Enables querying of robot MCAP recordings with SQL directly through Large Language Models.
This tool revolutionizes how users interact with robot data by allowing them to query MCAP recordings using natural language via an LLM. It eliminates the need for manual data processing or complex database setups, integrating SQL querying capabilities directly into the LLM's conversational interface. Users can ask questions about their robot's performance, sensor data, and session logs, receiving immediate insights without writing a single line of code or managing infrastructure.
Key Features
01LLM-native SQL querying of MCAP files
02No-install, no-database setup with `uvx`
03High-performance metadata retrieval (<1ms)
04Supports full DuckDB SQL, including ASOF JOIN
05Efficient data loading with filtering capabilities
062 GitHub stars
Use Cases
01Analyzing robot sensor data for anomalies or trends
02Correlating data from multiple robot sensors
03Comparing performance metrics across various robot operational sessions
