Bagel is a natural language interface for robotics and drone data, enabling users to troubleshoot physical systems using conversational AI. It targets engineers and researchers working with robotics data, offering a streamlined way to query and analyze logs without extensive manual parsing or specialized tools. The primary benefit is faster debugging and insights through intuitive, ChatGPT-like interactions with complex datasets.

How It Works

Bagel processes various robotics log formats (ROS 2, ROS 1, PX4, ArduPilot) by converting them into deterministic DuckDB SQL queries. This approach addresses the unreliability of direct LLM mathematical reasoning by leveraging SQL for precise data manipulation. The system then executes these queries against the log data, providing accurate, verifiable results in response to natural language prompts.

Quick Start & Requirements

• Installation: Use uv sync --group ros2 for PyPI packages, or run via Docker.
• Prerequisites: ROS2 Kilted and Claude Code are used in examples. Requires Docker for containerized execution.
• Setup: Local setup involves sourcing ROS2 dependencies and installing Python packages. Docker setup requires mounting local data volumes.
• Docs/Demos: Tutorials available for Claude Code, PX4 ULog, Gemini CLI, ROS2 Bag, and ArduPilot Dataflash.

Highlighted Details

• Supports multiple robotics data formats including ROS 2 (.mcap, .db3), ROS 1 (.bag), PX4 (.ulg), and ArduPilot (.bin).
• Integrates with LLMs via a Message Communication Protocol (MCP) server, allowing flexibility beyond specific models like Claude or Gemini.
• Roadmap includes Computer Vision modules, Video Language Models, anomaly detection, and support for additional robotics formats and LLMs.

Maintenance & Community

• Active community engagement encouraged via a Discord server for questions, feedback, and feature requests.
• Regular feature releases are planned, with updates to the README reflecting shipped features.

Licensing & Compatibility

• The specific license is not explicitly stated in the README. Compatibility for commercial use or closed-source linking is not detailed.

Limitations & Caveats

• The README does not specify the exact license, which may impact commercial adoption. Support for certain data formats or LLMs might still be in beta or under development as indicated by the roadmap.