Don't trust LLM math with your rosbags? Bagel has a solution

Hi ROS Community,

Couple weeks ago, we launched an open source tool called Bagel. It allows you to troubleshoot rosbags using natural language, just like how you use ChatGPT.

We released a new feature today to address a common concern raised by the ROS community.

:speaking_head: “We don’t trust LLM math.”:man_gesturing_no:

That’s feedback we kept hearing from the ROS community regarding Bagel.:bagel:

:ear: We heard you, so we built a solution: whenever Bagel needs to do a mathematical calculation on a robotics log, it now creates a DuckDB table and runs a clear SQL query against the data.

:brain: That makes Bagel’s math:

• Transparent – you see the query and verify it manually.

• Deterministic – no “LLM guesswork.”

We’re launching this today and would love community feedback :rocket: . Love it :heart: or hate it :-1: ? Let us know! Or yell at us in our Discord :rofl:.

https://github.com/Extelligence-ai/bagel

9 Likes

Love it.

If I just wanted the ROS2 → DuckDB table conversion only, is it possible to do that?

1 Like

Yes, Bagel can do it. It can cast any topics to DuckDB tables. You can check our tutorial here: bagel/doc/tutorials/pipelines/0_basics.ipynb at stage · Extelligence-ai/bagel · GitHub