Documentation
How Engineering AI works
When you send a message to the Engineering AI assistant, four things happen in sequence. The point of this page is to make those steps visible so you can see why the assistant behaves differently from a general chatbot.
1. The assistant reads your request
It interprets what you are asking for, including units, dimensions, materials, and the type of answer you expect. A short question like “density of air at 25C” is treated differently from a multi part request like “size a pump for 300 gpm at 80 feet of head, then check voltage drop on the motor feeder.”
2. It decides which tools it needs
The assistant has access to two kinds of tools: the reference database of 375 engineering tables, and the calculator suite of 50+ deterministic tools. For a typical request it picks one or more of these rather than answering from general knowledge. A lookup question uses the database. A sizing or checking question uses a calculator, which in turn uses the database to fetch its reference data.
3. It runs the tools and receives real values
Database lookups return rows from reference tables. Calculator runs return numerical results along with the inputs and key intermediate values. The assistant sees the raw results the same way you would if you ran the tool yourself from the database or calculators page.
4. It explains the answer
The response is streamed back in real time. It includes the final value, the tool or table that produced it, and a short explanation of the steps taken. When relevant, the assistant lists the assumptions it used and the units it picked.
Why this matters
A general AI assistant will happily state that a 10 inch schedule 40 pipe has a particular wall thickness, even if it is wrong. The Engineering AI assistant reads the actual value from the pipe dimension table. The answer is the same number you would get from the published reference.
The same principle applies to everything else the assistant does. It does not do engineering math in its head. It calls the tool that does the math, then reports the result.
What you see on your side
While a turn is running you see the assistant stream text, call out which tools it is using, and report results inline. If a run takes more than a moment the streaming output makes it clear the assistant is still working rather than waiting silently. When the turn finishes you see the final answer plus the credit cost for that turn.