Team 34

ChemBee

ChemBee is a voice-and-chat agent that runs process engineering and equipment buying for a chemical plant, and it leaves a full audit trail from live plant data and simulation to a drafted purchase order.

Check your own project

Community votes recognize the build and do not affect official judging.

Review the project

Start with the source code, then open the demo or video if available.

View GitHub repository
Project description
ChemBee is an AI agent for chemical plant engineers that figures out what equipment the plant needs and helps buy it, with every decision traceable. Here's an example. An engineer asks, by voice or chat, "Can we run Unit 100 at 14 gpm?" ChemBee simulates the unit at that rate and finds that the current reboiler can't supply enough heat. It looks up replacement reboilers from suppliers and simulates each one in the plant. It compares them side by side and recommends one. Then it drafts a purchase order, plus a PDF report with plots. Nothing is sent without the engineer's approval. Every answer is grounded in real data. Live sensor readings come from SingleStore. Plant history and documents, like the 2024 memo explaining why the rate was cut, come from a Cognee knowledge graph. That memo opens as a PDF you can check yourself. When an engineer corrects ChemBee, Firedrill re-tests the conversation with the correction. If the corrected version passes, the lesson is saved so ChemBee doesn't repeat the mistake. That's the audit trail: every purchase links back to the data, the simulation and the documents behind it.
Tools used