browser states, extract selectors, and diagnose failures, optimizing AI context.
• API Reverse-Engineering: Captures network traffic to generate efficient, direct API scripts, surpassing UI automation speed and reliability. Includes optional request security analysis.
• Interactive Debugging: Enables autonomous debugging by reproducing failures, pausing execution, inspecting live states, and fixing issues.
• Session & Profile Management: Stores runtime data (logs, network, actions) and allows saving/reusing authenticated browser sessions via profiles.

Maintenance & Community

Maintained by Saffron Health. Community support is available via Discord (discord.gg/NYrG56hVDt) and GitHub Discussions. Bug reports should be filed as GitHub issues.

Licensing & Compatibility

The provided README does not specify the project's license. Verification of licensing terms is recommended for commercial use or integration into closed-source products.

Limitations & Caveats

Snapshot analysis requires configured LLM provider credentials and incurs associated API costs. The README does not detail specific platform limitations or known bugs. The absence of explicit licensing information is a key point for due diligence.