Retrieving the non-metadata titles of 45,000 various PDF, docx, etc. without a bunch of rules/regexs that would fail half the time. “Derp derp, hallucinations”

Retrieving the non-metadata titles of 45,000 various PDF, docx, etc. without a bunch of rules/regexs that would fail half the time. “Derp derp, hallucinations”. Eh, no, not in practice, not when the entire context and document is provided and the tools are used correctly.