Hi! Datasheets can definitely be inconsistent, and that’s one of the tougher parts of doing this well. LLMs are very much “garbage in, garbage out,” which is exactly why the tool doesn’t search the web or pull from any sort of automatic datasheet library. It only reasons from the netlist and the PDFs you upload, so you stay in full control of the context and the primary sources it can pull from. If the datasheet is clear, the results are usually very solid; if the datasheet is vague, the model reflects that instead of pretending otherwise.

I’d really recommend trying it with one of your designs: upload the netlist + a component’s datasheet and ask a specific question about the part in the design. It’s the easiest way to see how well the ingestion works in practice. Would love to hear your feedback after you try it!