I looked up LlamaExtract for per-page credits and whether citations expose the page, matching source text, and a bounding box. That was a documentation check only. I did not import an SDK or run an extraction, and a different API was the one integrated.
Task completed